Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                

Jun 97

Download as pdf or txt
Download as pdf or txt
You are on page 1of 349

BOS Fluids Version 4.111 September 25, 2003 www.paulin.

com

Table of Contents
BOS Fluids Version 4.111
Released September 25, 2003

Chapter 1 Section 1 – Installation, Setup and Security


Section 2 – Program Overview – Getting Started
Section 3 – Running a Test Job
Section 4 – Network Installation

Chapter 2 Section 1 – Steady State and Transient Program General Input


Section 2 – Program Utilities
Section 3 – Valve Closure and Transient Pressure Waves
Section 4 – Waterhammer Example Problem
Section 5 – Acoustic Frequencies Example Problem
Section 6 – Solution Theory
Section 7 – Safety Relief Valve Notes and Examples
Section 8 – SRG and SHW File Formats

Chapter 3 Section 1 – SPLASH 2D CFD Calculation

Chapter 4 Section 1 –Reciprocating Compressor Pulsation and Mechanical Analysis


Section 2 – BOS Fluids and API 618
Section 3 – API 618 Compliance for Reciprocating Compressors - Summary
Section 4 – BOS Fluids API 674 Compliance
Section 5 – Reciprocating Equipment Notes

Chapter 5 Section 1 – Slugs and Flashing


Section 2 – Vortex Shedding at Tees & Control Valves

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Installation and Setup BOS Fluids 4.110


BOS Fluids is installed with FE/Pipe. BOS Fluids owners will have
demo access to FE/Pipe and full access to BOS Fluids. FE/Pipe
owners have full access to FE/Pipe and demo access to BOS Fluids.
BOS Fluids is distributed using the standard Microsoft installation program: setup.exe,
and can be run on Windows 95, Windows 98, Windows NT 4+, Windows 2000 Windows
XP or Windows ME. The installation path should not contain blank characters.
(Certain viewing features requiring DirectX7.0a or later will not be available on Windows
NT systems.) The BOS Fluids installation CAB files can be found on the delivered CD in
the FPPACKAGE folder. To install BOS Fluids and FE/Pipe (remember, they cannot be
installed separately since they use many of the same system data processing
components):

1) Using “Windows Explorer” or “My Computer” navigate to the installation CD.


2) Open the folder FPPACKAGE.
3) Double click on the file setup.exe.
4) Be sure the installation pathname does not contain spaces.

BOS Fluids and FE/Pipe will be installed in the installation path shown in the popup
window. Only the start:programs menu options will be modified. If the user wishes to
have a shortcut on the desktop, it may be copied from the start:programs menu and
pasted on the desktop. No system files will ever be overwritten. Only old FE/Pipe or BOS
Fluids files will ever be overwritten by the program installer. It is the users responsibility
to be sure that compatible versions of Internet Explorer, directX, video drivers, etc. are
installed on the target machine. The installation program requires that read/write
access is available for the Windows folder. If problems are experienced during
installation it is usually because the installer does not have read/write privledges to the
Windows or to the installation folder.

Paulin Research Group and Dynaflow use a software security mechanism based on the
user name and computer name. A 7 digit key is provided for each username and
computer name pair that should be able to access the software. Each license seat
includes a key for one user/computer name. (Any version may be run over a network,
but special setup is required for limited seat network access. See Chapter 1, Section 4.)

BOS Fluids and FE/Pipe will always run in a limited DEMONSTRATION mode. The
certification is controlled by allowing access for authorized user names and computer
names. The user name and corresponding computer name that should have access to
BOS Fluids should be emailed to Paulin Research Group at support@paulin.com to
obtain the corresponding site key. Remember to include your serial number in the email
to get your full authorization. A 7-digit hexidecimal site key for each
username/computername combination will be returned by email.

The site key can be entered under program control when the program
starts, or the user can enter the site key outside of FE/Pipe by
starting the program SECURITY.EXE in the installation directory.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 1.1.1


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The file “secure.key” is created in the installation directory when the site keys are
entered correctly. The “secure.key” binary file can be moved via a copy command to any
computer on a diskette. If one machine is replaced with a newer one, the same site key
can be used. It must only be copied from the old computer to the new computer.

When a non-authorized version of BOS Fluids starts the following demonstration prompt
will appear.

If you have your site key then press YES and follow the instructions to record the site
key and unlock the program. If you don’t have a site key then you can continue on in
DEMONSTRATION mode until you get one.

Site Keys are based on your username and computername. When you have your site
key and press YES in response to the above prompt you will see the following dialog
box:

Press YES, and you will see another dialog box that contains the current user name and
computer name:

If a system administrator is loading the software for you:


If a system administrator is loading the software for you the current user name and
computer name will not be the one that should be authorized. The site key should be
generated for the actual user of the program. If you do not know your user name and
computer name a program called USERNAME.EXE has been provided that will show it

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 1.1.2


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

to you. Run the program USERNAME.EXE on the computer that should be authorized to
get the appropriate user and computer name.

If you are loading the software yourself:


If you are loading the software yourself then the username and computer name that
appears in the screen above should be the one the site key is obtained for. The program
USERNAME.EXE is provided that will show the current username and computer name.

You will get a site key for each username/computer name pair for each license you have
purchased. Email your company name, user and computer names with your software
serial number to support@paulin.com to receive your site keys. Once you have received
your returned site key via email press OK when the above screen appears and a site key
entry window will be shown:

Enter your site code(s) and press F1 to write the file “secure.key.” (Network licenses can
be purchased that allow any number of user name/computer name combinations.) The
file “secure.key” will be written in the program directory when the F1 key is entered. The
“secure.key” file should be placed in the program directory for any version of the
licensed software the user wishes to run. The “secure.key” file may be generated
outside of program control by running SECURITY.EXE.

When BOS Fluids is unlocked it will startup without any prompt for licensing information.
(When NOT unlocked the word DEMO will appear across the window handle on the top
of the screen and input will be limited.)

Starting with Version 4.1006 of FE/Pipe and BOS Fluids, the


environment variable or path statement is not required.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 1.1.3


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Overview
BOS Fluids is a computer program that models steady state and transient flow in liquid or gas carrying
piping systems. The program can input neutral files from the pipe stress programs CAESAR II, PipePlus,
Triflex, or from FE/PIpe or the user can enter the piping topology using a standard input modus. A 3d GUI
is under development, and is scheduled for release in the second quarter of 2003.
Bos Fluids estimates slugging conditions and has a 2d user-defineable CFD simulator for free surface flow
problems such as tank sloshing or slug impacts.
Bos Fluids contains the elements required to model most unsteady flow conditions, and includes the
simulation of
1) Standard Valves,
2) Relief Valves,
3) Damped Check Valves,
4) Air Valves,
5) Pumps,
6) Equipment,
7) Surge vessels,
8) Orifices,
9) Tube Ruptures,
10) Long Pipe Boundary Conditions,
11) Two phase – homogeneous flow,
12) Deflagration,
13) Isolated Slug Evaluations,
14) Reciprocating equipment,
15) Pressure Regulator Valves,
16) Relief Valve Gas Transients including Joule Thompson Effects,
17) 2d-CFD Simulations of Free Surface Flows
18) Column Separation
19) Pipe Evacuation Analysis
Graphical 3d output includes pressures, flowrates, velocities and unbalanced forces. The 3d output
graphics can be viewed using a DirectX model viewer that allows the user to zoom, cut, slice and
interactively rotate the results model. Unbalanced forces can be output in files suitable to be read into pipe
stress programs for the calculation of displacements, forces, moments and stresses. The program also
includes separate algorithms to compute characteristics of slug flow, gas relief and free surface fluid
systems.
The steady state and transient fluid calculations are based on a simplified form of the Navier-Stokes
equations.
The assumptions made are:
1. Fluid behavior is one dimensional i.e. fluid properties are constant at any one pipe cross section.
2. Fluid transport velocity is small compared to wave speed.
3. Wave fronts remain plane while propagating.
4. Gas simulations assume that flow velocities are below sonic, and that the total pressure drop through
the system is less than 30%.
The friction model used is Colebrook-White. The Darcy-Weisbach flow model is used for steady state
pressure drop calculations.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 1.2.1


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

BOS Fluids is capable of simulating both the steady and transient behavior of closed conduit systems of
pipes, valves, pumps and surge relief devices. Rectangular sections can be modeled using a total whetted
area approach. Vapor formation models exist that can simulate fully evacuated piping systems.
The following special features are available:
1. Pipe stress neutral files from PipePlus (pnf), Triflex(phb) or CAESAR (CII) can be imported for fluid
analysis. Neutral files can also be read or written to FE/Pipe. The user can also setup input in an
Excel or other spreadsheet and read it into BOS Fluids using the TEXTIN feature.
2. The user can pick different fluids from a database or add his/her own fluids to the database.
3. Two different models are available to simulate column separation: Concentrated Air Pocket (CAP)
model and the Vapor Cavity Model (VCM). (The CAP model supports the large cavity model, i.e. a
completely empty pipe section.)
4. Isotropic (Metals) and Orthotropic (FRP) materials are included.
5. Typical pump properties can be automatically generated.
6. Buried and above ground systems can be simulated.
7. The transient responses from multiple sources can be handled simultaneously in both looped and
unlooped systems, i.e. Pump starts, Pump Failures, and Valve Operations can occur simultaneously
in the transient simulation.
8. Harmonic option allows the user to analyze the occurrence of standing waves AND the effect of
reciprocating equipment.
9. Maximum and minimum pressures and velocities occurring during transient and/or harmonics are
traced.
10. The force processor allows the user to survey the time history of the unbalanced forces on pipe
sections and preprocesses the force time histories to be used in the dynamical module of the pipe
stress program.
11. A spectrum breakdown of force time histories is available. The user can see the natural frequencies
of the fluid response that will tend to excite the piping system.
12. Metric (SI) or English Units can be selected.
13. Exchanger tube ruptures can be simulated to predict the maximum pressure on the shell side of a
heat exchanger to permit lower design pressures for the shell-side thicknesses.
14. Damped check valves can be simulated to estimate the loads and backflows that can occur on pump
shutdowns, etc.
15. Mechanical models of relief valves included mass, stiffness and damping can be input, or the user
can input the set, full flow and reseat pressures for a relief valve and the program will back compute
the properties needed.
16. Air valves (vacuum breakers can be input.)
17. Pressure Regulating valves can be input.
18. Long Pipe (nonreflective element)
19. Simplified models of the effect of two phase flow on the speed of sound and net flow effect on the
piping system.
20. Deflagration (explosion) waves and pressures can be simulated.
21. Lumped gas volume simulation that can be used as a compliance element in the discharge of
pumping systems to eliminate vapor formation during pump trip, (“large volume surge vessel), or for
the simulation of pulsation bottles on the discharge of piston driven pumps, or as any other blatter-
filled dampener element, or as the gas charged section of a standpipe, or to assist in the proper
simulation of a single or multi-cylinder compressor bottle.
22. Output from long simulations can be isolated either by geometry or time window and long outputs
minimized. Maximum pressure trapping can be used to find the maximum pressures in pipe sections
with a minimum of output.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 1.2.2


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The program has been used in the fluid analysis of:


A. Water transmission and distribution systems,
B. Main cooling water systems for chemical plants,
C. Sewage water systems,
D. Combined (power-drinking water) cycle power stations.
E. Oil product transport lines,
F. Tanker loading and unloading systems,
G. Dynamic behavior of chemical liquid transport lines.
H. High Pressure Heat Exchanger Tube Ruptures
I. Vessel Blowdown
J. Triplex pump surge bottles.
K. Compressor Piping Systems
L. Acoustical Natural Frequency Calculations

The basic theory applied in BOS Fluids can be found in Wylie & Streeter's "Fluid Transients" published by
FEB Press. Various chapters herein also contain descriptions of the basic theory used in BOS Fluids.

API RP 521 2.3.14.1 Water Hammer states:


“The probability of hydraulic shock waves, or water hammer, occurring in any liquid-filled system should be
carefully evaluated. Water hammer is a type of over-pressure that cannot be reasonably controlled by
pressure relief valves, since the response time of the valves is normally too slow. The oscillating peak
pressures measured in milliseconds, can rise to many times the normal operating pressure. These
pressure waves damage the pressure vessels and piping where proper safeguards have not been
incorporated. Water hammer is frequently caused by the action of quick-closing valves. Where water
hammer can occur, the use of pulsation dampeners should be considered.”

Suggestions:
The fewer nonlinear elements included in a run the easier it will be for the user to intuitively understand
how the model is performing. Most startup runs are made to find mistakes in the model coding or in the
fluid assumptions and so introducing nonlinearities slowly and methodically is a good practice. Highly
nonlinear elements include:
1) Pumps
2) Cavity formation
3) Damped Check valves
4) Damped Relief Valves
5) Tube ruptures or air inlet valve elements
6) Complete line drainage (large cavity formation)

When starting a system analysis the user should probably:


1) Turn column separation and/or cavity formation off.
2) Keep check valves and relief valves linear in behavior. (Set the valve exponent to zero for check valves
and only use Pset, Prated and Preseat as specifications for the relief valve.)
Once the initial runs are made and the system is performing as expected, the user should introduce column
separation and damped check and relief valves to see how they effect the already understood solution.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 1.2.3


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Getting Started

BOS Fluids and PRGMAPS.EXE

BOS Fluids is accessed via a program called PRGMaps. PRGMAPS.EXE starts up one of several interfaces for
the program. The PRGMAPs main menu can be used to access FE/Pipe or BOS Fluids. The PRGMAPs main
menu is shown below:

Each is BOS Fluids program button is described briefly below:

The steady state and transient gas and flow simulator for pipe networks. Used for
waterhammer, steamhammer, slug flow, tube ruptures and pump startup or shutdown problems. Handles liquids
and gases where some compressibility effects are negligible. Interfaces with CAESAR, Triflex, PIpePlus,
FE/Pipe and semicolon delineated text files. Also includes the gas transient option for Joule-Thompson effects
in relief lines.

2D Free Surface CFD modeler. Used for sloshing in tanks due to pitch and roll loads or
earthquake. Can simulate single slug impacts on elbows and breaker plates.

Includes a Slug Velocity and state calculator, and processors for reciprocating pumps and
compressors. Also included is an evaluation tool for API 674 for reciprocating pumps and API 618 for
compressors.

The Running a Test Job example can be used to become familiar with the “General Liquids and Gases
program”. SPLASH and the “Miscellaneous” programs have separate chapters each dedicated to their
use. The first time user is encouraged to review these respective chapters before starting the program.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 1.2.4


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

BOS Fluids runs under the Microsoft operating systems Windows 95, 98, 2000 or XP. The NT operating system
is supported, but is being phased out.

All BOS Fluids documentation is delivered in electronic format. Please check for the latest pdf files in the
installation folder once the software has been setup. Also please check the downloads area on the Paulin
Research Group website at www.paulin.com for the latest document updates and patches.

Windows user's may have to experiment with large or small fonts on the computer via the control panel to be
sure that BOS Fluids performs most effectively. User’s are recommended NOT to run out of the installation
directory. Right click on your startup icon and change the “start in” subdirectory in the shortcut so that this does
not happen.

There cannot be two analyses running simultaneously in a single subdirectory. If two jobs must be run at the
same time, they can be run from different subdirectories.

If multiple jobs must be run, then the batch processor can be used, and in this case each job will be run one-
after-the-other and the jobs must reside in the same subdirectory.

Numerous input files referenced in this documentation exist in the <installation>\BOSMODELS subdirectory.
Most of these models are in a ready to run state for the user to test and explore.

Support

From time-to-time questions will arise while running BOS Fluids, or during the interpretation of output. When this
occurs, user’s are encouraged to read through the documentation as every effort has been made to include a
discussion of the most commonly encountered problems. If the user is still unsure of confused any questions
may be emailed to support@paulin.com. Please include the company name and serial number as part of
the support request. Support requests are typically answered in the morning, at noon, and prior to the end of the
day, if not more often. Include a copy of the job and any pertinent drawings in the email. Autocad DWG files may
be included if they will make understanding the problem easier. The Paulin Research Group or Dynaflow
Engineering can also be contracted on a consulting basis to review or setup models for first time users or for
more experienced users that simply need additional assistance.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 1.2.5


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Running A Test Job

Test Model

The following simple system will be used to demonstrate the use of the Bos Fluids transient flow analyzer. A
straight water evacuation system that is 80 meters long has a gate valve at the end that closes in 0.5 seconds.
The analysis should predict the pressure rise in the system upon valve closure.

Starting the Program – PRGMAPS.EXE


The BOS Fluids program is bundled with FE/Pipe and they are both installed together using the installation
routine found in FPPACKAGE. The default folder is FP4109, although the user can change the installation
folder to any name desired providing it does not include spaces.
Once the software has been installed the target startup program is PRGMAPS.EXE. This program can be
started from the Start button, from a command prompt, or the user can double click on PRGMAPS.EXE
while in explorer.
The PRGMAPS main screen appears below:

Select ”File” and choose New FE/Pipe or BOS Fluids Job. The menu that appears depends on the
operating system, but the Windows XP files menu will appear like the one shown above:
From this menu the user can:
1) Create a new file in the current working folder.
2) Navigate to another folder and create a new input file.
3) Create a new folder, then open that folder and create a new input file.
To create a new folder, right click in the file list window, then select “new” and folder. Give the new folder
name, and then double click the folder to open it. To create a new file, type the name in the File name text
box and hit the <enter> key or click on the save button. The new filename should appear in the blue

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 1.3.1


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

“jobname” box on the main PRGMAPS menu like the example shown below.

Units
FE/Pipe and BOS Fluids have two unit sets – English and Metric.
The units may be set on a program-default basis, or on a job-folder basis. To set the units on a program-
default basis, open a new jobname in the installation folder. When the new jobname appears select utilities

Then Configure:

And from the Configure options select “Units Specification”

Use the arrow keys to select either Metric or English options and then hit the <enter> key. (The example
model will be run in METRIC units.)

Then hit the key “K” to Save changes and Return. If the user starts the unit setting process when the
current working directory is the program installation folder then a new “config.bin” will be written in that
folder. The new “config.bin” will set the defaults for every new job that is not in a folder that has a local
config.bin file.
If the user starts Configuration process in a working folder, the new “config.bin” file will be written into the
local folder, and any job that is started in that folder will use the new config.bin settings. Jobs that were
already in the local folder will start with the already established units.

Once the desired unit set has been established, the user is ready to continue on with the model input.

JOB FILES SHOULD NOT BE STORED IN THE INSTALLATION DIRECTORY.

Although many users store their data files in the program installation directory this is not recommended.
Please select a new job in a folder that is NOT the installation folder. The new name should appear in the
blue Jobname box on the PRGMAPS menu and the working folder name should appear in the window
handle at the top. When the jobname is properly established, click on the BOS Fluids button:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 1.3.2


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The button will expand into two buttons:

Since a new job is to be entered, click on the New Job Button.

At this point the user must decide what type of analysis should be performed:
General Liquids & Gases – Typical waterhammer, steamhammer, gas transient analysis. Choose for API
618, Pumps, Tube rupture, relief valve analysis, rapid valve closure, etc.
SPLASH - CFD-2D Sloshing or slugging local load, pressure, velocity analysis. Free surface, two
dimensional, computational fluid dynamics evaluations. Can be used for periodic, or time history loadings,
seismic evaluations of liquids in tanks, and vertical, or horizontal vessels.
BOS Fluids Misc. Calcs. – Includes:
1) Slug Estimator: Determines the tendency for slugs to form in both vertical and horizontal pipe
runs, and calculates the maximum slug velocity possible in a close-coupled piping system.
2) Reciprocating Pumps: Computes the harmonic input frequencies and amplitudes that must
be checked for single, duplex or triplex pumps.
3) Reciprocating Compressors: Evaluates the compressor suction or discharge waveform to
determine the fourier coefficients that should be used in a discrete harmonic analysis of the
piping system to satisfy an API 618 analysis of the system.
Select the button General Liquids & Gases:
The following screen appears to let the user select the major pressure unit. This is purely a user
preference. Some analysts are more comfortable with pressure reported in feet or meters of fluid head,
others prefer pressure units in force per area.

At the start of each input file the user is given the option of selecting the major pressure unit for the job.
Once the major pressure unit is selected, it remains fixed for that job.

Most piping and vessel engineers prefer psi or bars as the major pressure unit. For metric applications the
example uses bars.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 1.3.3


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Once the major pressure unit is selected, the BOS Fluids main input menu is displayed:

Select 2-Pipe to enter the basic model geometry.

Enter the node numbers: 5, 10, 15 in the top data line. Two elements are created. The first element runs
from 5-to-10 and the second element runs from 10-to-15. This approach is taken to reduce the number of
element data screens.
Each element will be 40 meters long in the X direction. The pipe diameter is 250mm, with a wall thickness
of 5mm and a roughness of 0.05mm. The other inputs for steel, the ambient fluid temperature and the fluid
name are defaults. The fluid temperature effects the fluid properties extracted from the data base, but
otherwise is not used in the analysis.
Equipment, valves, pumps, surge vessels, and orifices can be defined for any pipe element. The key F10
can be used to invoke a flow/pressure unit converter.
PgDn and PgUp keys let the user navigate between pages. The F11 function key can be used to search for
a particular node or element in the existing data base.

Only a single node may be entered in the search screen, and the first element containing that node will be
found.
Using the function keys, boundary conditions at the nodes can be set [F3], or special properties can be

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 1.3.4


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

added to the elements [F4-F8]. Hit the [F3] key to specify the boundary conditions at Node 5. Enter the
numbers shown below to define the fixed reservoir pressure of 10 bars at Node 5.

A wide variety of boundary conditions can be applied to nodes. These include DEADENDS, LONG PIPES,
SONIC HEADS, TIME heads or flows, HARMONIC heads or flows, and FIXED heads or flows.
Using the above screens, two pipe elements have been described along with the boundary condition at
node 5. Next the valve at the end of the system should be entered so that the flow can be shut down.
Escape from the boundary condition screen and hit the PgDn key to get to the next blank pipe element
screen and enter the dimensions of the valve from nodes 15 to 20.

Hit the [F5] key to define the valve for element 15-20. The valve can also be selected by clicking on the
button just to the left of the F5-Valve option on the bottom of the screen. (See the figure above.) The
program enters the node numbers automatically on the valve screen. The only inputs required to define the
valve will be the Valve Type and Identifier. “Gate” and “v100” are entered for the example below. Hit the ?
key to invoke help for any of the data cells. For this small example problem, the gate valve on element 15-
20 will be closed in 1/2 second and the resulting pressure rise will be calculated.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 1.3.5


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Use the <esc> key to exit from the popup valve window. The ESC key will exit from any
BOS Fluids Pop-Up window.
When back on the 15-to-20 element page use the [F3] key again to enter the boundary conditions for the
node 20 at the end of the valve. This will be a fixed head and will represent an open end into the
atmosphere. Note that the atmospheric pressure in gage is input as zero.

The user must always provide either pressure or flow at the start and stop of any system to
establish a steady state condition. The steady state solution calculates pressure if a flow is given,
or flow if a pressure is given. There must be sufficient boundary conditions to establish an initial
flow state from which to start the transient. An option exists to start the transient solution from a
NOFLOW state, but this is not typical. For pump or pipe sizing, the steady state flow condition is
the critical one and is governed by friction and valve loses. For transient flow problems, the
steady state condition is not as important since friction losses are of secondary importance. The
steady state solution must be generated only to establish the proper initial flow velocities.
Escape back out to the BOS Fluids Input Menu and select the Optional Input screen.

Set the Diameter Visualization Multiplier to 6 to enhance the 3D display for the test job. This parameter is
used whenever long lengths of pipe are modeled which appear as only fine lines when plotted to scale.
The Diameter Visualization Multiplier increases the diameter for all pipes in the model, so that each pipe

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 1.3.6


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

may be more easily viewed.


The parameters “Maximum Simulation Time” and “Maximum Timestep” have been left blank. These
important parameters are calculated automatically by BOS Fluids to capture the slowest moving transient
that may excite the mechanical piping system or that would cause high or low pressure problems with the
fluid system.
Escape from the Optional Input screen and select Action.

Change the Analysis type to “transient” and the select F2 to enter a description of the valve transient.

The valve ID is “v100” – the same as was entered when the valve was defined, and the valve will go from
100 % (fully) opened to 0% (fully closed) in 0.5 seconds.
Running the Analysis
Escape (Esc) from each of the input screens until the main BOS Fluids screen is shown below:

This screen has C-Save, D-Plot, E-PREPARE for Analysis, and F-SUBMIT for Analysis options on the
bottom:
C-Save – updates or creates a jobname.ifu file with the job input data. This is the file that contains the job

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 1.3.7


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

database in binary format that can be read to start the model analysis session. Job files are automatically
saved each time Save, Plot, Prepare, or Submit is executed.
D-Plot – creates a plot of the geometry, but does not prepare the model for analysis. Plot is used to check
that the geometry is entered correctly. The input file will be “saved” prior to plotting.
E-Prepare for Analysis – creates a plot, and also prepares files needed for the steady state and transient
analysis of the system.
F-Submit for Analysis – creates a plot and prepares files needed for analysis, but also starts the analysis
requested if there are no errors in the model.
At this point the user can “prepare for analysis” from the BOS Fluids menu, interrogate the resulting graphic
image to be sure that it is correct, and then start the run from the PRGMAPS, “Old Job”, Analysis menu
selection; or the user can “Submit for Analysis” from the BOS Fluids menu to go directly into job analysis.
Users are encouraged to experiment with both approaches to determine which one is most useful for a
given situation and problem.
When a transient run is started, the steady state module is first invoked to solve for the steady state
pressures and flows. When the steady state module completes, the transient module is started and the
screen below is displayed counting through the time steps needed for the transient solution.

Graphical, animated, and tabular reporting options are available once a run has completed. Several plots
and a summary of the tabular listing for this simple run are included and discussed on the following pages.

2D Plots – Pressure, Flow, Velocity, Force time history or frequency plots for single nodes in the piping
system or for valves, pumpts, etc. Plots are generated in a 2D – XY format with time or frequency along the
range and pressure, force, velocity, or flowrate along the ordinate.
3D Plots – Allow the user to plot pressure, flow, velocity and forces on a three-dimensional rendering of the
model. The maximum results from transient analyses can be plotted along with animated views of pressure
wave movements, etc.
Reports – Tabular summaries of the steady state and transient calculation are included.
The system acoustic response time can be found from the approximate expression:
2L/c
where L is the distance from the source of the disturbance to the point in the piping system where the
disturbance will be reflected, and "c" is the speed of sound in the fluid. For rough estimates of the
response time in liquid filled pipes c may be taken as 1000 meters/second (approximately 3000 feet

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 1.3.8


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

per second), and the basic system acoustic response time can be found from:
[L(in meters)] / 500] (sec.)
<or>
[L(in feet) / 1500] (sec.)
When the duration of the transient event is much larger than the response time estimated above the
system will respond in a "slow" or pseudo-static mode. When the transient event duration is less than the
response time estimated above the system will respond in a "fast" or dynamic mode.
The system acoustic response time for the example can be found from:
time = (80 meters) / (500) = 0.16 seconds.
The transient event duration is 0.5 seconds which is larger than the system response time, and so "slow"
responses will be observed. The increasing pressure wave will return to the source and begin canceling
further increases before the full value of the unbalanced pressure will be applied to the system.
This phenomenon is experienced in the plant, and in many households. When valves are shut quickly
there is often a resounding "water hammer" through the pipes. When the valves are closed slowly there is
hardly a whisper. BOS Fluids helps the user know where these breakpoints are, and more importantly,
aids in quantitatively assessing the magnitudes of pressure rises and unbalanced loads that result.
The maximum pressure rise that could occur in any system due to a theoretical "instantaneous" valve
closure is:
pressure rise = (fluid density)(c)(V)
where (c) is the speed of sound in the fluid, and (V) is the initial velocity of the fluid.
BOS Fluid calculates the initial velocity of the fluid and makes this available for plotting. For the system
analyzed above the initial velocity is 18 meters/sec. (An artificially high flow velocity because of the
extreme pressure drop used in the input.)
Instantly stopping this flow would produce pressure rises on the order of:
pressure rise = (1000 kg./cu.m)(9.8)(1000)(18) = 176,000,000 N/sq.m.
This is equivalent to 1700 bars. The BOS Fluids calculations shows that the pressure rise is only about 25
bars as seen from the plots on the following pages. Designing for the actual pressure rise of 25 bars will be
a considerably simpler task.
Bos Fluids includes the effect of vapor formation when the line pressure drops below the vapor pressure.
This effect will be a part of the solution to this problem because when the fluid attempts to “bounce back”
from the fully compressed column state, the pressure will drop to below the vapor pressure for water. The
Concentrated Air Pocket column separation model is "on" by default for BOS Fluids runs but may be turned
off if the user wants to allow negative pressures. The column separation effect can be seen in the transient
solution when pressures begin to go negative. This occurs in the 2D pressure plot below at around 0.6
seconds. At this point a rarefaction or negative pressure wave exists. The curve is flattened out because
the lowest pressure possible is the vapor pressure of the fluid. The pressure during the simulation can
never drop below this value when a column separation option is activated. Once vapor is formed it remains
in place until the line pressure returns. When this happens the bubble can collapse and produce positive
pressure spikes in the system.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 1.3.9


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Results

When the user clicks on the 2D Plots button the following screen appears:

Under the 2D menu option the user may select:

Pipe 2d
Non-Pipe 2d

The “Pipe 2d” option menu is shown below:

From this panel the user can select the quantity to plot and the type of plot to display. A node number is required
because the pressure, flowrate, velocity or unbalanced force for a single node will be plotted. The pressure at
nodes 5, 10 and 15 for the example problem is shown below:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 1.3.10


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Node 5 Node 10 Node 15


The flowrate and nodes 5, 10 are shown in the plots below:

3D Plots

When the user on 3D Plots a three-dimensional line-figure of the piping system appears as shown below. The
plot options can be found under the Transient/Steady menu option:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 1.3.11


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The following four options are available from the Transient/Steady menu:

Steady State Options


Transient Options
Animate
Time

The steady state option selection can be used to superimpose the steady state pressure or flow solution on the
three dimension plot. A steady state pressure plot for the example system is shown below.

Varieties of transient and animation options are available from the 3D plotting menus. These plots can also be
transformed into the 3D viewer and rotated, zoomed, panned, ect. in real time. The user is encouraged to
experiment with these options to find those that best suit the current output review needs.
Tabular report data is also available. The tabular reports contain an echo of the input in a more readable format,
and various warning and parameter reports.

Output reports can be edited by the user and saved or printed. This facility allows the user to choose only those
results that are pertinent to an analysis, while deleting or omitting others. The user can also add discussion
of results directly into the body of the BOS Fluids report for improved discussion and readability. The
delete key deletes any line and the insert key inserts one or more lines into the body of the report for additional
comments.
Any request from the output processor to send tabular results to a printer writes a <jobname>.lis file. This file
(which many users employ for the start of their documentation) may also be written directly from the output
processor. If the user outputs a <>.lis file, and then subsequently sends a report to the printer. The printer
report .lis file will overwrite the previously written .lis file.
Several different report types are shown below.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 1.3.12


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Input Echo

The steady state report shows that the linear iteration to solution took five iterations. BOS Fluids also supports
an Intelligent System Reduction Mechanism that remeshes the structural geometry so that a more efficient fluid
system results.

The steady state solution gives the pressures and flowrates from the steady state solution. These are the
pressures and flows that are used to start the transient analysis.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 1.3.13


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The Transient Warnings and Messages report gives the user a variety of useful information about the transient
analysis. In particular any unusual events that occurred during the transient analysis are reported here.

The “Pipe Parameters” report is used to check the calculated wavespeed for the system. In this case the
calculated wavespeed is 1240 m./s. The # of reaches reported in the far right column is the number of internal
system nodes that have to be created to trap the correct maximum pressure at any point in the pipe length.

The maximum and minimum Transient report gives the maximum and minimum values of pressures, flowrates,
forces and cavity sizes that occur at any point in the transient. The values printed on the far right of the report
tell when in the element the maximum occurred. For example in the report above, the maximum pressure
occurred at a distance of 40 meters from node 5. The maximum cavity size at any node is reported in percent of
total cross section. The cavity size is that amount of the local pipe section that is completely filled with vapor.
When the value reaches 100% then the pipe is completely empty. When extremely large areas of the pipe will
be exposed to void sections, the user should activate the large cavity model to properly track the void
movement.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 1.3.14


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Network Versions of Paulin Research Group Software

What is a Network Version?


A network version of PRG software has the following characteristics:
1) The software MUST be loaded on a remote server.
2) Many users will have access to the program, but simultaneous use is limited to the
number of licenses purchased.
3) A network administrator must be available to maintain the account.

Network Versions require:


1) All users must have read/write/create access to all files in the installation subdirectory
on the server, including the secure.key file.
2) The server drive id cannot change. If the server drive id changes, then the software
must be reinstalled. The encryption algorithm uses the server drive. This number must be
available to the OS via the Windows API procedure: “GetVolumeInformation.”
4) The startup program for BOS Fluids is PRGMAPS.EXE. Shortcuts or batch files
should point to this executable to start BOS Fluids.
5) The network administrator must collect the username/computername pairs for all users
that will possibly want access to the system. The username/computername pairs must be
entered into an ascii file named username.txt in the installation subdirectory on the
server. The network administrator can change this file at any time.

Network Version Pricing:


1) Network versions require some setup fee depending on the number of users that will
have access to the system and the number of licenses supported. Contact
sales@paulin.com for more information on network pricing for each PRG product.

To Install a network version:


1) First you must decide which of the PRG products should be available in a network
serviced format. BOS Fluids, FE/Pipe and ALLPRO can all be networked apart or
together. Just about any configuration can be designed to suit the users needs. For each
product purchased, the user must decide the number of simultaneous users, and the
maximum number of users that will want to have access to the program.

If the user access to BOS Fluids, ALLPRO, NozzlePRO, AxiPRO, or FE/Pipe will be
different then the programs will have to be loaded into separate installation directories. If
they are to have identical user access then they may be loaded in the same subdirectory.
Either program can be loaded first.

The network delivery version is no different from the standard demonstration version of
the program delivered on all CD’s.

Be sure that read/write access is available for the Windows folder on the machine used
for the installation. The Microsoft setup.exe program uses this folder to store temporary
files.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 1.4.1


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

2) Run the program NETDRIVE.EXE from a command prompt to determine the installed
drive id of the server! Include the root path in the program invocation, i.e.

NETDRIVE C:\

The message box that appears should give the network installed drive identifier. An
example is shown below:

The ID number, (in the example –1124132195), must be returned to Paulin Research
Group along with the other network setup parameters, i.e.

1) Maximum possible number of users.


2) Number of simultaneous users per program.

3) Collect from each user their login username and computername. The base security
system used at Paulin Research Group identifies users based on these two identifiers.
Users may determine their own username and computername by running the delivered
program USERNAME.EXE. Example output from the program appears below:

4) Return the server ID number to the Paulin Research Group. One “secure.key” file will
be returned for each separate network installation required. The email address at Paulin
Research Group is “support@paulin.com”

5) Enter the username and computernames into a single ascii file named:
USERNAME.TXT. Do not include any spaces in either usernames or computernames.
For example, if a username is “Bill Smith,” the entry into the USERNAME.TXT file

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 1.4.2


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

should be “BillSmith”. The username should be first. And any number of spaces can be
used to separate the username from the computername An example few lines from a
USERNAME.TXT file are shown below:

administrator HP-0013A
BillSmith HP-0012B
PipeStress HP-00132B

Each user must have read/write access to this file to be able to use the program. The
network administrator can change the entries in this file as needed when users or
computers enter and leave the network.

To maintain a network version:


Inevitably users will be logged into the system when other users want to run the software.
When this happens the following message will appear on the user’s screen that is trying
to run the software:

If users that are “logged in” do not get “logged out” when they leave the software either
because of system crashes, dropped network connections, etc. the security system will
believe that a license is still in use that should be available. When this happens the
system administrator can run the program:

NETRESET.EXE from the installation directory on the server machine. This program
will reset the security system, logging out all users.

Additionally, if a user suspects that he/she may be logged in, and wants to be sure to “log
out” of the system, the program LOGOUT.EXE can be run from the user’s machine to
logout only a single user. (The user logged out will be the one identified by the
username/computername combination where the program LOGOUT.EXE is run.) The
file NETUSE.TXT contains each login to the system. NETUSE.TXT is an ascii text file
that contains the time of login and the username/computername combination. Only
successful logins are recorded.

The file secure.key that was originially transmitted from Paulin Research Group can be
reloaded at any time and the entire security system will be reset.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 1.4.3


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

There should only be one copy of secure.key on the server drive for each network setup
established.

Summary:
There is considerable flexibility in how the network installations can be configured. The
user must decide how many people should have access to each software program, and
how many will ever need to use it simultaneously.

The simplest configuration would allow the same number of users to access all PRG
programs installed. In this case, the network administrator should load all programs in
the same installation subdirectory on the server. (Either can be loaded first.)
Alternatively, each program can be kept in a separate subdirectory.

Each user must have read/write access to the installation directory.

The network administrator should run the program NETDRIVE from the server with the
argument pointing to the root of the drive where the program will be loaded, i.e.

NETDRIVE C:\

The drive id returned should be transmitted to the Paulin Research Group.

The administrator should collect the username and computername combinations that
should be able to access the system. These should be entered into a single ascii file and
placed in the installation subdirectory on the server. There should be no spaces in either
the username or the computername. When entering usernames and computernames in the
USERNAME.TXT file, just ignore spaces in the names. Do use spaces to separate the
username from the computername.

Paulin Research Group will provide a file: secure.key for each network installation.
Copy this file to the program installation directory on the server and the system should be
ready to use.

Any number of username/computername combinations can be entered into the


USERNAME.TXT file, but only the first <n> will be used, where <n> is the maximum
the user has requested for his license.

Network Checklist:
1) Make sure that the USERNAME.TXT file is in the application directory on the server.

2) There should not be spaces in either the username or computername. The only
space(s) should be to separate the username from the computername on a single line. If
the username or computername has spaces in it, then ignore them when entering the
username and computername into the USERNAME.TXT file.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 1.4.4


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

3) Be sure that any shortcuts to the programs point to PRGMAPS.EXE for BOS Fluids
and FE/Pipe, and to STARTNP.EXE for NozzlePRO, and to STARTAXI.EXE for
AxiPRO.

4) Make sure that the user has read/write/create priveledges to all files in the installation
directory. Insert a username/computername for the network administrator in the
USERNAME.TXT file and see if the network administrator can start BOS Fluids in a
non-DEMO mode.

5) Be sure that the secure.key file is copied into the program installation directory on the
network.

6) If one secure.key file is returned, then all PRG programs must reside in a single
installation directory. If two or more secure.key files are returned, then be sure not to
confuse the files. (Each secure.key file will have a different program setting in this case.)

7) Rerun the program NETDRIVE.EXE on the server drive and make sure that the
number (drive ID) returned is the same one that was sent to PRG.

8) PRG Network protected software will not run in DEMO mode. If the program starts in
DEMO mode, then most likely either:

a) A shortcut is incorrect and the wrong program is being started.

b) The secure.key file is corrupted or in the wrong place.

Problems with network versions should manifest themselves with network-related error
messages.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 1.4.5


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

General Liquids & Gases


Each of the BOS Fluids, “General Liquids & Gases” input screens are described in this section. The input for
each cell is described in detail and numerous reference suggestions are given. In all cases the “?” key can be
entered when the cursor is in a data cell to bring up help for the data cell. TextIN, TextOUT, and TextADD
options appear across the top of each input window. These features allow the user to write out, or read in a text
file that contains the basic screen data. Writing out a text file is useful when the user wants to perform an
operation on the data in excel, notepad, or some other program that is more suitable for the type of change
required than the BOS Fluids editor. User’s can also use these files to create BOS Fluids input from a
spreadsheet. Input and output is to the standard text file TEXTIO.TXT. The TextIN, TextOUT and TextADD
commands all read or write to the file TEXTIO.TXT. Semicolons are used as cell delimiters in TEXTIO.TXT.
This is because some cells in BOS Fluids (the coordinate cell for example) permit the entry of commas to
delineate entries. The user wishing to use these features are encouraged to experiment writing, viewing and
modifying existing example files to gain an understanding of how these alternate input features can be used.

The Gas Transient module is used to assess the temperature drop downstream of control or relief valves. These
temperature drops can drop the line temperature below the nil ductility temperature for the pipe producing a
potential for brittle failure. Superheated steam systems can also be analyzed and the drop in temperature
checked to be sure that excess condensate does not form in the outlet flow. This module and its input is
described at the end of this section.

The following flow tolerance exists on the optional data form:

For small diameter pipe systems, or for systems where very small flows are expected the flow tolerance should
be set accordingly. Previous versions defaulted this tolerance to 0.1, and for certain sized piping or tubing
systems this tolerance may not be sufficiently accurate.

Input Menu

Save
Saves all inputs and/or changes made to the job, without having to plot the model. Plot, Prepares and
Submit always “Saves” first before continuing. “Save” does not prepare the model for analysis. A “saved”
model should not be analyzed. The “save” feature is intended to update the image of the input on the hard
disk so that work is not lost in the event of a power failure or system crash.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.1


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Plot
Generate a plot of the model. Only a graphical image will be generated. Solution files are not written. This
is a model check feature only. To plot and prepare the model for analysis, use the “Prepare for Analysis”
feature. The Plot feature will generate plots when there are fatal errors in the model if possible so that the
user can interrogate the plot to try to find any mistakes. “Prepare” often will not generate the plot of a model
if fatal errors in the model are detected.

Prepare for Analysis


“Prepare for Analysis,” plots the model, and writes all necessary data files to run BOS Fluids. After the
“prepare” step, the user can go straight to Analysis from the main menu. A “Prepare for analysis” will check
for fatal errors, exit the input program, write all data files needed for the solution requested and generate a
plotted model of the geometry for interrogation. When the plot session is ended control is returned to the
PRGMAPS main menu where the user can “ANALYZE” the BOS Fluids model.

Submit for Analysis


Submit for Analysis plots the model, performs all fatal error checking, prepares the job for analysis, writes
all necessary data files for analysis and starts the analysis directly. There is no need to return to the
PRGMAPS main menu to start the analysis. This option should be used only when the user does not need
any preview of the model geometry before running.

Pipe Elements

Start, Middle, End Nodes


Enter two or three node numbers. Two node numbers are entered to define a single element, and three
nodes are entered to define two elements.
If three node numbers are entered, i.e. 5, 10, 15, the first element goes from 5-to -10, and the second
element goes from 10-to-15. If two elements are defined by this screen then enter two sets of delta
coordinates (dx, dy, dz) in the coordinate data cells.
Any defined nodes may go to the tangent intersection points (TIPT's) of bends. The two pipe segments
may be co-linear to define a straight run of pipe or at any angle. Bends may be placed at either end, or at
the middle node of the segments. Separate the node numbers by spaces or commas.
Several node numbering examples are shown below:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.2


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Bend Tangent Node


Enter up to three node numbers. The node numbers should be nodes that are located at tangent
intersection points on bends. (In the two examples on the right above, node 10 is a tangent intersection
point for bends. In the far right example, 15 is a tangent intersection also.
Tangent intersection points only have to be defined once, and may be defined on any element on which
they appear. The bend radius is inserted automatically by BOS Fluids. The user can run from tangent
intersection-to-tangent intersection as shown in the far right sketch above.
Separate node numbers by spaces or commas.

dx, dy, dz
Enter the delta coordinates to define the length and orientation of the pipe elements.
For example, if the node numbers entered are: 5, 10, 15, the first set of delta coordinates would be entered
to describe the element from 5-to-10, and the second set of delta coordinates would be entered to describe
the elements from 10-to-15. In English units a feet-inch format can be used, i.e. 5 feet, 10 inches can be
entered as 5-10. In Metric units the exponential format is most useful for entering dimensions in meters, for
example 80 meters is entered 80E3.
Separate numbers by spaces or commas.

Pipe Diameter, Thickness and Roughness


Enter the actual outside diameter of the pipe followed by the pipe thickness, followed by the pipe
roughness. The roughness may be omitted and will default to 0.05mm.
There is no provision for converting nominal pipe sizes to actual.

Bend Radii
Enter the radii for the bend tangent nodes given. If there are three bend tangent nodes then enter three
radii.
If left blank or zero the radii will default to 1.5 times the nominal diameter of the pipe.
For example, if the data sheet defines the two elements from 10 to 15, and from 15 to 20; and the nodes 15
and 20 are bend tangent nodes and the bends are short radius bends, i.e. 14in. then a part of the input
data sheet would appear:
10, 15, 20 Start, Middle, End Nodes
15, 20 Bend tangent Node(s)
14, 14 Radius (Bends Only)
If 10, 15 and 20 are ALL bend tangent points that define 14" short radius bends, then the input data sheet
would look:
10, 15, 20 Start, Middle, End Nodes
10, 15, 20 Bend Tangent Node(s)
14,14,14 Radius (Bends Only)
If 15 and 20 are bend tangent points, and only at point 20 does the bend have a short radius the input data
sheet would look:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.3


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

10, 15, 20 Start, Middle, End Nodes


15, 20 Bend Tangent Node(s)
DEFAULT, 14 Radius(Bends Only)
The word "DEFAULT" implies that the first bend should have the default long radius, and the second bend
should have a 14" radius. (The words DEFAULT or NONE may be used synonymously.) Note that bend
tangent nodes and bend radii only need to be defined ONCE, on any data sheet where the bend tangent
node appears on an element segment.
Pipe Material
Enter either of:
STEEL for all low carbon and alloy steel materials (E=210,000 MPa, u(xy)=0.3, u(yx)=0
GRE55 for 55 degree wound Glass Reinforced Epoxy (22,775, 0.65, 0.38)
GRE63 for 63 degree wound Glass Reinforced Epoxy (24,515, 0.62, 0.26)
GRE73 for 73 degree wound Glass Reinforced Epoxy (26,965, 047, 0.15)
GRP00 for continuous wound Glass Reinforced Polyester (17,715, 0.3, 0.25)
FUJIK Helical Wound GRP (18,091, 0.69, 0.39)
UPVCO (2,757, 0.4, 0)
PVC (2,895, 0.15, 0)
Other materials as needed can be added on request.
Pipe Restraint Description
Enter any of:
STD Above ground, typically supported piping.
BURIED Below ground pipe, axially restrained due to the restraining friction action of the soil.
BELLOWS Pipe, of expansion joint, or bell and spigot construction that cannot support axial loads
without separating.
IGNORE Pipe elements that are NOT A PART OF THE FLUID MODEL. This option is used most often
when piping systems are brought into BOS Fluids from an external source. These models
often contain constructions required for mechanical analyses that have no meaning for the
fluid simulation. These elements MUST be ignored, or BOS Fluids will try to find a flow path
through the collective element system.
LUMP Used for very short pipe segments to set the speed of sound in the element to infinity and to
eliminate the element from the time stepping determination.

The different element types determine how the speed of sound is calculated for the segment. STD
elements will dilate radially and contract axially. Buried elements can dilate radially, but cannot contract
axially. Bellows elements have essentially no pipe stiffness and plastic pipe elements have an interaction
requiring the assumed winding angle and Poisson’s ratio. The LUMP element is used when there is a short
crossover between two piping systems and the length of the crossover is not important. For large systems,
the maximum time step allowed can be controlled by the smallest required element in the system. If the
smallest element in the system is a short crossover, the time stepping can get quite small and solution
times can become excessive. In these cases the user can change any pipe length not connecting to a
valve or other NON-Pipe element to a LUMP to speed the solution.

Temperature (Deg)
Enter the temperature of the fluid in the pipe. In English units this is in Degrees Fahrenheit, and in SI units
this is in Degrees Celsius. Enter this value anywhere in the fluid where the temperature changes. Multiple
temperatures are used to describe the effect of chillers or other heating or cooling equipment on the
properties of the fluid downstream of the equipment.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.4


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Description:
Enter any two lines of text to describe the element. The text cells are used only for element identification
purposes and are not included in listings or printouts at this time.

Fluid Name (F9)


Hit the key F9 or click on the small button just opposite the Fluid Name. The following fluid data screen will
appear:

The above list contains the choices currently available for fluid properties stored in the program. The option
for steam is for slightly superheated steam at 750 deg. F. Wherever there is any question about the fluid
properties the user is encouraged to enter the fluid properties directly by selected the USER option.

Liquids or gases can be added into the BOS Fluids data base if desired. The fluid data base resides in a
file named liquid.dat in the BOS Fluids installation directory. This file may be modified using Notepad or
any other compatible ascii editor. There are (6) columns of data in liquid.dat. A part of the entry for water
is shown below.

WATER 24
0.0 999.8 1.781e-3 2020.0 0.61e3 0.0
5.0 1000.0 1.518e-3 2060.0 0.87e3 0.0
10.0 999.7 1.307e-3 2100.0 1.23e3 0.0
15.0 999.1 1.139e-3 2140.0 1.70e3 0.0
20.0 998.2 1.002e-3 2180.0 2.34e3 0.0

The name of the fluid is on the first line followed by the number of data lines used to define the fluid. There
are 24 lines to define the fluid water. Each data line describes the properties at a different temperature.
Each column in the file is defined below:

Column 1 – Temperature in deg. C


Column 2 – Density in kg/cu.m.
Column 3 – Viscosity in Newton Seconds / sq.meter.
Column 4 – Bulk Modulus in MPa
Column 5 – Vapor Pressure in Pascals
Column 6 – Saturation Pressure in Pascals (Not Used)

If the user selects the “user” fluid property, the following data screen appears:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.5


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The Fluid Types available in BOS Fluids are:

GAS – Non-choked gas flows where the pressure drop is less than 30% of mean line pressure.
LIQUID – Newtonian Liquid Flow
BOTH – Homogenous two phase flow of liquid and gas.
DEFLAGRATION – Explosion wave front simulation.

Each of the items will be discussed below:

Speed of Sound – Optional input. If the user knows the speed of sound for a particular gas, liquid or
combination then its value should be entered. In general this field is left blank and the program calculates the
speed of sound based on fluid properties.

Gas Volume / Total Volume @ mean line pressure – Required input if BOTH is entered for the fluid type. This
input determines the volume ratios of gas to liquid. The speed of sound of two phase mixtures can be
significantly reduced by the introduction of small amounts of gas into the fluid. Slower speeds of sound will drop
response times and lower acoustic natural frequencies. The lowest acoustic natural frequency of a liquid-gas
system is often between the all gas and all liquid states.

Properties for Gases or the Flow Gas Component:


Estimated Mean Line Pressure (gage) – This is the estimated average pressure in the line. Typically some
pressure between the inlet and the outlet. This pressure is used to determine the speed of sound in the gas. The
effective bulk modulus of the gas is estimated as the entered mean line pressure times the ratio of the specific
heats.
Cp/Cv (Ratio of Specific Heats) – Enter this value for the gas.
Viscosity (cp) – Enter the viscosity of the gas in centipoises. Values on the order of 0.02 cp are typical.
Weight Density – Enter the weight density of the gas at the mean line pressure. Note that this value is a
typically English Units measure. For Metric users this is the mass of the gas multiplied by the gravitational
constant “g,” typically taken as 9.81 m./sq.sec.

Properties for Deflagration:


Cp – Specific Heat – Enter the specific heat of the reactant. Input should be in TU/mass unit/deg., where TU is
the thermal unit of choice. Typically BTU’s for English units and Joules for Metric units. The mass unit is kg for
Metric units and lbm for English units.
H(Enthalpy Increase per mole of Product) – Enter the enthalpy increase per mole of product expected during
the deflagration event.
M(Avg Molecular Wgt of Reactants)
M(Avg Molecular Wgt of Reaction Products)

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.6


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Ti(abs) Initial Temperature of Reactants (deg.)

The difference between a deflagration event and an explosion, is that the explosion wave front has a reaction
driving the pressure front at a speed determined by the thermodynamics of the reaction and the fluid properties.
The deflagration tends to be a slower event, and the wave front is driven by an essentially static reaction source.

The deflagration analysis proceeds as follows:

1) Build the BOS Fluids model of the vessel and attached pipe model. (Typically the deflagration source
will be a concentration of product in a vessel or other container volume.)
2) Enter the Deflagration data as described above.
3) Run a “Prepare for Analysis” and review the plot for accuracy.
4) Check the Tabular listings that describe the run to extract the calculated deflagration pressure multiplier.
5) Ramp the pressure in the vessel from the startup pressure to the product of the pressure and pressure
multiplier in 5-to-10 ms. Make sure the deflagration model type is selected so that BOS Fluids knows to
use the traveling deflagration front celerity as the system speed of sound.
6) The results from the transient are maximum pressures and forces and moments on the piping system.

Properties for Liquids or the Flow Liquid Component:

Weight Density – Enter the weight density of the liquid or liquid component of a two phase flow. Note that this
is an essentially density measure and is the equivalent of the mass multiplied by the acceleration of gravity. For
water this value is approximately: (1000 kg/cu.m )(9.81 m/sq.sec) = 9810 N/cu.m. in metric units and 62.4
lb./cu.ft. in English units.
Viscosity –(cp) Enter the viscosity of the fluid in centipoises. Water is about 1cp.
Bulk Modulus – This is the elastic modulus of the fluid. Typical values are:

Liquid psi MPa


Water 286,000 2000
Kerosene 204,000 1430
Benzene 147,000 1030
Ammonia 305,000 2135

Vapor Pressure (ABS) – Enter the absolute measure of the vapor pressure for the fluid. Typically this is a value
that is looked up in a table for a given temperature of the fluid. This value is only used for certain slugging
simulations and if the user wants to permit column separation of the fluid. Column separation occurs when a low
pressure is developed in the fluid and vapor may come out of solution and form an air pocket in the line. These
air pockets can potentially cause pressure spikes when they implode, and if the large cavity model is activated
they can travel down the pipe in a “slug-like” manner. The vapor pressure is an input that has a special purpose
when modeling traveling slugs. The user should review the section on slug simulation for more on the use of this
input in that instance.

The user can always edit the user data screen by entering 8-Flow Elements

And then F9-User Fluid Properties

to bring up the fluid properties data sheet whenever it should be reviewed or changed. Any of the individual user
data can be edited or changed via the 8-Flow Elements data screen.

Search (Hit F11 from the Pipe Element menu)

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.7


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The search menu window allows searching through the element list for single nodes or for elements that contain
at least two nodes. Repeated searching for a node or element will find the next occurrence of that node or
element. The user can also employ the TextOUT feature to write out a TEXTIO.TXT file in the current work
folder that contains a semicolon delimited file of the screen input data. From this file the user can also do a
variety of searches or changes. Once the TEXTIO.TXT file is modified the user can read it back in to replace the
current input or add it to the end of the already current input. The semicolon delimited fields were setup to work
with excel.

Multiplier (Hit F1 from the Pipe Element Screen)


The multiplier screen shown below allows the user to enter the resistance of affixed number of 45, 90 or 30
degree elbows to any particular straight section of pipe. This input is used to let the user enter a single long
straight section of pipe and the approximate number of fittings to get a reasonable resistance model of a section
of the flow network that is not important to model. Often a long straight section with a number of bends will be
entered followed by a long pipe boundary condition. The long pipe boundary condition prevents the reflection of
pressure or flow waves, and the long straight section with fittings provides for a proper flow resistance for other
parts of the model, i.e. pumps or relief valves.

Enter the number of 90, 45 or 30 degree bends on the respective straight section of pipe. Note that the node
numbers for the possibly two sections of straight pipe that can be defined on each pipe data screen are given at
the top of the “Multiplier” form.

Unit Converter

The unit converter screen can be accessed from most of the BOS Fluids screens. This screen is intended
to help the user convert from metric to English units, or from one set of units to another when preparing
input. The user can change any of the data cells. All items on the screen are updated when one value is
changed.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.8


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Fluid Boundary Conditions


(F3 from the Pipe Element data Screen or F3 from the Flow Elements Data Listing Screen.)

Boundary Condition Node


Specify the point in the model of the piping system where fluid boundary condition applies. These are
typically points of specified head(pressure) or flowrate. (See F4-Equipment for alternative methods of
specifying equipment-type fluid boundary conditions.)
Dead ends are found and applied automatically by BOS Fluids to any node in the model connected only to
a single element that does not already have any other fluid boundary condition defined.
BE SURE TO SPECIFY FLUID BOUNDARY CONDITIONS WHERE YOU KNOW THEM!!
Only a single boundary condition is allowed per node, i.e. either the flow or the head may be specified. If
the pressure is specified at a node then the program will calculate the flow required to support that
pressure. If a flow is specified at a node then the program will back-calculate the pressure to support that
flow.
Usually at least two boundary conditions must be specified for each model, and one of those boundary
conditions must be a pressure. BOS Fluids can solve for the flows from pressures, but it must have at least
one pressure at a boundary condition to solve for the surrounding flows. This is demonstrated in the sketch
below.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.9


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Boundary Action
A fluid node can be of the following different types:
FIXFLOW The flow into OR out-of the system at the specified node is fixed at a given value for
all time. When FIXFLOW is specified the flowrate is fixed BOTH for steady state and
for transient calculations. A FIXFLOW of 0.0 is the same as a DEADEND. A positive
value of the flow for a FIXFLOW boundary puts fluid into the system, and a negative
value of the flow for a FIXFLOW boundary withdraws fluid from the system. (For
conditions where the user knows the flow in the steady state condition through a
piece-of-equipment, but is not certain that the flow will be maintained during the
pressure buildup of a transient, then the EQUIPMENT element model can be used to
provide a fixed resistance, but variable flow boundary condition.)
FIXHEAD The head (or pressure) at a particular node in the system is fixed for BOTH the
steady state and transient conditions. The pressure specified at the FIXHEAD point
should NOT include any pressure head due to the elevation of the fluid. The pressure
specified is the so-called “static-head,” the pressure measured by a Pitot tube that is
held at right angles to the flow. The total energy of the flow is equal to this entered
pressure head plus the flow velocity squared divided by the acceleration due to
gravity (the dynamic head). (The sum of the static head and the dynamic head equal
the total pressure – which is measured by a Pitot tube that is turned into the flow, i.e.
that measures line pressure and the component of the flow that brings the fluid to a
rest.)
DEADEND There is no flowrate past this point. DEADENDs are assumed at each node in the
piping system where only a single element is attached that is not predefined by some
other nodal entity.
LONGPIPE A long pipe connection is a boundary condition that does not reflect either pressure
or flow waves. This connection simulates the attachment of an infinite amount of
straight pipe. Any wave passing this boundary is assumed to continue on forever
without returning. LONGPIPE boundary conditions can use a linear or second order
approximation of the incoming flow properties to keep the wave from reflecting.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.10


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

LONGPIPE is used for the LINEAR approximation (2 terms), and LONG2PIPE is


used for the second order approximation, (3 terms). When using LONG2PIPE the
user should only be sure that the attached pipe has at least two reaches. (See the
Pipe parameters report for how many reaches are in a single pipe length.)
The LONGPIPE boundary condition is used when the user is concerned with local
disturbances and does not want to be concerned with reflections from far-away
locations. LONGPIPE is neither a FIXHEAD or a FIXFLOW type boundary although
the user can specify either the initial pressure or flow at the boundary node. (If both
pressure and flow are specified then the flow will be ignored.) The pressure or flow is
used in the steady state solution to establish the basic flow pattern. Once the
transient begins, both the pressure and the flow can change at the LONGPIPE
boundary condition to accommodate the incoming pressure or flow wave
without reflecting it.
The longpipe boundary condition requires a few time steps to establish the state of
the transient flow. During this time there may be small perturbations in the flow. The
user should check for the presence of these disturbances and start any transient of
concern after these perturbations have dissipated.
LONG2PIPE Second order LONGPIPE approximation of the incoming waveform. The user must
make sure that there is at least two reaches in the attached pipe when LONG2PIPE
is used.
SONICHEAD Used to allow the user to specify the sonic FLOWRATE at a choked location. The
steady state solution will calculate the resulting pressure to support the specified
flow, and then apply that pressure as a constant during the transient solution. Used
to specify the boundary condition at a high-pressure gas letdown station.
TIMEHEAD The head at the node varies with time during the transient analysis. The head as a
function of time should be specified in the field given for it below. When entering a
TIMEHEAD boundary specification, the steady state value of the pressure should be
entered in the FIXHEAD field, and the pressure-time history should be entered in the
TIMEHEAD field. Up to 10 different pressure vs. time curve points may be entered.
Up to 500 pressure or flow vs. time points may be entered for an individual time or
flow head in an optional text input file. The user should imitate a shorter TIMEHEAD,
or TIMEFLOW history generated by the program and then add the additional points
using an ascii editor such as notepad. Up to 500 points may be added in this way.
The ascii text input should be stored in the model data base in a file with a .TXT
extension. BOS Fluids is instructed to read this file on the Optional data form.
BOS Fluids will linearly ramp the head or time between changing times. For example
in the input below:
Time Head
0.0 sec 100 psig
1.0 sec 100 psig
2.0 sec 200 psig
3.0 sec 200 psig
The pressure will remain constant up until 1.0 seconds. Then it will linearly increase
to 2.0 seconds where it will remain constant for the next second at 200 psig.
TIMEFLOW The flowrate "to" or "from" the node varies with time. The flowrate as a function of
time should be specified in the field available for it below. The steady state flowrate
should be entered in the FIXFLOW field.
HARMONICHEAD The head at the node oscillates in a periodic way during the transient. The mean
head of the flow should be given below along with the frequency of the oscillation and
the percentage of the mean value of the head that is varying. The flow may also be
started up over some period, (ramped), and shutdown over a different period.
HARMONICFLOW The flowrate at the node oscillates in a periodic way during the transient. The mean
value of the flowrate should be given below along with the frequency of the oscillation
and the percentage of the mean value of the flow that is varying.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.11


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Flowrates are positive if TOWARDS the node, i.e. into the system, and negative if AWAY from the node,
i.e. OUT of the system.

Harmonic Pressure/Flow Time Waveform Filter

Flowrate (Volumetric) (cu.ft./sec.) or (cu.m/sec)


Enter the desired fluid flowrate for FIXFLOW, HARMONICFLOW, TIMEFLOW, SONICHEAD or LONGPIPE
For a node in the piping system this is the supply of fluid ADDED to the system if positive, and TAKEN
FROM the piping system if negative. For LONGPIPE either the flowrate or head may be entered. If both the
head and the flowrate are given then the head will be used and the flowrate will be ignored.)

Head (Height of Liquid Column or Force-per-Area)


Enter the desired system pressure in the appropriate units.
Height of Fluid Column: SI=meters English=feet
Force per Unit Area SI=bar English=psi
The head SHOULD be entered if the boundary condition type is FIXHEAD, HARMONICHEAD OR
TIMEHEAD. For HARMONICHEAD or TIMEHEAD the head or pressure value entered into this column is
used to establish the steady state value of the system flows. For LONGPIPE either the flowrate or the head
may be entered. If both the head and the flowrate are given then the head will be used and the flowrate
ignored.
The pressure specified in BOS Fluids is the so-called “static” pressure. The “dynamic” pressure is found by
BOS Fluids when the velocities are computed, and the so-called, “total pressure” is the sum of the static
and dynamic pressures.

Time for TIMEHEAD (sec)


Enter the INCREASING times in the head vs. time curve.
Input in the TIMEHEAD, Time field will only be used if the Boundary Action specified above is TIMEHEAD.
The time starts from 0.0, and this point can be omitted if desired. (The zero time head is taken from the
FIXHEAD input above.) There is a maximum of 10 time history data points available from the input screens
although up to 500 may be entered using the text input file method described above. Times entered do not
need to start at 0.0 seconds. The steady state solution is assumed to exist at 0.0 seconds and the user
only needs to enter changes from that condition. If there are questions about data entry formats, the user is
suggested to build a small test model and apply the time history in question. In the output the time history
for that particular node can be plotted and the user can see exactly how the program interpreted the
particular input data set. BE SURE TO DEACTIVATE MAXIMUM PRESSURE TRAPPING when checking
boundary condition inputs. To produce readable reports the pressure is only reported at the user input
nodes. Internally the maximum pressure may occur at some other point in the pipe however, and if this
happens then the maximum pressure will be reported at the node closest to the point where the actual
maximum pressure occurred. This maximum pressure trapping can disturb the maximums reported for a
boundary condition point and so should be turned off to see just the pressures at the nodes. THIS IS TRUE

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.12


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

WHENEVER THE USER IS ONLY INTERESTED IN PRESSURE AT A SINGULAR POINT AND NOT THE
MAXIMUM PRESSURE AT ANY POINT IN THE CONNECTED PIPE SEGMENT

Head for TIMEHEAD


Enter the head values that correspond on a 1-to-1 basis with the times entered above. (These two sets of
values define the time history of the nodal boundary pressures.)
This input will only be used if the Boundary Action specified above is TIMEHEAD.

Time for TIMEFLOW


Enter the INCREASING times to describe the flow vs. time boundary condition. Up to 10 points in the flow
vs. time history may be entered. Up to 500 may be entered in the ascii format input file mode. This input will
only be used if the Boundary Action specified above is TIMEFLOW.

Flow for TIMEFLOW


Enter the flowrate values that correspond on a 1-to-1 basis with the times entered above. (These two sets
of values define the time history of the nodal boundary flowrates.)
This input will only be used if the Boundary Action specified above is TIMEFLOW.
HARMONICFLOW OR HARMONICHEAD
For HARMONICFLOW or HARMONICHEAD boundary conditions enter the frequency of the flow or head
oscillation and optionally the phase angle on the first line in the text cell, and the amplitude of the oscillation
as a percentage of the mean flow or head on the second line. To minimize transient effects the harmonic
flow or head boundary can have a ramp superimposed on the sinusoidal excitation during startup and can
have a specified stop time and shutdown duration. These inputs are shown schematically on the figure
below:

HARMONICFLOW or HARMONICHEAD Startup Duration (sec) – See the figure above. This is the linear
startup time for HARMONICFLOW or HARMONICHEAD boundary conditions. This specification is used to
soften the startup transient.
HARMONICFLOW or HARMONICHEAD Stop Time (sec) – See the figure above. This is the total time for
the steady state portion of the harmonic oscillation. The total constant amplitude harmonic oscillation will
occur for [(stop time) – (startup duration)] seconds.
HARMONICFLOW or HARMONICHEAD ShutDown Duration (sec) – See the figure above. This is the
rampdown time after the stop time has been encountered during the transient. Used to simulate
reciprocating pump rundown.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.13


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Equipment Data –The equipment element description is accessed from the pipe element form by
hitting F4 or from the 8-Flow Elements menu by hitting F4. The equipment data form is shown below.

The Equipment Element feature is allows the user to simulate the transient effect of heat exchangers, coolers
or similar equipment that has a known flow rate in the steady state condition, but a fixed resistance to flow in
the transient condition that is not known a priori.
It is not unusual to know the flowrate of cooling water through a heat exchanger, but to be unsure of the exact
pressure drop given other system conditions. The Equipment Model lets the user specify the flowrate for the
steady state condition, and then finds the required pressure drop that can be accommodated by the
remainder of the piping system that to provide that flowrate. From this calculation, the net resistance to flow is
determined and used in the transient simulation to interact with the changing pressures and temperatures.
Equipment Nodes
Enter the nodes on the equipment in the steady state flow direction. The equipment model is used most
often to describe large boundary conditions, i.e. heat exchangers, condensers, plant connections, etc. that
will be used in the transient simulation where a FIXFLOW boundary condition is too restrictive in terms of
transient pressure wave evaluations. (See sketch below.)

From the steady state solution the friction characteristics (flow resistance) of the equipment are determined
and used in the transient solution to properly consider the effect fluctuations in head and flowrate on the
system performance.

Equipment Identifier
Enter an up to 10 character alphanumeric identifier for the piece of equipment being modeled.
Output for the equipment will be reported on a component identifier basis. 2D plots of the flow, velocity and
pressure drop will also be available based on the equipment identifier entered.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.14


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

DO NOT USE SPACES OR COMMAS IN THE EQUIPMENT NAME.


Flowrate (Volumetric) (cu.ft./sec) or (cu.m./sec.)
Enter the flowrate through the equipment that should be established in the steady state condition. This is
the steady state fixed flow from the first node to the second node. (THE ORDER OF THE NODE
DESCRIPTIONS ON THE EQUIPMENT DATA SCREEN DETERMINE THE DIRECTION OF FLOW.)
From the steady state solution a global friction factor for the equipment will be determined and used in the
transient solution. For the transient calculation the equipment is modeled as a constant flow resistance.

Valve Data – The valve data screen allows for a variety of in-line flow restriction devices. Common
BUTTERFLY, BALL, GATE and GLOBE valves can be modeled along with damped or undamped CHECK
valves, damped or undamped RELIEF valves, vacuum breakers (AIR), pressure REGULATOR valves, and
exchanger tube RUPTURE components. The component is identified by entering the starting and stopping node
for the valve. The Valve Data screen can be accessed from the Pipe Element screen by hitting the key F5, or
from the 8-Flow Elements menu by hitting the key F5. Each of the available valve types and a discussion of the
input for each is given below.

Valve Type
Available Valve Types are:
BUTTERFLY – Wafer type butterfly valve that provides a resistance to flow as a function of opening
position.
BALL – Full or partial port valve that provides a resistance to flow as a function of opening position.
GATE – Common valve that provides a resistance to flow as a function of opening position.
GLOBE – Moving spherical valve that provides a resistance to flow as a function of opening position.
RELIEF – Valve that permits flow once a set pressure is exceeded. Relief valves can be theoretically
instant open devices, or may incorporate the effects of mass, stiffness and damping. Additionally the user
may specify the set, rated and reseat pressures and omit the mass stiffness and damping values. If the set,
rated and reseat pressures are entered, the valve performance will be adjusted to satisfy these settings
and in so doing will simulate the effect of mass, stiffness, and damping. See the section on the relief valve
model below. BOS Fluids may be used for safety relief valves – used for compressible and incompressible
fluid surface, and relief valves used in incompressible fluid service. The methods in BOS Fluids may not be
well suited where incompressible service is to be simulated and the results should be checked carefully.
There are two capabilities in BOS Fluids that can be used to simulate safety relief valve gas behavior. The

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.15


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

method described in this section employs the method of characteristics, constant density and restrictions
on flow velocity to simulate gas flows. The BOS Fluids Gas Transient module described elsewhere
incorporates a thermodynamic model of the gas be exhausted and will predict Joule-Thompson cooling
effects in the relief piping. The user should review this and other documentation carefully when addressing
problems involving safety relief systems.
CHECK – Valve that permits flow in only one direction that is a function of valve opening position. The
check valve can be instant open and instant close, which is the ideal check valve, or the user may enter
mass, stiffness and damping to simulate the effect of real check valves. Mass, stiffness and damping slow
the performance of the valve and in some cases allow for excessive backflow through the valve which can
then result in large waterhammer loads when the valve finally closes. See the section on check valve
modeling below.
AIR – Valve that permits a local air pocket to be collected at the valve location when a low pressure is
experienced in the pipeline. The air is later expelled when the line pressure returns.
REGULATOR – Valve that opens and closes based on a sensor node at some other point in the system, in
an attempt to maintain a set pressure at the sensor node. The response time of the regulator can
determine how effectively the valve can interact with the system to control the transient. A regulator
response time is determined by the properties of the valve. See the section below on pressure regulators.
RUPTURE – Valve type that simulates a single or multiple tube rupture in a heat exchanger. See the
section on tube rupture simulations below.
Currently all of the types BUTTERFLY -through- GLOBE
BUTTERFLY, BALL, GATE or GLOBE valve types all support transient position control, and all have the
same ability to provide fluid resistance characteristics. The inputs described below apply to each valve
type unless specifically stated otherwise.

Valve Identifier
Enter an up to 10 character alphanumeric identifier for the valve. DO NOT USE SPACES OR COMMAS IN
THE IDENTIFIER.
This identifier will be used to cross reference the valve transient characteristics with its steady state
parameters. 2d plots are also available based on the valve type and all tabular report is given using the
Valve Identifier.

Valve Discharge Coefficient


(Default = 0.5)
This is a measure of the friction loss across the valve. Typical values are between 0.0 and 1.8. Examples
are given below:

The user is cautioned that should the major head loss in the system be attributable to this valve, and
therefore be greatly affected by the magnitude of the discharge coefficient, then a range of values should
be studied unless test, or other data is available.
If a resistance coefficient "K" is given, then the relationship between Cd and K is given by: Cd = 1/sqrt(K).
If a Cv is given, then the Cv must be converted into the appropriate dimensionless Cd or K values. Cv is
defined in terms of volume per minute, typically U.S. Gallons per minute. The relationship between Cv and
Cd is:
Cv = (58,880) (Cd)(A) … A in sq.meters.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.16


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Cv = (38.0) (Cd(A) … A in sq.inches


<or>
Cv = (29.9) d2 / (K)0.5 … d in inches.
For example, a 12” full bore valve with a Cd of 0.3 has a K value of 11.1 and a Cv of 1248.
gallons/minute/psi.

Valve Bore
Largest inside diameter of the valve measured normal to the flow cross section. If omitted, the valve bore
(id) will default to the inside diameter of the pipe.
This is the inside diameter associated with the effective flow area for relief valves and should not be
defaulted to the ID of the incoming pipe. The user should get this value from the manufacturer and
enter it for the valve!

Valve Closure Exponent


Used only for transient solutions.
For GATE, GLOBE, BUTTERFLY and BALL valves, this value determines how the valve flow area varies
with amount of valve closure as a function of time.
Typical values range from 0.15 -to- 1.0. Values for the closure exponent that are less than 1.0 cause the
majority of the flow cutoff to occur at the beginning of closure, while closure exponents larger than 1.0
cause the majority of flow cutoff to occur at the end of closure. (See the figures below.) The default value
of 0.2 simulates a wide variety of typical gate or swing valve closure behavior since the flow will tend to
slow the closing as the flow area decreases, and will tend to speed the opening as the opening starts. The
user should either check the behavior for a particular valve, or make sure, using a sensitivity study, that the
valve exponent entered will not significantly affect results.
Many valve manufacturers will provide Cv values for a range of valve opening conditions. The Cv constant
is the number of gallons per minute for a one psi pressure drop across the valve. From above the basic
equation relating Cv and CdA is:
[Cv]Full = (co)[(Cd)(A)]Full
And for the case where the valve is half opened:
[Cv]half = (co)[TAU]half [(Cd)(A)]Full
TAU at half opened during a full valve cycle is (t/Tc)em so that the ratio of [Cv]half over [Cv]full is:
[Cv]half / [Cv]full = (0.5)em
This can be solved for em from the manufacturer’s values for Cv:
em = (-3.322) log10( [Cv]half / [Cv]full )
The steps for entering accurate valve data from manufacturers Cv data is as follows:
Step 1: Find Cd for the full flow case from Cd = (Cv /[ (58,880)(A) ]) where A is in meters squared, or
from Cd = ( Cv / [(29.9)(d2)] where d is the inside diameter in inches. This is the value of Cd that should be
entered into BOS Fluids.
Step 2: Find the closure exponent by the equation: em = (-3.322) log10( [Cv]half / [Cv]full )

Basic Valve or Non-Pipe Flow Equation:


The flow through the valve is related to the pressure drop across the valve by the equation:
Q = (Cd)(A)[2gH]0.5
where Q – flow through the valve. (cu.ft./sec) or (cu.m./sec)
Cd – valve discharge coefficient (unitless)
A – Cross section inside area of the valve.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.17


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

g – gravitational constant
H – Pressure drop across the valve measured in fluid head.
As the valve is opened or closed, the value of the (Cd)(A) product varies and can be expressed:
(Cd)(A) = (TAU)[(Cd)(A)]@100% OPEN
The variation of Tau with time during opening or closing is calculated during the transient by:
TAU = TAUI – (TAUI – TAUF)(t/Tc)em

where t – is the time from the start of the change of state of the valve.
Tc - is the total time for the change of state.
em – is the valve coefficient
TAUI - %Opening at the beginning of the change of state
TAUF - %Opening at the end of the change of state.
TAU for three different values of “em” is plotted below for a valve closing and opening. (100% -to- 0%)

(Closing) em = 0.2 (Closing) em = 1.0 (Closing) em = 2.2


When em=0.2 the valve change of state starts rapidly, but ends slowly. For em=1.0 the change of state is
linear, and for em=2.2 the change starts slowly but finishes rapidly.

(Opening) em = 0.2 (Opening) em = 1.0 (Opening) em = 2.2


On opening (the same as for closing), small values of “em” cause the change to start rapidly and end more
slowly. To summarize
Value of em Opening Closing
Small Start Fast – End Slowly Start Fast – End Slowly
1.0 Linear Linear
Large Start Slow – End Quickly Start Slow – End Quickly

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.18


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The user can specify up to 10 valve change states in the standard input data screen, or up to 20 may be
entered in an ascii data file. (See Chapter 2 Section 8 the ACTION command and the bottom of the
OPTION form for ascii file input.) Each state change that represents a different opening for the valve will
invoke the TAU calculation described above to compute the (Cd)(A) product for the valve. An example is
given below:

Time ValveOpening TAUI TAUF Duration


0.0 100. 100 50 0.25
0.25 50
1.4 50 50 20 0.55
1.95 20 20 0 1.35
Most typical BOS Fluids applications involve the full opening or closing of a valve at a particular point in
time, and for generic valves an em=0.2 value most closely resembles their opening and closing
characteristics.
For check valves the closure exponent is used to determine if the user wants a
theoretically instant-open, instant-close check valve or a more realistic one. If
em=0, or blank, then an instant open valve will be modeled. For any value greater
than zero, a damped value will be simulated.

Instant Opening Non-Instant (em=0.2) Non-Instant (c1=0.1 ft.lb.sec/rad, em=0.2)


When a CHECK valve is instant opening (or closing), the direction of flow is used to determine the state of
the valve. When a check-valve is Non-Instant Opening, the mass, stiffness and damping of the valve are
used in addition to the flow state surrounding the valve. A positive flowrate calculated will force the check
valve opened. (Although the check valve may close of its own accord if the pressure holding the valve open
is less than the set pressure for the valve plus the weight of the disk. The equation to be satisfied is:
(I)d2a/dt2 + (c)da/dt + Tarm - (Kdi2/4)(cos a+da/2) = (dp)(Ai)(di/2) + w(di/2)cos( a ) + (pset)(Ai)di/2
where Tarm is the torque due to a lever arm and external damper, and the flow area Ai is raised to the
power (em.) The angle (a) is zero when the valve is fully opened, and 90 deg. when the valve is fully
closed. (This is what produces the minus sign in front of the stiffness term in the equation above.)
User’s can estimate check valve damping using a part of the equation above:
(c)da/dt = (dp)(Ai)(di/2)
If a valve should close (or open) in one second, then da/dt = 90 deg./1 second. Calculations are in radians
so da/dt = 1.5708 rad/sec to close or open the check valve with a constant angular velocity in one second.
If the approximate “average” pressure across the valve will be 5 psi, then the damping constant can be

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.19


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

calculated from:
C = (dp)(Ai)(di/2) / (da/dt)
The example problem in the \bosmodles directory chk3 demonstrates this calculation. Note that the valve
does not open in 1.0 seconds because the pressure across the valve is not a constant 5.0 psi. For the
“chk3” model, Ai=132.7 sq.in., di=13 in. and the pressure drop driving the valve closure is 5.0 psi. The
valve damping coefficient is then:
C = (5)(132.7)(13/2) / 1.5708) = 2745.6 lbf.in.sec./radian.
BOS Fluids uses the damping coefficient in terms of lbf.ft.sec./radian, so to use the value for c, it must be
divided by 12. (See chk3 for this input, i.e. 2745.6/12 = 228 lbf.sec.ft./radian.)
User’s may be given coefficients for hydraulic arms, etc, but typically the user must experiment with a
range of damping values to determine which value causes the valve to close in the period anticipated. If
only the nonzero valve exponent is entered, BOS Fluids will enter a mass for the valve if the flow is
reversing direction. This will close the valve. Also a small rotational damping amount is entered if the user
has not entered one. The default values used are calculated by the program and are printed in the
transient message section of the output.
If damping and mass is not provided the program will select small values that may result in valve flutter.
The user is encouraged to try to estimate reasonable values for the mass of the disk or ball and the the
damping coefficient (c1).
The “pset” value for check valves is only applied to initiate opening of the valve. Once the valve is open and
moving in a further open or further closed position, “pset” is not used.
For Pressure Safety Valves the valve exponent is used only when the mass, stiffness, or damping is
greater than zero. Typically the user will not have the mass, stiffness or damping of the valve, but will have
the set, full flow and reseat pressure. When the set, full-flow and reseat pressures are given, the flow
through the valve will be calculated using the following equation which does NOT include the valve
exponent.
flow = (C) [(P-Pset)/(Pfull-Pset)]
where:
flow- is the flow through the PSV.
C- is a constant
P- is the line pressure
Pset- is the valve set pressure
Pfull- is the full flow pressure for the valve
em- is the valve coefficient.
When the stiffness, mass or damping of the valve is entered then the linear displacement of the valve seat
is multiplied by the opening of the valve raised to the (1/em) power.

Valve Opening Percent


Enter values between 0.0 and 100. This is the starting position for the valve. The value entered here will
be used in the steady state solution unchanged and then will be the initial value taken in the transient
solution.
This percentage is the ratio of the open flow area to the total flow area for the valve.
Transient changes to a valves position can be input in the "ACTION" screen under the "Transient Valve
Data" section by hitting the F2 key, or the user can enter the 8-Flow Elements screen and use the F6-Valve
Transients selection.
The valve opening percent is the initial value of TAU for the valve. This input is not used for CHECK,
RELIEF, AIR INLET OR TUBE RUPTURE valves. The initial steady state opening for PRV valves is
determined in an iterative calculation during the steady state solution. The initial value entered by the user
is used as a starting guess for the iteration.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.20


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

CHECK Valve Direction (PLUS/MINUS)


This input is required for CHECK valves and defines the direction of allowed unrestricted flow.
Enter "PLUS" if the allowed direction of flow is from the FROM node on the element to the TO node on the
element.
Enter "MINUS" if the allowed direction of flow is from the TO node on the element to the FROM node on the
element.
This input item is not used for any other valve type.

Safety Valve Set Pressure


Enter the value of the set pressure for the relief valve. The set pressure is the pressure at which the valve
begins to open. This is a required input for the relief valve and is the point where the valve just begins to
life.

Safety Valve Full Flow Pressure


Enter the value of the differential line pressure that will support full flow through the relief valve. The
pressure safety valve will only be partially open when the driving line pressure is between the set and full
flow pressures.
GAS APPLICATIONS ABOVE 14 psi or 1 bar, TYPICALLY REQUIRE SONIC MODELING TECHNIQUES.
BOS Fluids includes a choked flow reduction to satisfy the sonic flow requirements at the valve orifice but
the user should be sure that the model is sufficiently accurate for the intended purpose.
Safety Valve ReSeat Pressure
When a safety valve closes, the pressure at which the valve shuts is typically different from the opening
pressure, and is typically about 10% lower. Enter the reseat pressure with the full flow pressure and BOS
Fluids will operate the relief valve within these ranges. The difference between the set and reseat pressure
is often called the blowdown pressure.

Special Valve Types


There are several special valve types that require or allow added inputs and considerations, and they are
listed and described below:

RELIEF VALVES
CHECK VALVES
RUPTURE MODELS
PRESSURE REGULATOR
AIR VALVE

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.21


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

SAFETY-RELIEF VALVES
The amount of mass produced and the ability of the relief valve to exhaust that mass are hand or simple
programmatically problems that should be addressed outside of BOS Fluids. Many of the peculiar problems
associated with relief valve flow that can be analyzed by a method-of-characteristics solution can be
analyzed by the BOS Fluids Liquids and Gases Model. A special steady-state Gas Transient model
exists for the steady state temperature prediction of downstream temperatures in relief piping.
Safety valves must be sized properly using manufacturer’s methods for the intended quantity of flow to be
discharged. This maximum relieving capacity should be stamped on the valve. Methods in B31.1 Appendix
II can be used to verify that the BOS Fluids model for steady state conditions is suitable. For closed
systems B31.1 recommends that the design pressure in a closed system is multiplied by 2 since choking
can occur in the downstream tailpipe during the transient introduction of steam into an air-filled exhaust
line. Oversized relief valves will chatter and undersized valves will permit an excessive pressure buildup.
To prevent chatter most Codes limit the pressure drop in the inlet piping to 3% of the set pressure on the
valve. Standard relief valve sizing techniques address the primary issues with relief valves and should be
employed whenever BOS Fluids is used to study a valve installation. The BOS Fluids relief valve model
has several special purposes however:
1) Calculation of the pressure pulsations due to valve operation.
2) Determination if the relief valve can open quick enough to provide pressure protection.
3) The Gas Transient Steady State capability predicts the temperature profile in the relief valve
discharge line to be sure that ductility limits are not violated.
4) Ability to discharge from one or more fixed volumes.
5) Pump control system failure.
6) Prediction of overflow due to valve closure. (For example, at quick-disconnect couplings for
loading fluid loading services.)
7) Estimation of manifold backpressure during HAZOP study. (When one valve opens into a
manifold that collects the outflow from numerous safety valves, the backpressure from the one
valve discharge might increase the back pressure on other valve outlets preventing their proper
opening.) Note that changes in gas densities in relief exhaust piping is typically large however,
and the effect of the change in diameter is not seen in the BOS Fluids Liquids and Gases
program.
The pressure drop in inlet lines that can cause chattering in relief valves can be predicted by BOS Fluids.
Most sizing methods for relief valves include hand evaluations of this pressure drop, however where
complicated piping configurations are employed, and where acceleration effects can be significant the use

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.22


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

of BOS Fluids can assure that excess pressure drop will not occur and produce chatter.
Per API 520 2.X.X “When a pressure relief valve is installed on a line directly connected to a vessel, the
total non-recoverable pressure loss between the protected equipment and the pressure relief valve should
not exceed 3 percent of the set pressure of the valve, …”
“When a pressure relief valve is installed on a process line, the 3 percent limit should be applied to the sum
of the loss in the normally non-flowing pressure relief valve inlet pipe and the incremental pressure loss in
the process line caused by the flow through the pressure relief valve. The pressure loss should be
calculated using the rated capacity of the pressure relief valve. Keeping the pressure loss below 3 percent
becomes progressively more difficult as the orifice size of a pressure relief valve increases.”
“An engineering analysis of the valve performance at higher inlet losses may permit increasing the
allowable pressure loss above 3 percent.”
“The designer is responsible for analyzing the discharge system to determine if the reaction forces and the
associated bending moments will cause excessive stresses on any of the components in the system.”
“Pressure relief devices that relieve under steady-state flow conditions into a closed system usually do not
create large forces and bending moments on the exhaust system. Only at points of sudden expansion will
there be any significant reaction forces to be calculated. Closed discharge systems, however, do not lend
themselves to simplified analytical techniques. A complex time history analysis of the piping system may be
required to obtain the true values of the reactions forces and associated moments.”
The BOS Fluids empty pipe slug capability can be used in long, closed discharge systems to estimate the
transient effect of liquids rushing into an air filled relief header. The momentum change loads in this case
can cause significant loads on the piping system and should be evaluated.

If the spring constant is not known it can be estimated from:

K = (Prated-Pset)(1.2)(Aorifice)/[(0.27)(dorifice)]

The weight of the disk and assembly can be estimated from:

M = (2.5)(Aorifice)(0.3) = Weight in pounds, (Aorifice in inches)

A critical damping value of 3% is typically used for non-controlled relief valves.

The user employing the above should be cautioned that these are approximations only to be used when
actual data is not available, and that the resulting estimates should be checked carefully before being used in
any application.

CHECK VALVES
The check valve model can be either instant opening – instant closing or damped. The damped model
opens and closes more slowly than the instant opening model and thereby introduces flow perturbations. At
the discharge of a pump, for example, if the check valve closes slowly when the pump trips, reverse flow
can begin to move back through the pump and a waterhammer wave imitated when the valve finally shuts
terminating the backflow.
The damped check valve model will be used whenever the valve exponent is greater than zero. Even if all
other parameters defining the valve are zero, if the valve exponent is greater than zero a value for the
torsional damping will be entered to provide some small resistance to the check valves opening and
closing. The default value for torsional damping if no other check valve properties are entered are
calculated from:

IF the steady state pressure drop across the valve is zero THEN
C1 = (0.01)(g)(rho)
ELSE
C1 = (dH)(g)(rho)

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.23


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

ENDIF
3
Torsional Damping Constant = c1 * Di * pi / 8.0 * dt / 0.7 / 1000.0 … units (N.m)/(rad/sec.)

If the pressure across a 10” (254 mm) valve in the steady state solution is 1.5 bar, and the time step for
solution is 0.0005 seconds then:

C1 = (1.5)(105) = 1.5E5 Pa.


Torsional Damping Constant = (1.5E5)(0.254)3 (3.14 / 8) (0.0005 / 0.7 / 1000) = 6.89e-4 (N.m)/(rad/sec)

There are six inputs that can be used to control the speed of the check valve closure. Each is
demonstrated in the illustration below:

A positive flowrate will force the check valve opened. The initial state of the check valve is determined by
the steady state solution. If the steady state solution shows that the valve is opened entering the transient
calculation the Pset and Weight of the valve are changed to zero until the valve has closed at least once.
The equation to be solved is:
(I)d2a/dt2 + cda/dt + Tarm – Kdi2/4 cos(a+da/2) = (dp)(Ai)(di/2) + w(di/2)cos(a) + (pset)(Ai)di/2

where:

(I) = inertia of the disk


a = gate angle
c = torsional damping constant
Tarm = (Arm/dt) [ (cos(a) – cos(a+da)] (damp)(arm)(sin(a+da/2))
K = stiffness
di = inside diameter
dp = pressure drop across the valve
Ai = Inside area of the pipe
pset = Pressure that must be overcome to open the valve once it’s closed.
The user is encouraged to experiment with the values given to produce the valve closure times estimated
by the manufacturer. Small check valves with spring loaded balls can also be simulated by entering the
mass of the ball for the weight of the gate and the stiffness of the spring.
Note that the valve weight is entered in “weight” units. “Weight” units are newtons for metric unit users
and pounds for English units users.

RUPTURE MODELS
The BOS Fluids exchanger tube rupture model is based in some part on the paper, “High Pressure Gas-
Filled Tubing Rupture in Liquid-Filled Heat Exchangers,” by L.L. Simpson, given at the Sixty-fourth Annual
Meeting of the AICHE in December 1971, and on the API “Model to Predict Transient Consequences of a

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.24


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Heat Exchanger Tube Rupture,” 1976, and API RP 521, “Guide for Pressure-Relieving and Depressuring
Systems”. The phenomena described in the Simpson paper was extended by Tony Paulin and Dr. Hans
Bos using the method for air release described by Wylie and Streeter and as implemented by Andy van
Esch in BOS Fluids.
The problem addressed is described in the diagram below:

When a high pressure tube ruptures into a low pressure shell, there will be an increase in the pressure in
the shell that should be protected by a relief valve. Tube rupture and the resulting pressure release occur
on the order of milliseconds, and peak shell pressures are often attained in commercial heat exchangers
roughly 10ms after the tubeside rupture. (Simpson). Relief valves may begin to react in this time, but the
pressure wave must have time to get to the pressure relief valve before it can open it, and in this time the
pressure in the shell may increase.
API RP 521 2.3.13 States:
“In shell-and-tube heat exchangers, the tubes are subject to failure from a number of causes, including
thermal shock, vibration, and corrosion. Whatever the cause, the result is the possibility that the high-
pressure stream will overpressure equipment on the low-pressure side of the exchanger. Economical
design usually dictates that the higher-pressure stream flow through the tubes, since this can result in a
thinner shell, but this is not always the case. The pressure relationships must be known to evaluate the
results of tube failure properly. The ability of the low pressure system to absorb this release should be
determined. The possible pressure rise must be ascertained to determine whether additional pressure relief
would be required if flow from the tube rupture were to discharge into the lower-pressure stream.”
3.18.2 States:
An internal failure can vary from a leaking tube or tube sheet to a complete tube rupture where a sharp
break occurs in one tube. The exact nature of anticipated failure should first be defined for heat exchangers
or for tube assemblies inside a pressure vessel. For relatively-low-pressure equipment, complete tube
failure is not a viable contingency when the design pressure of the low-pressure side is equal to or greater
than two-thirds of the design pressure of the high-pressure side. Minor leakage can seldom result in
overpressure of the low-pressure side during operation. Conversely, complete tube failure is considered a
viable-contingency when the design pressure of the low pressure side is less than two-thirds the design
pressure of the high pressure side. However, if the high pressure side of the exchanger operates at 1000
pounds per square inch gauge (6895 KPa) or more and contains a vapor or a liquid that can flash or result
in vaporization of liquid on the low-pressure side, complete tube failure should be considered, regardless of

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.25


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

the pressure differential.”


The data required for the rupture model is listed below:
Time – Estimated time of tube rupture. Typically taken to be 1 ms. Entry would be 0.001.
Tube ID – Inside diameter of the tube. (in. or mm.)
No. – Number of tubes discharging into the low pressure side of the exchanger. API recommends that (2)
be used to simulate a worst case condition. If any operator has experience with greater numbers of tube
failures simultaneously, then any number of tube ruptures can be entered.
Cp/Cv – Ratio of specific heats for the gas on the tubeside.
R – Gas constant tubeside material. (Nm.Kg/degC or ft.lbf/lbm/degR)
Temp – Stagnation pressure on the tube side. (deg.F or deg.C)
Pressure – Stagnation pressure on the tube side.
Dens – Stagnation weight density on the tube side. (lb/cu.ft. or N/cu.m.)
Make sure that the time step selected for the transient solution is not smaller than the tube rupture time.
This generally will not be the case, but the user should verify that this requirement is satisfied.
The model determines if the tube discharge is sonic or subsonic and switches between the two as the
pressure changes in the shell side.
These inputs and the model building will be demonstrated by running the test provided in the API Model
document. Dimensions not given in the API document were assumed.

The tube bank resistance to flow must be estimated and a single equivalent area provided. The equivalent
internal area of the single effective tube must be the same as the total area of the tube bank.

Tube OD = 0.75
Tube ID = 0.532
Tube Area = pi/4 0.5322 = 0.222 in2
Total Tube Area = (150)(0.222) = 33.3 sq.in.
(pi/4) Deff2 = 33.3 sq.in.
Deff = 6.5 in.
Per Wylie:
f / D = fs / Ds

where f = friction factor for the effective diameter


D = effective diameter
Fs = friction factor for a single tube
Ds = Diameter of a single tube.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.26


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

From Wylie’s ratios, the effective friction is: f = fs (D/Ds)


The friction for a single tube can be found from a simplification of the Colebrook-White friction factor
expression:
1/ (fs)0.5 = -2log10( eps / (3.7 Ds) )
= -2log10( 0.002 / [(3.7)(0.532)]
= 5.986
fs = [1 / (5.986) ]2
fs = 0.0279

f = (0.0279) ( 6.5 / 0.532 )


= 0.341
Having the friction factor, the effective roughness can be calculated which will be input into BOS Fluids for
the effective tube bank element.
eps = (3.7)(Ds) [10] ( 1 / [(-2)(fs)^0.5] )
= (3.7)(6.5) (0.139)
= 3.348 in.
The thickness of the equivalent tube can also be estimated so that the speed of sound in the effective
diameter pipe will be equal to the speed of sound in a single tube. This is done by setting the thickness of
the effective diameter to the thickness of the tube times the diameter ratio:

teff = ( D / Ds ) ts
= ( 6.5 / 0.532 ) ( 0.109 )
= 1.33 in.
The BOS Fluids model for this exchanger-only geometry is shown below.

The tube rupture boundary condition will be applied at node 16. The channel will fill with the high pressure
from the shell side and the pressure will dissipate somewhat as flow continues down the tubes.
Tube rupture “valve” elements must be connected to two pipe elements. The high
pressure influx will be applied to the system in between the two pipe segment
ends. The end of the tube rupture element should have a fixed head pressure equal
to the tube rupture pressure.
Be sure to set the output time increment and the total time so that the transient of

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.27


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

interest is trapped. Usually this requires smaller output time steps and shorter
times than those that are typically used.

PRESSURE REGULATOR
The Pressure regulator valve adjusts its position during BOTH steady state and transient solutions in an
attempt to keep the pressure at a sensor node, removed from the valve at a fixed value. Cd is used for
pressure regulators just for any other valve. The value of the closure exponent gives the tolerance on
the set pressure. For example, to assume that a 1% variation of the set pressure is sufficiently
converged, the input for the valve opening coefficient would be 0.01. The opening percentage is used
on the startup steady state solution as a first guess at the valve opening. The “Time:” input for pressure
regulator valves is the estimated time it takes for the valve to go from a fully closed to fully opened
condition.
The valve may function in two modes: Overpressure Mode, or UnderPressure mode. In Overpressure
mode, the sensor node is typically downstream of the PRV valve. The PRV is closed when the pressure at
the sensor node exceeds the set pressure. The valve closes in an attempt to reduce the downstream
pressure.
In Underpressure mode, the sensor node is typically upstream of the PRV valve. The PRV is opened when
the pressure at the sensor node is greater than the set pressure.
Note that the Pressure Regulator Valve (PRV) will cause iterations to occur in the steady state solution.
This is because the steady state solution tries to find a valve location where the regulator opening is not
changing when the transient starts. If the steady state solution cannot find a PRV setting that holds the
pressure at the sensor node at its set value, then the user will likely find the PRV valve fully opened or fully
closed when the transient starts.
AIR VALVE
The air valve is used at a high point (or points) in a pipeline to protect the pipe from vacuum conditions.
When the line pressure drops below atmospheric pressure. A schematic of the air valve setup is shown in
the sketch below.

1) Note that a fixed head boundary condition must be placed at one end of the air valve. The fixed head
input is not used in the analysis, but the program must see a fixed head attached to the node opened to
the atmosphere so that it can properly orient the valve. When the pressure at the end of the valve
attached to the pipeline, (12 in the sketch above), drops below the fixed head pressure for the valve,
the valve opens and an air volume is taken into the pipe until the pressure at the valve attachment to
the pipe exceeds the fixed head pressure at which time the valve opens and the air that was previously
taken in is expelled. Note that there are two Cd’s associated with this valves behavior. One is for air

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.28


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

inlet, and the other is for expelling the air. Any air taken into the system is assumed to remain at the air
inlet valve location and should be smaller in volume than a single pipe reach. (The size of a pipe reach
can be found in the “Pipe Parameters” report for each element in the system.) The total flowrate of air
into and out of the system can be plotted in the 2D Plotted Results. User’s should inspect this report to
determine if the size assumptions are valid. User’s should be careful sizing air valves. The spacing and
size of the valve opening can determine the effectiveness of the valve. Vacuum breakers can be
undersized, or not placed at a sufficient number of locations along a pipeline. Each of these errors can
lead to collapse of the pipeline if the pipe is not designed for external pressure. The air valve model is
identical to the one described by Wylie & Streeter in Section 6-7, “Fluid Transients in Systems.”
Depending on the valve orifice size the flow out of (or in to) the valve may be subsonic or sonic. Sonic
flow conditions will limit the volume of air that can be expelled. If the size of the collected air pocket
becomes large the user should consider either installing more valves, or using the large cavity model to
allow for the entire pipe section to be evacuated. In this case, large pressure spikes can occur when
the flow is rejoined and the user must be very careful with this model.

Centrifugal Pump Data


There are three major areas in the centrifugal pump data screen and they are shown in the figure below.

Reciprocating pumps of single, duplex or triplex types are analyzed using a completely approach. See the
Chapter “Reciprocating Pumps and Code Compliance.”
In steady state the centrifugal pump runs at a fixed RPM if is started. For each speed the pump follows a
parabola on a plot of head vs. flow. A typical pump curve is shown below:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.29


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

BOS Fluids permits the input of three different pump curves for three different running speeds. The section in
the upper right section of the Pump input screen shown above is to let the user auto-generate pump curves from
the several basic parameters that govern centrifugal pump flow. If the user does not have a pump curve, then he
should fill in as much of the data in the top of the screen as is known and then hit the key F1 to generate a pump
curve. The curves auto-generated by BOS Fluids tend to be very good approximations.
BOS Fluids uses an extension of the method described by Wylie & Streeter in Chapter 7 of “Fluid Transients in
Systems.” An abbreviation of the Wylie & Streeter description of the pump transient problem is given below:
“The pump motor, by exerting a torque on the rotating shaft, conveys energy to the impeller to cause flow
through the pump and to develop a total dynamic head increase from the suction flange to the discharge flange.
… The instant that power is lost, the reactive torque of the liquid on the impeller causes it to reduce its rotational
speed, which in turn reduces [the head] and causes rarefaction waves to be transmitted downstream and
pressure waves to be transmitted upstream through the connected piping. If the system develops its head
primarily by lifting the liquid to a higher elevation, the flow at the pump reverses in a few seconds, and then a
short time later the rotation of the pump reverses, and it starts to run away backward. As it spins up to a high
reverse speed, it causes a high resistance to the flow, which develops a high head at the pump that must be
provided for in the design, [unless there is a check valve at the discharge]. There is also the possibility that the
discharge pipeline, due to its profile, may suffer low pressures, perhaps causing air to come out of solution and
vapor cavities to form. The collapse of the vapor cavities and compression of the air pockets may cause
dangerously high pressures to develop. If the pumping acts against a friction head, as in the case of long lines,
there are no adverse forces to cause reversal of the flow. The flow at the pump does decrease rapidly when
power is cut off, however, which may overpressure the suction line unless it is designed to withstand it. On the
discharge side of the system the slowing of flow at the pump causes a succession of rarefaction waves to move
downstream, causing air and other gases to come out of solution and liquid column separation. Depending on
the profile of the line these pockets of gas and vapor may cause high pressure upon closure. Also, the initial
low-pressure waves may collapse the pipe unless [it is] designed to withstand sub atmospheric pressures.”
Four quantities can be used to describe the performance of a pump:

1) Dynamic Head

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.30


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

2) Torque
3) Flowrate
4) Shaft Speed

Two of these quantities are independent and once known, can be used to determine the other two. Two
assumptions are used:
1) The steady-state characteristics hold for unsteady-state cases also. Even though the flow and speed
change with time, for example, the head and torque can always be determined from the instantaneous
values of flow and speed.
2) Homologous relationships are valid. (Test results from one pump can be used to analyze a different, but
mathematically similar pump.)

The Homologous Relations for head, flow and speed are:


H1/(N1D1)2 = H2/(N2D2)2
Q1/(N1D13) = Q2/(N2D23)
Where the subscripts “1” and “2” refer to two different-sized pumps of the homologous series.
The homologous theory assumes that the pump efficiency does not change with the size of the unit, so:
T1N1/(Q1H1) = T2N2/(Q2H2)
Equations can be written that relate data from one pump to another. BOS Fluids stores results for four different
pumps. The specific speed of the pump entered by the user is matched against the closest specific speed of the
data stored and the closest actual pump data is selected to describe the transient behavior of the pump.
This homologous data modeling approach is only used when the pump is started or stopped. When the pump is
running at a constant speed the pump will always be constrained to the pump curve specified. The specific
speed is given by: NQ1/2/H3/4.
Inertia of the shaft, impeller and the motor must also be entered. Wylie sites a linear regression analysis from
five pump manufacturers that gives the following equations for total rotational inertia.
Ip = 1.5(107)(P/N3)0.9556
Im = 118(P/N)1.48
The total rotational inertia is: I = Ip + Im. “P” in the above equation is the shaft power supplied to the pump in
kilowatts, and N is the rotational speed in rpm. “I” is in kg.sq.m. Note that the inertia value required by BOS
Fluids is in units of N.sq.m. so “I” from the equation must be multiplied by 9.81.
To restate the solution procedure for a pump shutdown:
1) The steady state solution is found first. The input pump curve is used to determine the location on the
pump curve that matches the possible states for the pump at the specified running speed.
2) The transient calculation is started from the pump steady-state solution. The pump that most closely
matches the specific speed of the user’s pump is selected from the data base and homologous relations
used to relate the user’s pump characteristics to the characteristics of the piping system attached to it.
(Simultaneous equations are iteratively solved that relate the flow state in the pipes adjacent to the
pump to the pump flow and head state.)
3) The interaction of the pump without power and the piping system surrounding it is solved iteratively for
each time step in the solution.
The data for each input cell is described below:

Pump Identifier
Enter an up to 10 character alphanumeric identifier for the pump. Do not use spaces in the name.
This identifier will be used to identify the pump in the tabular listings and in the 2d output plots.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.31


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Suction Node
Enter the node number that describes the inlet flange to the pump. The pump equations will be used to
describe the flow condition from the suction node to the discharge node. Pipe or valves can be directly
attached to any pump suction node, (but only a single pipe or valve). The user can also specify only a
boundary condition at the suction side if the inlet piping is not modeled.
Discharge Node
Enter the node number that describes the outlet flange from the pump. The pump equations will be used to
describe the flow condition from the suction node to the discharge node. Pipe or valves can be directly
attached to any pump discharge node, (but only a single pipe or valve.) The user can also specify only a
boundary condition at the discharge side if the outlet piping is not modeled. Most often a check valve is
attached directly downstream of the pump. This valve can be attached directly to the downstream
(discharge node) side of the pump.
Steady State Condition
Enter either “ON” or “OFF.” This switch determines the state of the pump for the steady state solution. If the
pump is “ON,” the steady state iteration will find a point on the running speed parabola that satisfies the
system flow-head characteristics. If the pump is “OFF” there will be no flow through the pump in the steady
state solution.
Pump Speed (RPM)
Enter the speed of the pump when running. For the steady state solution the pump will be set to this speed
when it is running and the appropriate location on the pump curve that corresponds to the system curve will
be found.
Rated Efficiency
Constant value used during the transient simulation to determine the pumps effectiveness in transferring its
rotary inertia into flow energy.

Typical values are between 0.6 and 0.9.


Force Moment of Inertia
The force polar moment of inertia of the pump impeller and its rotating assembly. This value is used only
during the transient solution. For SI units users the mass moment of inertia in Kg.sq.m should be multiplied
by 9.81 to get N.sq.m. For English units users the mass moment of inertia is the same as the force moment
of inertia, i.e. lbm and lbf.
Discharge Diameter
Used only for the F1-Build Pump Option. The discharge diameter interacts with other information in the
design data screen to improve the estimate of the pump flow/head/speed characteristics when the "Build
Pump" option is activated. The “Build Pump” option is only needed when the user does not have pump
curves for the pump being analyzed.
Suction Diameter
Used only for the F1-Build Pump Option. The suction diameter interacts with other information in the
design data screen to improve the estimate of the pump flow/head/speed characteristics when the "Build
Pump" option is activated. The “Build Pump” option is only needed when the user does not have pump
curves for the pump being analyzed.
Rotor Diameter
Used only for the F1-Build Pump Option. The rotor diameter interacts with other information in the design
data screen to improve the estimate of the pump flow/head/speed characteristics when the "Build Pump"
option is activated. The “Build Pump” option is only needed when the user does not have pump curves for
the pump being analyzed.

Increasing the rotor diameter increases the flow and head through the pump. Increasing the rotor diameter

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.32


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

also has a strong effect on the transient pump rundown solution by significantly increasing the estimated
polar moment of inertia and thus slowing down the rundown transient.

Pump flowrate is a third power of the rotor diameter, while pump head is a second power of the rotor
diameter. (See the homologous relations above.)
Specific Gravity of the Pumped Fluid
Used only for the F1-Build Pump Option. The specific gravity affects the estimated polar moment of inertia
of the pump rotor. The specific gravity is only used when the “Build Pump” option is activated. It is not used
otherwise. The “Build Pump” option is used only when the user does not have pump curves.
Discharge Head
USED ONLY WHEN BUILDING THE PUMP CURVE CHARACTERISTICS. Used only for the F1-Build
Pump Option The suction head is subtracted from the discharge head to get the rated flow for pump
design. Input of the discharge and suction heads and desired pump flowrate creates the best pump
models for use. The discharge head is only used when the “Build Pump” option is activated. It is not used
otherwise.
Suction Head
USED ONLY WHEN BUILDING THE PUMP CURVE CHARACTERISTICS. Used only for the F1-Build
Pump Option The suction head is subtracted from the discharge head to get the rated flow for pump
design. Input of the suction and discharge head, and the desired pump flowrate creates the best pump
models for use. The suction head is only used when the “Build Pump” option is activated. It is not used
otherwise.
Rated Flowrate
Used only for the F1-Build Pump Option. The input of the flowrate interacts with other information in the
design data screen to improve the estimate of the flow/head/speed characteristics when the "Build Pump"
option is activated.

The rated flowrate value provides a key data point for the pump curve. The rated flowrate should always
be entered when using the F1-Build Pump Option.
Pump Operation Curves

At least a single curve must be entered for the running speed of the pump.
Two or three pump curves may be entered for different speeds and the used curve will be interpolated from
the entered curve for the actual speed of the pump.
All of the data points above will be automatically filled in by BOS Fluids if the user hits the button: "F1-Build
Pump." The Build Pump option uses information in the "Auto Generated Pump Design Data" part of the
form to create a "typical" set of pump curves. The more information the user provides to the "Design Data"
part of the form in the upper right hand corner, the closer the pump design will be to the actual pump in
service.
If the pump curves are available, the flowrate values may be read and entered directly from the
manufacturer's pump curve. IF THE USER IS UNSURE WHAT VALUES SHOULD BE ENTERED INTO
THIS TABLE. HE should enter as much information as is available in the “auto-generation” box in the top
right of the screen and the running speed. Once the data is entered, then hit the F12 key to build the data.
Note the curve points generated automatically by the program. These points will be close to the actual

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.33


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

pump curve data.

Most pump curves give the head across the pump in feet, which does not change based on the liquid
pumped. To change the head from feet to psi the following relation should be used:
psi = Head X Specific Gravity / 2.31
The flowrate from the above curve is given in US gallons per minute. To get from US Gallons per minute to
cu.ft./sec.:
cu.ft/sec = (US GPM) X 0.002228
cu.m/sec = (US GPM) X 6.3083E-5
The three points for the 13.0 inch impeller shown on the pump curve above are:
Maximum Head = 38 ft. (16.45 psi.) Minimum Flow = 500 GPM (1.114 cu.ft/sec)
Rated Head = 27 ft. (11.688 psi) Rated Flow = 1500 GPM (3.342 cu.ft./sec)
Minimum Head = 14 ft. (6.06 psi.) Maximum Flow = 2500 GPM (5.57 cu.ft./sec)
These values are entered as shown below for water for the 880 RPM running speed:

There is no reason to be concerned with the “Rated Flow” if the user does not know exactly what the rated
flow will be. The user is encouraged, however, to select the “Rated Flow” point to be as close to the rated
flow as possible so that that part of the flow-head curve will be modeled as accurately as possible.

Build Pump Button


Move the cursor into this field and hit "Y", or hit F12 from anywhere on the pump data screen to take the
displayed pump design data create the characteristic parabolic pump operation curves shown on the
bottom of the screen.
The F12-Build Pump Button can be hit with any amount of Auto-Generation Pump Design data filled in on
the screen, but the more data provided by the user the better the generated curves will fit the users actual

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.34


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

pump curves.
The most significant error when using pumps in BOS Fluids is to put a pump into a system that will not
support the combined head and flow. The user should always start pump problems without including the
pump in the model. The suction and discharge lines should be analyzed separately to determine if the
specified boundary conditions will support the flow. The following example illustrates:

A transient analysis is to be run on the above system. The process engineer wants to know how much flow
will go through the injector nozzles once the power to the unit is shut down. The BOS Fluids model with the
pump is shown below! The user running this model is not following the recommendation above. The auto-
pump model was used to simulate the pump since this system is still just a concept in the process
engineer’s head. The auto-pump model results are shown below for the pump described above.

The rated head is 35 psi, which is the difference between the discharge head of 45 psi and the suction
head of 10 psi entered in the auto-generation section of the pump data form above. The process engineer
is hoping for a 2.4 cu.ft/sec flowrate from the pump to the injectors. This flowrate was entered also. The
BOS Fluids model is shown below:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.35


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The steady state flow calculation is shown below.

The pressure drop across the pump P101 is 47.46 – 9.999 = 37.461 psi. at 1.17 cu.ft/sec of flow. This is
outside the minimum flow area described on the pump data sheet above. The minimum flow is 1.92
cu.ft/sec and the head at that flow is 35.3 psi. The quickly generated pump is not suited for the specified
system. The pump would have to operate well outside of its design range. (See the figure below.)

Pumps should not be operated outside of their design range because the pump efficiencies fall off very
quickly. The error above was that the process engineer did not correctly include the fluid head at the
injector outlet in his head calculation for the pump. If the pressure that must be maintained at the injectors
for the required flow is 30 psi. and the vertical length of pipe from the discharge of the pump to the injector

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.36


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

manifold is 40 ft., then an additional elevation head must be overcome:


Elevation head = (40 ft) (62.4) lb/cu.ft (1/144) = 17 psi.
There must be at least 30+17 = 47 psi of discharge head at the pump, plus any friction losses to properly
size the pump. The analyst should have left the pump out of the analysis in the first run and placed a 47 psi
fixed head at the discharge node 5. The steady state calculation would give the flowrate with this discharge
pressure and the analyst can select the pump based on this data. From the pump solution above that is
out-of-range it can be seen that a 47.46 psi discharge pressure will produce a 1.17 cu.ft/sec flow. This
system will need a pump discharge pressure greater than 47.46 to get the desired 2.4 cu.ft./sec of flow.
The pump is removed from the system and a few steady-state runs made with varying discharge
pressures. It was found that a 48 psi discharge pressure will produce a 2.49 cu.ft/sec flow that is desired.
Output from this run is shown below.

The pump can now be adequately sized and inserted in the run for the transient to start. Most pump
problems that occur in the transient solution occur because the pump is not selected, sized, or entered
properly for the steady state solution. The user is encouraged to check the steady state results thoroughly
first for consistency before starting a transient analysis with a centrifugal pump.

Surge Vessel

The surge vessel element is a multi-purpose point compressible gas model that can be used to simulate a
variety of transient fluid devices:
1) Liquid Surge Protection on Overpressure or Underpressure
2) Blowdown Pressure Source
3) Reciprocating Pump Pulsation Damper
4) Gas-over Air Model in Tube Rupture Exchanger Simulation
The surge vessel has one and two node versions, and so can be placed inline with other elements or as a single
node attachment at the discharge side of a pump for example. The surge vessel is shown schematically in the

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.37


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

figure below:

The container attached to the pipeline is cylindrical in shape and can be either horizontal or vertical. The surge
vessel is basically a transient device. There is no interaction during the steady state solution except to pass the
pressure from one node on the surge vessel to the other if the optional second node is specified. The steady
state solution however, is used to establish the initial gas pressure in the surge vessel gas space. The user
specifies the liquid height in the vessel when the transient solution starts. It is assumed that this is the charged
space in the surge tank at time t=0. There is no check to see if the initial gas mass expands into the line
contents at any time during the transient, so the user can view the liquid height in the vessel as a convenience
item that may be left zero if the gas space is to be specified directly. The solution procedure can be described in
more detail below:
Surge Vessel Solution Procedure:
A steady-state solution is run. The surge vessel does not affect the steady state solution. The pressure at the
node where the surge vessel is attached is found from the steady state solution. The volume of the surge vessel
is found by subtracting the liquid volume from the full volume of the vessel and the polytropic expansion
constant calculated from PVk = C. “C” will remain constant throughout the transient. The surge vessel gas
pocket can now interact with the pressure and flow in the attached pipeline via the polytropic expression. PiVIk =
C is solved for each time step so that Pi and Vi are pressures in the surge vessel at time i. The liquid height and
the air volume is calculated and available for printing in the 2d output plots for each surge vessel after the
solution has completed.
Surge Vessel Blowdown Solution Procedure:
The system should be started from a NOFLOW boundary condition and the air pressure for blowdown specified.
A valve should be placed in line with a pipe inlet to the system that is to be studied. A schematic of this
configuration is shown below.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.38


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Note that light resonance can be established between a surge vessel and its inlet piping. A Helmholtz resonator
can be setup with the mass of liquid in the branch and volume of the vessel. Where the inlet line to the surge
vessel is long the user should include it in the calculation.
The gas in the vessel is treated as ideal. There is no friction losses included at the entrance to the vessel. If
entrance loses should be included the model the user should provide a straight pipe section and orifice at this
location.
Surge vessels can be used in overpressure or underpressure situations. In overpressure, the gas in the vessel
is compressed and the overpressure is reduced by compressing the gas. In an underpressure situation – the
gas is expanded and provides excess fluid to the pipeline so that vapor cavities do not form which could be
harmful when they implode. Underpressure service is often provided at the discharge side of pumps.
Surge vessels can be used in air, steam or other gas lines to provide a concentrated compressibility effect
where needed.
Each input data cell is described below:
Surge Vessel Node
Enter the node in the piping system where the surge vessel attaches. If this node is part of the mainline of
the piping system there will no consideration of the inertial effects of the column of water that connects the
surge vessel with the main line. This means that the connecting piping is large with respect to the volume
rate of liquid that must be delivered to the pipeline to prevent column separation.
If this assumption is not valid then enter the connecting pipework between the surge vessel and the line to
be protected.
Vessel Outlet Node
Leave blank the Vessel Outlet Node blank for the majority of applications where the surge vessel is
attached either directly to the main line or to a branch pipe connected to the main line. The vessel outlet
node can be used in compressor bottle modeling or some acoustic dampner modeling. These special
applications will be discussed in detail in a separate section.
When entered the gas “cushion” is used to relate the pressure and flow at the inlet, or “Surge Vessel” node
to the pressure and flow at the outlet node. When left blank the outlet to the surge vessel is assumed
capped.
Surge Vessel ID
Enter an up to 10 character alphanumeric identifier for the surge vessel. Do not include spaces.
Output for the surge vessel will be reported on a component identifier basis.
Total Surge Vessel Volume
This is the total volume of the surge vessel when empty.
Air Volume
This is the volume of air in the vessel that can be compressed at normal operating conditions. This may be
equal to the total vessel volume if the vessel is completely full of gas at the start of the transient. The
steady state solution is run to find the initial pressure at the start of the transient (unless NOFLOW is
specified). If NOFLOW is specified as the initial startup condition then a blowdown pressure should be input
to define the initial pressure of the gas.
Liquid Height
Enter the normal operating level of the liquid in the surge tank from the tank bottom. If the surge vessel is
oriented in the horizontal direction then this is the distance measured along the vertical direction radial to
the tank centerline. If the surge vessel is oriented in the vertical direction then this is the distance
measured in the vertical direction ALONG the tank centerline. The liquid height is used primarily to
adjust the head from the nozzle at the bottom of the vessel to the free surface of fluid in the tank.
Zero may be entered if the tank should be fully filled with air when the transient startup.
Vessel Diameter
Enter the inside diameter of the vessel.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.39


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Surge Vessel Orientation (HORIZ/VERT)


The surge vessel is assumed to be cylindrical in shape, and may be oriented on its side, i.e. a horizontal
vessel, or vertical. The orientation of the vessel determines the magnitude of the exposed area between
the gas and line contents.
Ratio of Specific Heats
Enter the ratio of specific heats for the gas in the surge tank above the pipe liquid level. For air this ratio
defaults to 1.2.
A surge vessel is often used to simulate a Helmholtz resonator (low pass filter). The natural acoustic frequency
of the resonator can be calculated by the equation [(k)(A)Po / ρVoL] 0.5 (1/[2π]) where (k) is the polytropic
exponent, A is the inlet area of the pipe neck. Po and Vo are the initial pressure pressure and volume
respectively, ρ is the liquid density in the neck, and L is the length of the neck. This configuration is shown
schematically in the figure below:

Orifice Data

In steady state and transient cases an orifice is simulated mathematically just like a fixed opening valve.
The difference between an orifice and a valve in practice is that the orifice edge is often contoured to
produce the desired flow turbulence in some cases and streamlines in others. Orifices are often placed at
locations of high velocity in a line to reduce pressure fluctuations. A typical location is at the nozzle flange
of large vessel openings. Large opening orifices (di/do = 95%) can often significantly reduce pressure
pulsations when placed at these positions in the line.
The orifice model is an "elemental" model as opposed to a "nodal" model. The nodes on either side of the
element should be viewed as being on the flanges on either side of the orifice plate. The element that

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.40


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

describes the orifice should have a finite length as shown in the figure below:

Orifice Identifier
Enter an up to 10 character alphanumeric identifier for the orifice element. Do not use spaces. Output for
the orifice will be reported on an identifier basis.
Orifice Inside Diameter
Opening flow diameter of the orifice.
Often very small percentages of the pipe inside diameter are used here to simulate restrictive effects on the
fluid flow, i.e. nozzles, sprinklers, inlet or outlet effects, etc.
Orifice Discharge Coefficient
Assumed to be 0.5. As the orifice opening gets smaller the discharge coefficient typically increases.
Cd values are also typically functions of the anticipated flowrate, but vary from 0.9 to 0.21.
If a resistance coefficient "K" is given, then the relationship between Cd and K is: Cd = 1/sqrt(K).

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.41


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Action Screen

The Analysis Types are:


STEADY – Used to run a steady state calculation only. All transient data will be ignored. Users should
always start with a steady state calculation before jumping into a transient calculation. The steady state
analysis should be checked carefully to be sure that flows and flow paths are as expected, and that pumps
and check valves are entered and operating properly. The user can easily switch between STEADY and
TRANSIENT analysis when inputs are changed to check different characteristics of the system. Any data
input for the transient analysis such as valve closures, etc. do NOT have to be removed. Transient data is
ignored in a steady state run.
TRANSIENT – Used to run a transient fluid calculation after an initial flow state is established. These
calculations are typically made when there is a change to the fluid initial condition to be studied. These
changes are typically pump starts or stops, valve closures, tube ruptures, or time varying harmonic or
transient inputs. Time varying inputs include flow or pressure changes as a function of time. A transient
analysis can also be run after a steady state solution without a transient flow perturbation specified. In this
case the user would be evaluating the equivalence of the steady state and transient solutions. There
should be little or no perturbation in the transient solution in this case, i.e. all transient behavior 2d plots
should be straight, horizontal lines. When this is not the case there is a malfunctioning nonlinear element
and the user should look more closely at the solution.
GASTRANS – BOS Fluids contains a gas transient model that simulates the transient dynamic effects in a
relief valve vent stack. Joule Thompson effects are included in the calculation so that temperatures and
sub-to-supersonic transition zones can be predicted. The use of the GASTRANS module is described
below.
Pump Transients (F1) – Click on the pump transients button or hit the function key F1 to bring up the pump
transient data screen. Input for this screen is described below. The user can also enter pump transient data
from the Flow Element menu. The Pump Transients screen is used to tell BOS Fluids when to start or stop
centrifugal pumps.
Valve Transients (F2) – Click on the valve transients button or hit the function key F2 to bring up the valve
transient data screen. Input for this screen is described below. The user can also enter valve transient data

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.42


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

from the Flow Element menu. The Valve Transients screen is used to tell BOS Fluids when to open or
close valves.
Initial Flow State – When the Analysis Type is TRANSIENT, then the initial flow state setting is used to
determine the startup condition for the transient. Usually a STEADY state solution is selected, but
sometimes, (for example blowdown), the transient solution should start from a NOFLOW or zero velocity
state. The NOFLOW state starts with the fluid at a stand still and pressure determined from elevation in the
fluid.

Transient Pump Data

Used to start or stop centrifugal pumps.


The Pump Data Screen can be accessed from the 3-ACTION menu or 8-FLOW ELEMENTS menu
selections. This screen is used to turn on or off centrifugal pumps. The pump does NOT need to be turned
on if it is running in the steady state case. Pumps may be started or stopped at any time after 0.0 seconds.
If a running pump is being STOPPED, then enter the time when the power to the pump is stopped. The
rundown routine in BOS Fluids will use the system flow characteristics, the current speed of the pump shaft
and the rotary inertia of the pump components to determine how long it takes for the pump rundown to
occur. The rundown speed, head and torque can be plotted in the 2D results for the pump non-pipe
element. The time it takes for a pump to go from zero to running speed is an engine and controls feature.
This value must be entered by the user. BOS Fluids will linearly increase the pump running speed from
zero at the start time to the running RPM over the time specified as the startup duration.
Up to ten different pump change states can be entered on a single screen. Up to twenty different states can
be entered using the appended text files.
Each specific input is described below:
Pump Identifier (For Transient Simulations)
Enter the up to 10 character pump ID that corresponds to the pump already defined in the system that is to
startup or shutdown. Do not use spaces.
Names are not case sensitive.
Action Time (Sec.)

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.43


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Enter the time where a change in status occurs. The first time must be greater than 0.0 sec. The pump
state at startup is determined from the “Steady State Condition” setting on the pump definition screen.. Up
to ten changes may be entered per analysis. It the power to the pump is to shut off then a run down
simulation for the pump will be performed. The user should enter the time when the power to the pump is
first lost. BOS Fluids calculates the duration for the rundown simulation. The user can not enter the
duration for the run down – it is determined by the rotating inertia, pump and attached system
characteristics. If there is a check valve downstream of the pump be sure to enter it. Large pump discharge
lines often have check valves that are internally (or externally) damped to prevent damage to the valve
when it closes. The slowdown in check valve closure can permit backflow through the valve and produce a
large waterhammer pressure surge in the downstream piping when the valve finally closes. Damped check
valves must be used in this instance. The user will most often not know the properties of large check valve
damping, but by experimenting with different closure times, the user can determine how sensitive the
system is to check valve closure, and if the system shows to be very sensitive, then more investigative
work might be required.
Action (START/STOP)
For the time given specify whether the pump is STARTING or STOPPING. If starting up a pump then need
to enter the duration for the startup also. The duration is NOT USED when a pump is stopped. BOS Fluids
calculates how long it takes for the pump to run down once the power has been stopped to the pump.
Duration
ENTER ONLY IF THE PUMP IS BEING STARTED. The energy available to bring the pump up to speed
on startup is not a part of the BOS Fluids pump definition. Typical startup times where there is not a zero
flow startup procedure in place vary from 500 milliseconds to several minutes. (Shorter times for smaller
line sizes.)
If the pump is stopped, the performance is a function of the parameters entered on the pump data screen.
If the user is not sure of these values or has to use auto-generated pump data then a sensitivity analysis
should be performed to be sure that design conclusions are not a function of “guessed at” properties.

Transient Valve Data

Used to open and close valves. Valves may be opened and/or closed all or part-way. Valve types that can
be controlled are: GATE, GLOBE, BUTTERFLY and BALL. Up to ten different change states may be
entered. If the user needs more control then up to twenty different actions can be added using text input.
BOS Fluids will linearly vary opening positions from one time to another that change. The user can enter
the same time twice to implement step changes that occur in a single solution time step, i.e.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.44


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Time %Open
0.0 0.0
10.0 0.0
10.0 100.0
The above example will instantly open the value at the time 10.0 seconds.
Time %Open
0.0 0.0
10.0 0.0
15.0 100.0
The above example will linearly open the valve in the time span from 10 seconds to 15 seconds.
Time history plots of the valve opening during solution can be viewed in the 2D Plotted output. The user is
encouraged to check these plots to be sure that the behavior is as intended.
Each specific input field is described below:
Valve Identifier (For Transient Simulations)
Enter the up to 10 character valve ID that corresponds to the valve already defined in the system that is to
be "opened" or "closed" some amount during the transient solution of the fluid network. Do not use spaces
in the valve name.
% Open
Enter the percent opening of the valve at the corresponding time. A 100% opened valve would be entered
as 100, and a fully closed value would be entered as 0.0.
The opening percent at 0.0 seconds is taken from the valve data screen. If the opening percent at 0.0
seconds is changed in the transient data screen, the same change will be made on the valve data screen.

Column Separation Options

Column separation is the phenomena associated with:


* Entrained gas coming out of the flowing fluid as the pressure drops
* Full Voids forming in a pipe section as the pressure drops
* Slug Flow or Deluge Fill loads when valves are instantly opened or pumps started
There are primarily four ways to approach column separation:
1) Ignore it. (Use NONE as the column separation model. In this case the user may see pressures
drop below the vapor pressure and large negative pressures can exist. Flow and pressure changes will
usually be more uniform.
2) VCM. The vapor cavity model is designed to evaluate the vapor size from the flow relations after
the line pressure is held at the vapor pressure. The cavity size is assumed to be small compared to the
volume of the pipe. The cavity cannot move along the pipe. The reference temperatures, pressure and void
fractions are not used for the VCM model. The VCM and CAP models have similar behavior characteristics
when the void size is small, but the VCM approach tends to have quicker changes in maximum and
minimum pressures and flows. The VCM result plots are “rougher” in appearance. The minimum pressure

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.45


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

in the pipeline should never get below the vapor pressure.


3) CAP – Without Large Cavity Option. The Concentrated Air Pocket model assumes that small
amounts of air are entrained in the fluid. These small pockets grow or are reduced in size as a function of
pressure in the pipe. The ideal gas equation of state is used to determine the volume of air at a particular
location at the pipe. A reference state determines how much air is originally trapped in the flowing fluid.
Default reference data is given, but the user can change this if an analysis of the flow is conducted. For
user’s that are unsure of the void fraction can be increased and reduced by an order of magnitude in each
direction to perform a sensitivity analysis. In general, transient pressure rises will be smaller as the void
fraction of air in the line is larger since the air helps cushion transient effects. The CAP model without the
large cavity option will perform similarly to the VCM option. There is no limit on cavity size, but it is
assumed that the cavity is small compared to the total volume of the pipe in the vicinity of the cavity. Cavity
sizes can expand to be larger than the pipe they are contained in if the small cavity assumption is not valid.
The pressure at voids may get smaller than the vapor pressure if the small cavity size limit is exceeded.
The initial size of the cavity is based on the pressure in the line at the start of the transient and the
reference temperature, pressure and void size as specified in the column separation spreadsheet.
4) CAP – WITH Large Cavity Options. The Concentrated Air Pocket model supports moving cavities
and pipe voiding. Completely empty sections of pipe can be created as pressure in the pipeline drops and
the cavities can move throughout the system as the front of the cavity/liquid moves through the system.
The minimum pressure in the pipeline will never get below the vapor pressure. The initial size of the cavity
is based on the pressure in the line at the start of the transient and the reference temperature, pressure
and void size as specified in the column separation spreadsheet. In some cases the initial vapor sizes are
increased on a per pipe element basis so that complete drainage of horizontal pipes can occur. The
adjustment is usually insignificant when the user is experience large cavities.
The CAP model – WITH Large Cavity Options can also be used to simulate slug flows and deluge system
filling. These special features of the CAP model with large cavity options are described in a separate
section.
Additional options to check when evaluating column separation models and output:
Fluid vapor pressure and temperature – The temperature in the line can vary, and the vapor
pressure will change with the temperature.
Activation of maximum pressure trapping – The activation of maximum pressure trapping can
make the solution appear to be “rougher” than it really is. When a user is evaluating the effect of cavities on
a system often the maximum pressure trapping flag should be switched off and on. It should be “on” to be
sure that the maximum pressure is trapped due to a possible vapor bubble collapse, but it should be off so
that the varying peaks and valleys of pressure along a long pipe is not reflected in a rough pressure plot at
the end point. This feature tends to be more of a user preference. The user should only be aware that it
exists, and should be aware of the effect it has on the output plotted results. More is discussed on the
Optional form data below.
Large Column Separation Option - Enter YES to activate the large column separation option. This
option restricts the size of growth of a cavity and allows for cavities to migrate through the system. This
option is also used for slug and deluge filling simulations.
Plot Cavity Size or Velocity Option – The cavity size may be plotted instead of the velocity in the 2d
plotted results. (One is replaced by the other in the fixed size output file.) When the user is expecting large
cavities the plotting of cavity is often more convenient. The cavity sizes plotted are the cavities at the node.
Potentially larger cavity percentages might exist along the pipe in between the nodes. The full cavity profile
can be plotted as described below.
Output Exclusion/Force List “Write Cavity Profiles” Option – When large cavities are formed that
completely empty the pipe, the location of the cavity and its movement through the system is often
interesting and important. The user can print the cavity profile on a per reach basis for up to 20 different
times. This generally produces a significant amount of output, but the user can also limit the amount of
output by providing node range limits also. (Output outside of the exclusion ranges is not printed.) The user
should experiment with the exclusion lists and the times for cavity writing. Slug or deluge fronts can be
tracked as they move through the empty piping system.
Column Separation Model (NONE/VCM/CAP)
Select the column separation model type.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.46


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The Concentrated Air Pocket (CAP) model is on by default with a fairly conservative value for gas void
fraction of 5.5E-6. For sensitivity studies this value should be adjusted both up and down by an order of
magnitude.
Column separation only affects the VCM model when the local line pressure drops to the vapor pressure.
For the CAP model, the size of the cavity is a function of the local line pressure, but the pressure cannot
drop below the vapor pressure so long as the cavity size is small. When the cavity size becomes large the
CAP model without Large Cavity Options activated can produce pressures smaller than the vapor pressure
of the fluid. The maximum cavity size is printed at the top of the maximum and minimum pressure report
and a warning is issued to the user if the cavity size exceeds 50% of the reach size and the large cavity
model is not activated.
Reference Temperature
Reference temperature of the fluid when the air void fraction is measured. Given in degrees Fahrenheit for
English units and degrees Centigrade for SI units. Typically the void fraction is not measured, and so this
input is not changed from the ambient default.
Reference Pressure
Reference pressure where the void fraction of gas-to -liquid is measured. Typically the void-fraction is not
measured, and so the reference pressure is defaulted to the barometric pressure.
Void Fraction
Volume fraction of gas-to-liquid when the liquid is at the specified reference pressure and temperature.
This value is typically not known. The default is 5.5E-6 which is a typical value for air entrained in water
flows. Gas void fractions of 1.0E-7 and smaller produce situations where no gas is available to increase in
size and there is no column separation. Gas void fractions of 1.0E-5 produces situations where there is too
much predicted entrapped gas and the solution is not realistic. Values in between should be tested and
realistic results used in design.
The VCM (Vapor Cavity Model) for column separation does not require the estimation of these parameters
and will typically produce results that are conservative when compared with the CAP model. The CAP
model correlates better with actual test data however. Only the CAP model supports large cavity
movement.

Program Interfaces

BOS Fluids input may be read from FE/Pipe, CAESAR, Triflex, or PipePlus. The use of each interface is
described below:
FE/Pipe – The FE/Pipe interface is provided via the TR0 file. This file can be read or written by BOS
Fluids. The TR0 file is used by other translators into FE/Pipe and contains node, diameter, thickness,
coordinates, restraint, displacement and local force data. For BOS Fluids the file only contains the nodes,
diameters, thicknesses and coordinates. Restraints and force data are ignored since they are not used in BOS

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.47


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Fluids. For this reason it makes sense to bring FE/Pipe models into BOS Fluids, because once in BOS Fluids
the support data is lost. BOS Fluids model files can be out to TR0 files and read into FE/Pipe without support
data. Fluid data is not stored in the TR0 file.
CAESAR – The CAESAR interface is supported using the <jobname>.cii file. Internally the converter
translates the CII file into the TR0 file and then calls the TR0 file reader routine. CAESAR Version 4.4 and
earlier version neutral files are supported. The CII file is an ascii file that can be inspected using notepad. The
<jobname>.TR0 file will be left in the user’s data directory. The TR0 file is also an ascii file that can be inspected
using notepad or a similar ascii editor.
TRIFLEX – The Triflex interface is implemented using the binary database <jobname>.phb file. The
Triflex converter translates the phb file into a TR0 file and reads the TR0 file into BOS Fluids. The user can
inspect the resulting TR0 file and compare it to the original input if there are errors in the translation. The phb file
is stored using a different topology algorithm than the PipePlus, CAESAR or FE/Pipe files and so there is more
programming effort in converting from one format to the other. User’s should check the converted data carefully.
PIPEPLUS – The PipePlus interface is supported using the ascii <jobname>.pnf file. The pnf file is
converted into a TR0 file and then read into BOS Fluids. PipePlus node names may be alpha-numeric and there
is a conversion routine that translates these into corresponding numeric values. The generated node number
depends on the alpha-numeric numbering scheme. It is preferable that the PipePlus user enter numeric values
so that there is a one-to-one correspondence between the PipePlus model and the BOS Fluids model.
OUTPUT – The current model will be written out to a TR0 file. Only the base geometry of the system is
included in this file. This is used to write out suction lines for example, that can be later included into the
discharge system so that both can be written together.
INPUT – The <Name_Entered>.TR0 file in the current working directory will be read in and added to the
end of the current element list. Note that fluid properties are not included in the TR0 file.

Name of File to be Read/Written


Enter the jobname to be read in without the extension.
BOS Fluids knows the extension to search for given the source of the interface file. The file name search is
not case sensitive. The interface file to be read should reside in the same data directory as the problem
model file.

Source/Destination Option
Enter the name of the program that was used to create the interface file, i.e., FEPIPE, CAESAR,
PIPEPLUS, etc. The names INPUT and OUTPUT can also be used. If the name OUTPUT is used then a
TR0 file will be written with the current model geometry. If the name INPUT is used then a file
<jobname>.TR0 will be read. This facility is used to read in multiple connections between equipment,
pumps, etc., that may have been analyzed separately. Note that there is no node number incrementing
implemented at this time so the user must keep the node numbers unique between systems.

Activate Import/Outport Now


Hit "Y" or "F1" to activate the data transfer process from the source file into BOS Fluids or from the element
database to the TR0 file. The transfer “OUT” to a TR0 file only occurs if the Source/Destination option is set
to OUTPUT.
Error messages will be displayed interactively to the user if for some reason the data transfer was
unsuccessful. Elements read in are added to the end of the existing data set. The TR0 contents for BOS
Fluids only includes the node numbers, basic pipe dimensions and coordinates.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.48


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Optional

The optional screen contains all major solution control options. It is designed so that the casual user does not
have to access it. Other users can find significant control of the solution and output on this screen. Each option
is discussed in detail below.
Fluid Mesh Reduction (OPTIMIZED/STANDARD)
Standard 3D models of piping systems are typically poor candidates for conversion to 1-D fluid models
because of the significant number of short pipe lengths and changes in diameter. BOS Fluids has an option
that allows the user to optimize the piping configuration so that these items do not cause a degradation of
the solution. OPTIMIZING a solution remeshes the geometry in a way that the fluid model is not sacrificed,
but that the time step does not become too small. Because the model is completely regenerated
automatically by BOS Fluids there are certain assumptions made about characteristics of the model that
are not significant and can be alternatively modeled. If the user thinks that any of these assumptions have
jeopardized the solution accuracy a check may be made by changing the OPTIMIZED option to
STANDARD and rerunning the solution.
The only difference between results in an optimized and non-optimized solution should be that the non-
optimized solution may run considerably longer.
There may also be slight differences in the loadings computed on elbows in the optimized solution because
the program fluid data points available to make pressure determinations will change slightly in location in
the optimized model.
Auto Pump Stabilizer

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.49


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

This option is useful when the pump design parameters are not available and the user needs real time
adjustment to his estimated pump curves to get realistic solutions from the fluid model. The default is OFF.
If the actual system parameters are off the pump curve, then the pump curve is extended to the left to the
zero flow state and to the right to the zero head state in an attempt to find a solution.
If the Stabilizer is ON, if the system solution is outside of the minimum and maximum flow characteristics of
the pump curve, the pump curve is adjusted upwards or downwards so that the limits are satisfied.
BOS Fluids tries to let the user utilize pumps with little information about the pump. The user should check
the results from these runs carefully and be sure that the solutions make sense. Primarily the user should
check the calculated suction and discharge pressures and the flowrate. If these values are within the
specified ranges then the solution is usually acceptable.
Diameter Visualization Multiplier
Pipe diameter multiplier used to improve the display of long thin pipeline geometries. Typical values are 2-
10. This has no effect on the calculated fluid or mechanical properties. If the pipeline shows up as a long,
thin geometry, and it is hard to see the contour shaded pressures or flowrates, the user is encouraged to
increase this value until the pipeline characteristics can be more easily viewed.
Minimum Element Length
The shortest element length that will be permitted in the solution. Short element lengths require small time
steps for solutions. The longer the element length the longer the time step and the less time spent in
solution.
If omitted, the Minimum Element Length defaults to two times the first diameter in the system.
If the system is composed of a large number of very short sections of pipe that are interconnected, then the
Minimum Element Length should be set to the length of the shortest section of pipe that should not be
changed. All pipe sections shorter than the Minimum Element Length will be changed to the minimum
element length for the flow solution. This adjustment is reflected in the Pipe Parameters report.
Pipe Roughness, Fluid Datum
This input is the measure of the irregularity in the pipe wall. It may vary between new and old pipe as scale
and corrosion build inside the pipe. The default value used is 0.05mm. This value may be changed per
pipe element, or may be entered globally on the default screen.
The fluid datum may be entered to shift the entire model either upward or downward in physical space.
This input only affects the output when the pressure output is in height of liquid column.
Both inputs are entered in the same data cell. The numbers can be separated by commas or spaces. If
only the roughness is entered, then the Fluid Data entry can be omitted or set to zero.
Maximum Number of Iterations
Change this value only if the steady state solution fails because of a lack of convergence. For the default
LINEAR solution method, the total number of iterations required for convergence is less than 15. Solutions
will still be allowed to continue even if convergence is not achieved. The *** Non-Converged *** message is
printed in the “Steady State Messages” report. The user should check the Steady State Messages report to
see that the solution has converged in some reasonable number of iterations.
Should the LINEAR steady state solution fail to converge the user should try the following two things:
1. Increase the number of iterations to 200 and retry the solution, or
2. Check the magnitudes of the flows in the solution. If the calculated flows are five-to-ten-or more
orders of magnitude greater than the flow tolerance, then increase the flow tolerance and retry the
solution. The flow tolerance was significantly reduced for 4.10 and later releases.
The LINEAR solution procedure used in BOS Fluids is typically very convergent and very quick. If a
solution does not appear to converge the user most likely has a check valve in the wrong direction or
a pump specified that is off the pump curves. Since the solution will always continue with the last
computed value, the user of a non-converged solution can check the calculation state at the last
iteration and figure out what happened.
Flow Convergence Tolerance, Valve Exponent

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.50


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Both inputs are entered on the same line and are separated by a comma or space.
When the calculated flow rates change by less than the flow convergence tolerance during a steady state
solution iteration the solution will be considered converged. The default is 0.0001 cubic length units per
second. The user should check the calculated flow solutions and be sure that this tolerance is small
enough for the problem being solved. Very small diameter tubing problems for example, may need a
smaller flow tolerance.
The valve exponent is also used during the flow convergence to determine when a valve is closed. A
valuve is considered numerically closed when the flow through the valve is less than 1(-Valve Exponent) The
default is 10, so that any flow through a valve that is smaller than 1D-10 is considered zero.
Maximum Simulation Time (Sec.)
This is the ending time for the transient simulation. BOS Fluids calculates this time automatically based on
the system properties and the transient events, and so this input may be left blank, but the user is
encouraged to carefully consider the value selected by BOS Fluids and to decide if a longer or shorter time
is appropriate. The solution time should be long enough to allow the user to view the transient events of
interest. This is typically enough time to let any perturbation reflect several times from each end of the
system.
If the user wants a good spectrum analysis of the lowest acoustic frequency of interest, then at least four
periods of the respond should be trapped. Longer run times will improve the low frequency resolution of the
Fourier transform. A single period of reflection can be estimated from 2L/c where L is the total system
length of wave traversal, and “c” is the speed of sound in the fluid.
After viewing the results the user should decide if the time spent in solution is sufficient. If not, then the
Maximum Simulation time should be input as a greater value. If the transient event of interest passes early
in the solution time, the user might shorten the time to speed both the calculation and the post-processing
time.
Lowest Mechanical Natural Frequency (Hz)
This is the lowest mechanical natural frequency of the piping system represented by the fluid model.
THIS IS AN OPTIONAL INPUT.
If entered this value will be used as part of the criteria employed to establish the total simulation time. One
criteria used for setting the total simulation time is that the low frequency harmonic responses due to
collective wave reflection must be trapped. The low frequency response is often due to the collective
harmonic contribution of multiple reflections.
BOS Fluids has several special features designed to isolate the frequency content of the fluid excitation.
Two dimensional frequency content plots of elbow-elbow pair force loadings are generated for all transient
solutions.
If the user enters the total simulation time then this value is not used.
Maximum Timestep (sec.)
LEFT BLANK BY DEFAULT.
The program will determine this value based on transient activity times, "point of information" times and
system geometry. The time step selected by the program is reported in “Transient Warnings and
Messages.” This value is usually left blank because BOS Fluids typically will select an adequately small
value to trap the desired transient. If the user has rapidly changing events that require a smaller timestep,
the maximum value may be limited to some smaller value. Large values entered for the maximum timestep
will not be used since BOS Fluids has a maximum timestep possible based on geometry and fluid. The
timestep may be smaller than this system maximum, but it cannot be larger.
Output Time Increment (Start, Stop)
LEFT BLANK BY DEFAULT.
BOS Fluids will automatically estimate this value if omitted and will produce output results starting from time
zero to the end of the simulation.
Transient solutions can produce sizeable output files to post-process because of the small time steps that

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.51


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

are typically required for a solution. Time steps on the order 0.00003 seconds are not unusual. A 10 minute
simulation with a time step of 0.00003 seconds will produce twenty-million time steps. This large number of
time steps might be necessary for the program mathematics to track the moving sound waves, but it is
seldom needed for post processing results. The more time steps included in the post processed solution
the longer it takes for Fourier transforms and time history plots to be generated. The Output Time
Increment field was designed to help the user get just the output of interest so that post processing can
proceed quickly and efficiently. The user should be careful specifying these parameters and should
thoroughly review the paragraphs that follow to be sure that the response of interest is being studied.
There are many types of phenomena that can be studied by BOS Fluids. The can generally be put into two
categories:
1) Single event, i.e. tube rupture, pump trip
2) Periodic events.
The user must be careful, because even single events, such as pump trip, will produce periodic events as
sound waves emanating from the single event reflect back and forth in the fluid system. To trap a single
event, the user must have an output time interval that is smaller than the single event time. For example, if
a tube ruptures in 5 milliseconds, (0.005 seconds), the output time should probably be some fraction of
0.005 seconds to properly characterize the initial development of the shock wave. If 2L/c for a pipe section
of importance is 0.002 seconds, then the output time should probably be some fraction of 0.002 seconds so
that the wave and loading in the branch pipe is properly reported. In no case does the output time need to
be smaller than the solution time step. If the output time increment is less than the solution time step it will
be ignored, and output will be generated for each time step in the solution.
Two additional values may be entered after the Output Time Increment, and they describe the time window
for output production. If output files are getting large or unwieldy, then the user can often only generate
output for a small part of the transient where he knows that the phenomena of interest occurs.
The Fourier transform operates only on the time slice output, and on the output time increment. The lowest
frequency reported in the Fourier plot is a function of many items, but one is the total duration of the output
sample. Shorter output samples eliminate lower frequency content from the frequency diagrams. Longer
output time steps eliminate high frequency response.
The figures below illustrate. The same waterhammer solution is shown in each plot. The difference in the
shape of each plot is caused by the Output Increment (Start/Stop) Settings.

Default Output Time Increment Output Time Increment Too Large

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.52


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Default Output Time Increment Window 2-to-3 Sec. Output Time Increment 0.1 sec. Window = 2-to-3 sec.

Output Time Increment = 0.0001 Time Stop = 1.5 Sec.

The user should be particularly careful with this window size when there is a desire to create animated 3D DirectX models
of the system transient. Ful 3D models of a transient require a large amount of disk space and time to create. While they
are spectacular and often prove very useful the user should be careful generating output files that are too large to
completely animate effectively. This is a parameter the user should experiment with. Results vary from machine to
machine and are a function of hard disk and processor speed along with graphics card capability.

Wavespeed Tolerance (%)


BOS Fluids will adjust the wave speed internally to aid in convergence and for other simulation purposes.
This value is used to limit the amount of adjustment the program will make. The actual value of the speed
of sound in the fluid medium is typically not known within about 10% of the calculated value, and so 10% is
used as the tolerance on the wavespeed. Seldom is an adjustment of the wavespeed required. A smaller
tolerance will produce smaller time steps. The wavespeed adjustment only affects a transient solution.
Length Tolerance (%)
BOS Fluids will adjust the lengths of pipe elements internally to aid in convergence and fluid mesh
construction. This value is used to limit the amount of adjustment the program will make. Larger
tolerances will make the solution run faster, but at less accuracy. Smaller tolerances will make the solution
run slower, but with greater accuracy. Seldom does this value need to be changed. The variation in used
element lengths is printed on an element-by-element basis in the Pipe Parameters report. The element
length is only adjusted for the transient solution.
Deactivate Max Pressure Trapping
By default, BOS Fluids will search for a maximum pressure during the transient solution at any point along

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.53


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

a pipe’s length. This maximum pressure may occur at a non-printed time step, or at some point along a
pipe element in between the node locations. With Maximum pressure trapped, the maximum pressure that
occurs at any point in time, and at any point in the piping system is trapped and reported at the nearest
node. In this way BOS Fluids is always giving the user the maximum pressure that occurs in the transient.
At times however, the user may want to see the "exact" pressure at a certain point in the piping system.
(For example at a boundary condition, or at a pressure gage.) In these cases the user can deactivate the
maximum pressure trapping feature to get a better plot of the pressure for just a single point. If the user
does not understand a 2D pressure plot for a particular node, then Deactivating the Max Pressure Trapping
should definitely be tried as a resolution. Often times deactivating the maximums trapping can result in
much more meaningful 2D and 3D plotted results.
If the user deactivates the maximum pressure results, he must realize that the maximum pressure
calculated by the program might not be reported. Think of an actual piping system. The operators only
know the maximum pressure at locations where there are pressure indicators. If a maximum pressure
occurs somewhere else besides at a pressure indicator the operators will not know.
The deactivation of maximum pressure trapping does not affect the reporting of flowrate, velocity or force,
which are always reported on a nodal basis.
Plot Cavity Size or Velocity
The output file from BOS Fluids is streamlined as much as possible to speed access. Only four values are
stored in the rapid access animation and 2D plot files. Typically the velocity is one of those quantities
stored. If the user would prefer that the local cavity size be stored instead of velocity, (which can be
obtained from flowrate anyway), then enter a YES in this field. This is useful when large cavities are being
watched or when the implosion of vapor bubbles is causing high pressure surges at high elevation sections
of pipe. This output is often used with the LARGE CAVITY MODEL and with the Cavity Profile Time output
on the Force/Exclusion form.
Use Large Cavity CAP Model (NO / SLUG / PLUG / GLUG / FOAM / FLASH)
When the possibility exists that vapor cavities or voids in the pipe will exceed the total pipe size, (e.g. when
it is possible that the pipe will be completely void of fluid),the large cavity model should be used to check
and provide for this occurrence. This option is not turned on by default because most cavity formation in
pipelines is small compared to the total volume of the pipe and so the large cavity calculations are not
required. Small adjustments are made to the vapor behavior to make sure that full drainage can occur.
When the vapor cavities are large these adjustments are negligible. When the vapor cavities are small,
these adjustments will disturb the accuracy of the CAP model. The disturbance is not significant, and would
probably not be noticeable by the user, but a separation of these capabilities was still considered prudent.
Each of the Large Cavity Models is described below:
SLUG – Impact Unstable Slugs. These slugs may produce high local pressures because of the collapsing
vapor and Joukowski’s effect. The slugs are very normal in character and do not mix before impact,
increasing the magnitude of the pressure wave. Usually associated with higher velocity slug movement.
GLUG – Non-Impact Stable Slugs. These slugs tend not to produce high local pressures because some
intermingling of the impacting fluid is permitted before the full deceleration of a slug section is experienced.
GLUG models tend to produce more realistic pressures in situations where slugging occurs in many places
of the line or over large portions of the line.
PLUG – Intermittent Unstable Slug Flow. The PLUG flow large cavity model will tend to produce sections of
liquid separated by areas of gas.
FOAM – Uniform Low Density Flow. Foamy flows do not tend to experience impacts but produce flow
continuity even through areas of low density.
FLASH – Unstable Variable Flow Expansion. This option is not really a column separation option. It is
associated with control valves and provides a capability in the valve to flash the incoming liquid so that the
downstream volumetric flowrate exceeds the incoming volumetric flowrate. The flashing constant used to
describe the change in incoming vs. outgoing flowrate is described in Chapter 5 of the BOS Fluids User’s
Guide.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.54


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The SLUG, GLUG and FOAM slug types develop, and are driven by a pressure and continue with a constant or
reducing flow depending on the boundary condition. The PLUG and FLASH slug models provide for an
expansion of a liquid or gas traveling in varying densities.

(P)(A), (M)(V) Multipliers, Drainage k1,k2,k3 Constants

The (P)(A) and (M)(V) multipliers are used during the unbalanced load calculations. The unbalanced loads
on a piping system subject to flow changes have a momentum (MV) and a pressure (PA) component. The
pressure component dominates in waterhammer and steamhammer events, and the momentum change
component dominates in a slug event. In general the momentum changes are ignored when
steamhammer and waterhammer calculations are made. This was true in BOS Fluids prior to version
4.111. After version 4.111 the addition of slug loads mandated that momentum changes be included in the
unbalanced load calculation. Enter 1’s as the leftmost two entries in this data cell if both momentum and
pressure loads should be considered when calculating unbalanced loads. For slug loads it is not
recommended to include the effect of pressure loads since violent local collapse of voids can produce peak
pressures that result in very large and unrealistic unbalanced loads. Omitting the PA (pressure term) for
slug load conditions often produces more realistic unbalanced load calculations.

To reproduce pre 4.111 BOS unbalanced loads, the leftmost two inputs should be 1, 0. The PA pressure
term was always included by default in pre 4.111 versions and the MV momentum term was always
omitted.

The k1 and k2 constants are used only with the large cavity CAP model. The constant k3 is not used at this
time. The constant k1 defaults to 0.1 and is a value that can over-relax or under-relax the speed of
drainage. This constant has the most effect for horizontal, or almost horizontal pipes. The default for k1 is
0.1 which introduces a small amount of gas in each reach that will permit the reach to fully drain. If values
for k1 are entered that are less than 0.1, a reach may not empty even though the downstream pipe has
become completely void of fluid. Values of k1 greater than 0.1 will make it easier for pipe sections to drain.
The constant k2 is an inertial term multiplier lower limit. The characteristic method uses two principle terms
to describe the equations of flow. One of these is an inertial term that accounts for the resistance of the
fluid in a reach to movement. When a reach empties due to large cavity formation, this inertial term should
approach zero. Letting the inertial term go to zero however introduces an uncoupled set of network
equations that is not solveable. The value of k2 is the limit for the reduction of the inertial coefficient. The
default is 20%, i.e. 0.2. Smaller values may reduce the local violent peak pressures often times inherent in
the SLUG model. Larger values tend to increase these perturbances. The largest effect is often seen in
going from SLUG to GLUG models however. The GLUG model will usually produce lower, but more
realistic pressures. The SLUG model will usually produce higher, but potentially realistic pressures. K2 is
a constant used most often with the SLUG model to attempt to attenuate unrealistically high peak
pressures. The user is encouraged to experiment with these values to see which produces the most
realistic simulation.

Model Building Filename to Include


If the user requires additional action items, time history data points, etc. an ascii file can be provided that
contains this input. The name of the ascii file to include should be entered here. Omit the .TXT extension.
To include the file put a YES in the field below. An example is shown below how additional time history
data points are generated to simulate the output from a triplex pump.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.55


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

7-Screen List

Use the function keys to view the related data information. Information displayed is not available for editing
or printing. These quick lists are only intended to help the user check columnar data that is repeated from
one element to the next.

8-Flow Elements

The 8-Flow Elements Form lets the user access and edit all fluid data associated with a model from a
single location. The user can always edit the fluid data from the element where the fluid data was defined,
but the user can additionally step through all of the subject-similar data from the Flow Elements screen.
The user is encouraged to use this screen liberally as it is very useful for comparing items in a list and for
editing and changing parameters. Note that text output and input is available from all Flow Elements
screens, so the user can write out the existing data to an ascii file, read it into Excel, operate on that data,
and then read it back into the exact same BOS Fluids screen.
The user may also enter any new Flow Element data from the flow element screen. Data such as Valves,
Pumps or Boundary Conditions do not need to be referenced from any particular element. (They can be,
but they don’t need to be.)
The F11-Update Page References option is useful because it searches through each item in the Flow
Element list and cross references it to the element that contains it. This allows the user to access, edit, add
or delete valve data, for example, either from the Flow Element listings or from the Pipe Element listings.
Effective use of this feature makes getting back and forth between specified data very simple. Page
references need updating for any item that is added from the Flow Element list.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.56


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Output Exclusion/Force & Cavity List

There are three types of input directives available on the Output Exclusion / Force & Cavity form:
1) Node range exclusions to reduce the size of output and to tailor output reports
2) Node Ranges to direct the force pair searching so that the correct elbow-elbow pairs are used to
compute unbalanced loads.
3) Cavity Profile times to tell when the detailed cavity report should be printed. These reports are
useful particularly when slugs or deluge type filling is simulated. At each time specified a report will be
generated that tells where in each pipe element and to what degree the slug/void/fluid interface is present.
Node Range (Output Exclusion List)
Enter a range of nodes that should be excluded from any output reports. Any element connecting to a node
number in a specified range will be omitted from the report. This facility can be used to minimize the output
for large systems with long run times.
Exclude Flag
The exclude flag is used with the Node Range. If the Exclude flag opposite a specified node range is YES
then the node range will be excluded from the output. If the Exclude Flag is NO then the node range will be
ignored. The user may enter node ranges and then activate them or deactivate them as needed during the
design phase. (Sometimes the analyst will want to see certain data, and other times they won’t.)

Node Range (Force-Pair List)


In certain cases the force-pair search algorithm employed in BOS Fluids may not produce the node pairs
the user is interested in for an unbalanced force calculation. The Force node range input lets the user
direct BOS Fluids to use certain nodes, and the paths between those nodes for force calculations. In some
situations, there may be multiple paths between the two nodes the user has entered. When this occurs, (or
if BOS Fluids does not select the path that is desired for some other reason), the user should try making
either the first or last node in the node range negative. A negative node number causes the pair searching
algorithm to start looking in the opposite direction first. Example output from a force/pair calculation is
shown below.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.57


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

In the example output shown above the user specified the first two force-pair paths, the first from 4090 to
3340, and the second from 3320 to 3348. All other force-pair paths are selected by the program. BOS
Fluids searches for user requested force-pairs first, and then marks those paths as not available for
automatic program searches. In the plot above the user wanted a single force-pair calculated for the node
length from 3340 to 4090. BOS Fluids saw the intermediate elbows and stopped the force calculation at
that point. The user however, wanted to make a conservative estimate of the total load in the length
upstream of the Y branch and so specified the force pair shown.
Write Cavity Profiles
The user should enter times where a snapshot of the vapor cavity condition should be taken. A tabular
dump of each element and each reach within each element will be produced for each entered cavity time.
Up to twenty cavity output times may be entered, ten per line. Separate the cavity output times with spaces
or commas.

Gas Transient Analysis


This facility analyzes gas safety relief systems that empty into a fixed pressure space. This is the
introduction of this facility into BOS Fluids with plans to expand the current steady state capability into the
transient solution so that HAZOP studies of looped relief valves firing may be studied. Additionally, the
temperature drop in flare systems can be analyzed so that minimum metal temperatures in the larger flare
header can be found and brittle fracture avoided. The gas transient data entry is selected from the main
BOS Fluids menu:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.58


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The following input screen should appear:

The user must identify which portion of the BOS Fluids model should be extracted for use in the Gas
Transient analysis and where the PSV or control valve is in the model. A gas transient analysis can be
performed on any system the user desires by entering the data on this screen and changing the ACTION
type to GASTRAN. Once the gas transient analysis is performed (separate from any other BOS Fluids
analysis), the user can deactivate the calculation. (See third text input cell above.) The user does not have
to build separate models just to perform the gas study. This separate model generates a comparison with
the Brandemaier and Knebler paper and can be used to compute temperature profiles in the vent stack,
flare header, or downstream piping. Temperature profiles can be important when the user wants to be sure
that the exiting gas does not cool the vent stack down to temperatures that would be below the brittle
transition temperature for the material. Additionally steam relief outlet pipes should not drop in temperature
so that excessive condensate forms in the outlet piping.
The Gas Transient model will also help the user be sure that the desired mass flow can established in the
downstream tailpipe and volume changes. Choked flow is calculated at points in the flow where either
friction or area changes produce a sonic condition.
For steam relief systems the user can estimate the total amount of condensate that will appear in the vent

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.59


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

stack. This is done by comparing the temperature and pressure along the vent stack to the Mollier diagram
for steam. Starting from the exit and working back toward the orifice, the user should find the point in the
flow stream (if any) where the steam is no longer superheated. If this occurs it will most likely be at the exit
and at some point back into the tailpipe, with the lowest quality steam being at the exit. The average
quality should be found and the amount of condensate that will drop from the flow estimated by the steam
flowrate and average quality. The user should provide for removing this quantity of condensate should this
be a significant value.
Non-ideal gas equations are used in the GASTRANS module to assure that accurate temperature
predictions are generated. The nonlinearity of the gas is characterized by Pitzer’s Acentric factor and its
effect on the gas compressibility. The acentric factor can be found in texts such as, “The Properties of
Gases and Liquids,” by Reid, Prausnitz and Poling, and a few common values of the critical temperature,
pressure and the acentric factor from The Properties of Gases and Liquids, are sited below:

Gas Tc (deg Kelvin) Pc (bar) Acentric Factor


Argon 150.8 48.7 0.001
Chlorine 416.9 79.8 0.090
Nitric Oxide 180. 64.8 0.588
Nitrogen Dioxide 431. 101. 0.834
Nitrogen 126.2 33.9 0.039
Oxygen 154.6 50.4 0.025
Sulfur Dioxide 430.8 78.8 0.256
Ozone 261.1 55.7 0.691
Sulfur 1314. 207. 0.171
Hydrogen 33.0 12.9 -0.216
Water 647.3 221.2 0.344
Ammonia 405.5 113.5 0.250
Phosgene 455. 56.7 0.205
Carbon Monoxide 132.9 35.0 0.066
Methane 190.4 46.0 0.011
Acetylene 308.3 61.4 0.190
Ethylene 282.4 50.4 0.089
Ethane 305.4 48.8 0.099
Propane 369.8 42.5 0.153
Isobutane 408.2 36.5 0.183
Benzene 562.2 48.9 0.212
Phenol 694.2 61.3 0.438
Cyclohexene 560.5 43.4 0.210
Cyclohexane 553.5 40.7 0.212
Toluene 591.8 41.0 0.263
Styrene 647. 39.9 0.257
Ethylbenzene 617.2 36 0.302
n-octane 568.8 24.9 0.398
Naphthalene 748.4 40.5 0.302

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.60


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

1 deg Kelvin = 1 deg Celsius + 273.15

The model type is PRESSURE or VOLUME. A constant pressure source is the default and accommodates
most relief or control systems that operate in a steady state opened condition. Where a vessel or smaller
volume of gas is to be relieved, the VOLUME option should be used, and the pressure drop in the volume
will be determined along with the corresponding effect on the flow system. Of particular interest here is the
change in flow pattern that occurs when sonic properties at the orifice are no longer supported.
When the VOLUME option is selected the GASTRANS module will perform a time history analysis of the
flow condition as the volume is discharged through the fully opened valve. Output for the thousand or more
time steps should be controlled by the user so that meaningful results are presented without excessive data
generation. For this reason, the user should carefully select the number of time steps and the time
increment. Output is printed for incremental time steps and is also controlled. The default number of time
steps for output is 100.
There is no graphical output from the GASTRANS module at this time.
Input will be illustrated by the following example:

The gas to be vented is CH4. The upstream pressure is 140 barg and the downstream pressure is 5 barg.
The above model was constructed in BOS Fluids and the properties for CH4 entered into the Gas
Description part of the Gas Transient Simulator screen shown below.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.61


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The relief valve description is shown in the BOS Fluids valve data screen below: Since the Gas Transient
module currently only evaluates the steady state flow condition of the PSV none of the parameters that will
effect the transient condition must be entered. Note that the valve ID is entered in the Gas Transient input
so that the location of the throat condition can be established. Input for this model is included in the
BOSMODELS folder, and a portion of the output is shown below:
Note how the supersonic condition beyond the valve is clearly shown in the cell isolation report. The
temperature change and thrust along the pipeline are also predicted. The user can see the effect of the
shock wave transitioning the flow from supersonic to subsonic, and can see that the thrust at the subsonic
exit is somewhat higher than that predicted by Brandmeir Knebler. The flowrate is limited by the sonic
properties at the valve orifice and for that both methods predict exactly the same flow. The difference is in
how the Gas Transient simulation handles the real properties of the gas downstream of the relief valve
orifice. The temperature drops after the valve and then recovers after the shock wave. In this example the
temperature does not drop enough to cause concern for the brittle transition of carbon steel, but this is
certainly not universally true.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.62


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

A comparison is made with the Brandemeir and Knebler paper in the Static messages section of the BOS
Fluids output. A portion of this report is contained below:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.63


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.1.64


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Utilities
The Utilities menu is accessed directly from the main PRGMaps screen. Click on the blue UTILITIES button or
hit the key “U” when the PRGMaps window has focus. The following utilities menu should appear:

Configure

The configuration screen allows the BOS fluids to control the static graphics colors, the default set of
units, and various data base sizes, (although the data base sizes will rarely need to be set in BOS Fluids).
Options C through E in the above menu let the user change the default colors in the static graphics
screens. The colors in the 3d animator are optimized for performance and cannot be changed. Option G is
used to set the units used for input to either English or Metric. The H - Graphic viewport specificiations are
hardly ever changed for BOS Fluids models. (They are used for huge piping systems, and have only ever
needed to be changed for artificially large systems used in testing.) The same applies for Data Base
Parameters. The control here is seldom required for a BOS Fluids job.
If under Windows NT the user has difficulty with the operating system allocating most of the hard disk to the
processes virtual memory space, he can limit the virtual memory space in the J-Miscellaneous Settings
Section. This value should be set to approximately the amount of actual memory in the machine times
0.75.
The configuration established can be saved locally or globally. If the current
working directory (shown in the window handle for PRGMAPS), is the installation
directory then when the user saves changes, they are written over the global
configuration file and used for all new jobs. If the current working directory is a
local directory that is NOT the installation directory, then the configurations saved
will only be used for jobs whose input files reside in those directories. The
configuration file is CONFIG.BIN. This file can be copied from folder to folder as
the user wishes to implement a particular selection of options.

Specify Graphics System


The screen below is not used for the Windows version of BOS Fluilds.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.2.1


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Set Text Attributes


As of version 4.110 BOS Fluids has switched to a “paper” interface that is hardwired. The following text options
do not affect the screen settings at this time.

Set Graphics Colors


BOS Fluids has two graphics engines. The first and oldest is a statics graphic engine that sits on top of the
model generation package and gives the user a “static” view of the system being created. The entities in this
package can be configured by the window below, although this is not recommended. The second and newest
graphics engine is the 3d viewer. The colors in the 3d viewer are fixed at this time.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.2.2


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Printer Hardcopy Setup


This setup screen is not used for Windows versions of BOS Fluids.

Units Specification
There are two units sets available for BOS Fluids: METRIC and ENGLISH. The user also has the option of
selecting pressure units when each job starts. Once the units for a job are selected, the input for that job
remains fixed in those units.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.2.3


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Graphics Viewport Specifications


The only option that might need to be changed in the following screens is the size of the display list. (See the
last option in the second screen below.) If the user happens to have a huge BOS Fluids model and gets a
display list error, the size can be adjusted by selecting B-Edit a current ViewPort, then entering a large value for
the display list size.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.2.4


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Define Database Parameters


The BOS Fluids and FE/Pipe model generators allocate a base construction space for points, curves and
surfaces used to create the three-dimensional images of pipes, pumps, etc. The settings below are usually more
than adequate for BOS Fluids models, but should the user get a message during the model generation phase
that a “block size” has been exceeded. One of the values below can be increased. Users should be careful
increasing the values shown below excessively as virtual memory is allocated for the predetermined size. The
user has up to approximately 4 Gigabytes of virtual address space, so initialization time and hard disk space are
more the real physical limits. Increases by more than an order of magnitude each step are not recommended.

Miscellaneous Settings
Only the Maximum Memory setting on the bottom of this screen is used. This value is typically not needed for
BOS Fluids models, but is recommended to be set to 3/4 of the hard memory on the machine in use.

Save Changes and Return


Save all changes made during this Configuration session and return to the Main Menu. Note that once

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.2.5


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

changes have been made they are NOT saved unless this option is used. Use the <escape> key to exit
from the Utilities menu editor WITHOUT saving any changes.

Interfaces
Click on desired section of screen below for detailed information. The interfaces accessed
from this menu do not support BOS Fluids models. The BOS Fluids interface is accessed
from the input interface. To read a file into BOS Fluids from CAESAR, Triflex, Pipeplus, ect,
the user should enter a new job name and go to Input for BOS Fluids. From the input menu
the user should select INTERFACES and enter the neutral file name to be read in. There is
more information on interfacing in the Input description of this manual.

Batch Processor
This is a job queue that can be used to submit multiple, long running BOS Fluids jobs back-to-back. The
system will run each analysis without stopping, and then will return to the main menu and await the user's
post processing commands for each job. The input for each job must be in the same subdirectory.

Add Job to Queue


Add the jobname for each job to be added to the batch que. All jobs MUST have previously been Prepared
for Analysis in order to be added to the batch processor. All jobs that are run together MUST exist in the
same subdirectory. If a job has NOT yet been prepared for analysis, the following message will appear:

Delete Job from Queue


Delete the jobs, one at a time, that you wish to remove from the queued jobs to run under batch control.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.2.6


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Exit - Start Batch Procedures


Exit this screen and start the batch analysis. All previously prepared jobs that are listed in the queue will be
analyzed.

Exit - Do Not Start Batch Procedures


Exit this screen without starting the batch process. Files whose names have already been entered will
remain in the queue and can be started at a later time.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.2.7


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Valve Closure Examples


This basic gas-hammer simulation is intended to demonstrate a number of the different options in
BOS Fluids that can be used to get different types of output. A number of program options are
used and discussed to help the user understand the program and the basic physical phenomena
under study.

Standing wave patterns and “resonances” in piping sytems are typically associated with the two
basic boundary conditions shown in the figure below.

At open ends the pressure is fixed at a single, or controlled value and the velocity (or flowrate)
is adjusted by the solution of the flow equations to satisfy the continuity and energy equations.
These boundary conditions are typically found at the entrance to large volumes, i.e. tanks or
pressure vessels.

At closed ends the velocity (or flowrate) is fixed at a single, or controlled value, and the pressure
is adjusted by the solution of the flow equations to satisfy the continuity and energy equations.
These boundary conditions are typically found at end caps, pistons, closed valves, or the
discharges of positive displacement machines.

By default BOS Fluids assumes that each pipe end, if not explicitly defined to be otherwise,
contains a pipe cap, i.e. is a closed end. If the end of a particular system is not a closed end, but
rather opens into a larger container or to the atmosphere, the user should place a boundary
condition at that end. Fixed pressure boundary conditions simulate an open end and allow high
pressure waves to pass. Fixed flow boundary conditions simulate a closed end in reflection
aspects and reflect pressure waves that encroach upon it, the velocity remaining at all times
constant.

Fixed flow boundary conditions are usually only dealt with in real systems at let down stations
where sonic conditions exist at the pressure let down valve, and at the discharge of positive
displacement machines.

Fixed head, (or fixed pressure) boundary conditions are more common, and exist at the entrances
and exits of vessels, and at the entrances and exits of centrifugal equipment.

In steady state simulations, a fixed flow boundary condition is often used to determine the steady
state pressure drop required to maintain a certain material flowrate. Once the pressure drop to

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.3.1


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

support the flowrate is determined, the fixed flow boundary condition is replaced with a fixed
head (pressure) boundary condition to more accurately simulate the open ended condition at the
source, (or discharge). This is demonstrated in the schematic below:

The heat exchanger discharges a hot liquid through the cutoff valve and into the process tower. If
the process experiences overheating problems the cutoff valve is closed in 1/2 second. This is a
rapid closure and we want to investigate the system for water hammer. The steady state flow
discharge from the heat exchanger is at 150 psi. and 0.3 cu.ft./sec. We know the elevation
difference between 5 and 45, but there will also be some friction losses. How can the system be
modeled?

Both the nodes 5 and 45 exist at open ends on the piping system and should be modeled as fixed
pressure (or fixedhead) boundary conditions. We only know the flowrate and the pressure at one
node however.

In this case, BOS Fluids can be used to find exactly the pressure at the opposite node to support
the design flowrate. This is important because the velocity is proportional to the flowrate and the
waterhammer pressure rise is proportional to the velocity. The procedure will be to run a steady
state solution first with a fixed head boundary condition of 150 psi at node 5, and a fixed flow
boundary condition of 0.3 cu.ft./sec at node 45. The steady state solution will calculate exactly
what pressure will exist at 45 to support the 0.3 cu.ft/sec. flow. This pressure will then be used
back in a transient analysis of the system to properly simulate the boundary conditions as open
ends when the transient valve closure occurs.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.3.2


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Take another example of a straight system with an inline closing valve as shown in the figure
below:

The system contains a gas. Both the inlet and outlet pressures are specified for the steady state
condition. We are given the following gas properties and so they can be entered directly into
BOS Fluids. (We don’t need to use the fluid or gas database.)

Note that the speed of sound is overridden since the speed of sound was given. Only the viscosity
and weight density will be used properties. For the gas calculation, the bulk modulus is estimated
as the mean line pressure divided by the ratio of the specific heats. If the speed of sound is
known, the bulk modulus can be calculated directly using the speed of sound and the gas density.
There may be some difference in the “used” speed of sound as printed in the output reports
however. Some adjustment within the tolerance specified in the optional data screen may be

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.3.3


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

needed to establish the internal characteristics grid. See the Pipe Parameters report for entered
and calculated speeds of sound.

To run a transient simulation of the valve closing in 5 ms. the user must:

1) Enter the physical layout of the piping system, or read it in from a pipe stress program.

2) Establish the steady state flow condition. This usually means setting the inlet and outlet
pressure or flowrate. At least one pressure must be entered. (Think about flow through a
straight horizontal pipe. For the steady state case the flow at both ends can be entered,
but this does not tell anything about the pressure to produce the flow. The pressure and
flow cannot both be specified at a single point. If the pressure is specified, the program
calculates the flow at the point. If the flow is specified, then the program calculates the
pressure to support it.)

3) Describe the transient, i.e. the valve closure, or pump startup.

4) Adjust the output and transient solution so that the output that is meaningful.

For the above problem BOS fluids produced the following two-dimensional output plot for the
node 15 in the system. The total solution time, time steps, and tolerances have all been
defaulted.

Several things are noticed about this plot:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.3.4


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

1) There are not too many points on the curve. We’d probably like a few more data points.
For larger problems the defaults selected generally do produce more data points, but the
amount of output can be controlled from the optional data screen. We can ask for a longer
time to be used in the calculation, or we can ask for a smaller output time step, or both.
2) The total time is 0.14 seconds. At a sound speed of 1200 feet per second, and a total
length of 60 feet, a wave will go from one end to the other, and return to the original end
in: 2L/c seconds = (2)(60)/(1200) = 0.1 seconds. BOS Fluids selected a time that is 40%
longer than that time. Often the user must select his own total solution time to see the
behavior that is sought, but if he does not, BOS Fluids will pick one for him just like it
did above. If the output from this calculation was to be used in a pipe stress program it is
usually desireable to produce at least five peaks of the lowest resonance. In this case that
would mean extending the time to (5)(0.1) = 0.5 seconds.
3) The increase in pressure is all positive. This does not seem to agree with the physics of
the problem. With a valve closed at the left end and an open pipe at the other end we
have a closed-open system. The wave reflection from the open end will return as a
negative pressure wave. The plot above certainly contains enough time to see the
returning negative pressure wave – why is it not there?

Ans:
A common difficulty with transient solutions is the huge amount of data produced for even
moderately sized piping systems. To speed post-processing BOS Fluids condenses data that is
distributed over part of the system and reports maximums of that data at the system nodal points.
The user’s entered node points can be thought of as physical locations on the pipeline where a
pressure or temperature indicator are positioned. These are points where the user wants to get
information out or put information in. These may not be the points in the system where the
maximum pressure occurs however. BOS Fluids tracks all of the internal maximum pressure
peaks even though they may not occur at user selected nodes. Since the nodes are the only
physical location where data is reported to the user, BOS Fluids will find the maximum pressure
and report it at the node closest to the actual place where the pressure peak occurred. What is
seen in the figure above is the maximum pressure in the pipe in the area of node 15. In the area
between node 15 and the next node point in the system the maximum pressure in that space of
pipe is always positive. The user has the option of turning off this maximum pressure trapping
feature. If this is done then the pressure that is plotted for the node 15 will be the pressure that
would be measured at that node by a pressure indicator. The pressure plotted in BOS Fluids is
the so-called “static-pressure,” or the pressure that would be measured by a theoretical pitot tube
that was perpendicular to the flow path. If pressure trapping is turned off, and the output time
step decreased from 0.01 seconds, (the default), to 0.001 seconds, (which will be better suited for
our short time period), the output plot of pressure for node 15 makes more sense. The text cells
for these inputs can be found on the OPTIONAL data form.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.3.5


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

With pressure trapping deactivated the plot makes more sense. The negative pressure wave
returns to point 15 at about 0.085 seconds. (This is when the pressure starts dropping.) The
pressure wave first entered 15 at about 0.015 seconds, so the difference between the time the
pressure wave first passed 15 and when it returned to 15 is (0.085) – (0.015) = 0.07 seconds.
The distance from 15 to the free end at 31 and back to 15 is (2)(42) = 84 ft. A wave moving at
1200 feet per second would cover this distance in (84)/(1200) = 0.07 seconds.

Several of the major events shown on the plot are explained in the figure below:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.3.6


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

BOS Fluids can also perform a Fourier transform on the time signal shown above and give the
frequency content. If this is done on the above signal the following frequency plot results:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.3.7


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

A closeup of the first peak shows a frequency of about 45 Hz. If the period of a closed-open
cylinder is 4L/c, then for the example problem the fundamental acoustic period is:

(4)(60)/(1200) = 0.2 seconds.

The frequency is the inverse of the period: 1/(0.2) = 5 Hz.

The period of a closed-open 20 foot cylinder is (4)(20)/(1200) = 0.06666 sec., and the associated
frequency is 1/(0.6666) = 15 Hz.

Our frequency plot shows the first acoustic natural frequency at 45 Hz, but hand calculations tell
us that there should be an acoustic frequency at 5 Hz. and at 15 Hz. Where are these lower
acoustic natural frequencies?

The problem here is that we are asking the fourier algorithm to do too much with too little
information. To get better resolution in the low frequency range the time of solution must be
about three-to-four times the fundamental period. Our fundamental period is 0.2 seconds, and the
total duration of the solution was only 0.14 seconds. Without adequate solution times the lower
frequency pulsations cannot be accurately plotted and aliasing problems can result in the plotted
frequencies. Aliasing is demonstrated in the sketch below:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.3.8


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Low frequency cutoff can be corrected by including more time in the solution so that lower
modes can be resolved. If the solution to the example problem is extended to 0.6 seconds, (3
times the fundamental period), the following plot appears:

We can start to see the aliased 5 Hz. fundamental mode and the 15 Hz mode. Going from a
smaller diameter to a larger diameter produces a closed-(approximately)open pipe acoustic
configuration. Going from larger to smaller produces the wave reflection that is more
synonymous with an (approximately) closed end connection, (depending on the area ratios).

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.3.9


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The diameter of the pipe has a significant effect on line pressure as can be seen in the
comparison of the two pressure waves from the points 15 and 22, which are four feet apart, but
are in two different diameters of pipe. Point 15 is in a 6” line, and point 22 is in a 10” line.

point 15 point 22

Compare these with point 29 which is in the 6” pipe just downstream of the 10” pipe:

point 29

Going from a smaller to a larger line size reduces the transient pressure magnitude, which is
generally NOT returned when going back to the smaller line size. (Exactly the opposite of what
happens in a steady state flow condition.)

Simplify the problem somewhat to look at basic pressure and flow relations in closed-closed
systems. The above model has been changed so that a constant 6” diameter exists throughout the
pipe length as shown in the figure below:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.3.10


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The pulse above introduces a single volume displacement into the left end of the pipe, much like
the movement of a piston in a cylinder. This volume displacement sends a pressure wave moving
from left to right down the pipe heading toward the closed end at node 31. The initial movement
of the pressure wave is comprised of a velocity and a pressure rise.

Note that the velocity of the flow is significantly less than the speed of the moving pressure
wave, (which is about 1200 fps.) The magnitude of the pressure wave can be estimated from:

p = (rho)(c)(V)
= (1.8) lbm/cu.ft (1200)(0.043) /(32.2)/(144)
= 0.02 psi.

When this moving pressure/velocity wave encounters the closed end at the right side of the tube,
the velocity introduced by the “cylinder” movement that initiated the pressure wave, is reduced
to zero and the change in velocity is converted into pressure (another rho*c*V component),
causing an increased pressure wave with a zero velocity of flow to move back towards the left
end of the cylinder. When the wave hits the left end of the cylinder the entire flow will be at rest

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.3.11


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

since the flow pulse influx has dropped to zero some time earlier. The pulse duration was only
0.03 seconds, and the return time for a wave is 0.1 seconds.

The pressure plot at the left end closed node 10 is shown below:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.3.12


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Valve Closure Simulation Procedure:

The following example details the solution procedure described in the first example
above. It is repeated below in a different example since the establishment of
boundary conditions is a very important part of a proper simulation.

Summary

The following steps are described to perform a valve closure simulation.

1) Find the pressure drop required to support the design or specified flowrate.

2) Put the pressures from step one back into BOS Fluids. Remove the flowrate specification.
(Leaving the fixed flowrate specification in from step 1 is the mistake that everyone makes.
When a valve is closed on a fixed flow system the pressure climbs to infinity. Valves are closed
on fixed pressure systems, and that is what needs to be simulated.)

3) Run the transient simulation with the pressures inserted from step 2.

The user will typically know the pressure at one point in the system and the intended flowrate.
The flowrate is usually the controlled quantity from a process point-of-view. Flowrate is also
directly proportional to velocity which is directly proportional to the pressure rise that occurs
when a valve closes. For this reason an accurate description of the flowrate is important.

Fixing the flowrate is one way to insure that a given volume of fluid moves through the pipe, but
the flowrate does not remain fixed when a valve closes. If a fixed flowrate is used and a valve is

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.3.13


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

closed, BOS Fluids will continue to insert fluid into the closed system causing the pressure to
rise to unrealistic values. To get around this problem pressures are specified instead of
flowrates. The user must, however find the pressures at both the inlet and outlet to produce the
desired flowrate.

Often pressures are given for the inlet and outlet of a system. These pressures however may be
the design values and are not concurrent values. Even if the pressures at the inlet and outlet are
given on a process diagram they are often calculated based on some assumed friction factors or
valve closure states, that could only be replicated with considerable work. Since the most
significant property is the flowrate, it is recommended to let BOS Fluids calculate the pressure
difference required to produce the flowrate. The pressure at either the inlet or outlet should be
specified, and then the flowrate given at the opposite end. Do not specify both the flowrate and
the pressure at one point in the system, they are mutually exclusive. (This is what BOS Fluids
does. If given the flowrate at a point, it calculated the pressure required to support that flowrate.
If given the pressure at a point BOS Fluids calculates the flow to support that pressure. The user
cannot specify both the pressure and the flowrate at the same point.)

In the example above the outlet from a boiler at node 5 is given as 1250 psi. The flowrate in the
line is 0.7 cu.ft./sec. This steam flowrate will be extracted from the system (a negative flowrate)
at the connection at 140 since the flow is from 5 to 140.

An initial steady state analysis is performed to determine the pressure at 140 (and throughout the
rest of the system) required to support a f low of 0.7 cu.ft./sec.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.3.14


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

From this steady state analysis the pressure at 140 is found. For the problem above the pressure
at 140 is found by BOS Fluids to be 1243 psi. For the user’s BOS Fluids model of the flow path
a 1250 psi to 1243 psi pressure drop between nodes 5 and 140 will produce the design flowrate.
It is not so important if the downstream pressure calculated is accurate. It is far more
important that the supported flowrate is accurate. The downstream pressure will be fairly
accurate for most plant piping systems however, as pressure drops through the relatively short
piping systems experienced in a plant environment are not that great. The problem is that the
pressure difference, though small must be accurate to get the flowrate, and thus the velocity
correct!

The input for node 140 is now changed from a FIXFLOW boundary condition to a
FIXHEAD boundary with the prescribed pressure. This input is shown for BOS Fluids
below. Note that the flowrate value can be left in, but will not be used when the boundary
condition is FIXHEAD.

The Valve Transient for C001 is specified:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.3.15


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

And the ACTION TYPE is changed to TRANSIENT.

Transient results and pressure rises now will be calculated accurately with the correct flow
velocity interruption, even if the pressure model of the pipe is not exactly correct. Pressures,
unbalanced loads, etc. can be used with confidence, and there will not be any artificially high
pressures associated with fixed flows into or out-of closed systems.

Recommendation:
When running a transient solution the user must always run a steady state calculation first.
The steady state calculation can be a NOFLOW analysis, in which case there are zero flows
throughout the system and only the heads based on elevation are calculated, but this is the trivial case.
The more difficult steady state condition to satisfy is when there are nonlinear elements, such as
pressure control valves and check valves in a steady state non-zero flowing case. To be sure that the
steady state and transient solutions agree, it is often a good idea to start up the transient solution after
the steady state case without any transient. Clearly in this case, the transient case should proceed from
the steady state solution with zero perturbation. If there is any fluctuation in either the pressure or flow
waveforms anywhere in the system then there is either an imbalance in the system or an unstable
condition was left from the steady state solution. This can occur in the following instances:
1) The steady state solution does not calculate vapor formation. When the steady state solution is close
to the vapor pressure and when the vapor cavities can effect the flow condition, then the steady state
and a non-perterbed transient solution will not agree.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.3.16


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

2) Damped check valves do not exist in the steady state solution. Special devices are used in BOS
Fluids to keep damped check valves from violating the steady state flow solution, but there are
instances where some perturbation will emanate from a check valve, especially close to a pump that is
damped.
3) Pressure regulator valves are only solved in the steady state solution to within a certain tolerance.
This tolerance is different during time stepping and so PRV’s may experience some oscillation when
going from a steady state to a non-perterbed transient solution.
4) Relief valves cannot open in the steady state solution. If the steady state pressure is close to the
relief valve opening pressure, then the transient, non-perturbed solution will not match the steady state
solution as the the valve will oscillate or open and close – likely producing large pressure swings.
5) Temperature effects are not incorporated in the steady state solution. Where changes in density due
to temperature produce pressure changes then the steady state and the transient solutions might not
match.
6) Where pumps are far off the pump curve there may be mismatch between steady state and transient
solutions.

Some items to check in all cases in steady state and transient runs:
1) Look at the minimum and maximum pressure reported in the tabular reports. Wild swings here can
indicate that there was a part of the transient that was unstable. The largest and smallest pressures
found at any time during the transient are printed at the top of the Maximum and Minimum Pressure
report. These values should be checked for “reasonableness.”
2) The maximum cavity percentage that occurred at any time during a transient is also printed at the
top of the Maximum and Minimum pressure report. If this value exceeds 20% of the local line size the
user should consider using the large cavity model, and should be sure that there does indeed exist a
condition in the pipeline where it may begin to drain or to form sufficiently large low pressure areas so
that almost the entire pipe section will be filled with gas.
3) When surge tanks are used the user should be sure to check the liquid level in the tank to be sure
that it does not go below zero if the tank is used to prevent vapor from entering a discharge line.
4) PRV positions should be checked in the transient solution to be sure that they have not swung to
completely open or closed. This is certainly a possible case, but reflects a generally poor design.
5) Check valves should not be alternately opening and closing the next time step. This reflects a
numerical instability and the condition should be checked.
6) Plots of the unbalanced forces should be checked before values from the tabular reports are used
for design. Closed valves in an elbow-elbow pair string can produce artificially high unbalanced loads in
the system. BOS Fluids makes every attempt to compute conservative values for unbalanced loads
throughout the system and sometimes the selection of areas for pressure, etc. while conservative are
excessively so. The user should simply make sure that the values printed “make sense.”
7) Pumps should always be checked to be sure that they are operating within reasonable limits on the
pump curve. Incorrectly specified steady state systems can show pumps way off the pump curves. In
these cases the energy introducfed into the flow is in error and any solution sensitive to the pump
effects will likewise be in error.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.3.17


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Waterhammer Example

Introduction
The following piping system is subject to a sudden valve closure at the pump suction and
resulting waterhammer. The valve closes in 0.4 seconds. BOS Fluids will be used to
calculate the pressure rise and the unbalanced force that results from the closure.

The geometry can be input directly into BOS Fluids or can be imported from a Triflex,
PipePlus, CAESAR II, or FE/Pipe neutral file. The fluid boundary conditions and the valve
characteristics are required input. The complete input for this job can be found in the
BOSMODELS subdirectory, jobname= HAMMER. (The BOS Fluids input filename is
HAMMER.IFU.)
There are two boundary conditions in this model. One is at the pump suction and will not
effect the solution since the valve will isolate the suction node boundary condition at node
number 1. The second boundary condition is at the source tank. This boundary condition
will reflect the pressure wave coming from the closed valve. How the pressure wave is
reflected will effect the problem solution and so this boundary condition should be modeled
correctly.

An open ended boundary condition is one where the velocity of flow at the boundary is not

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.4.1


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

zero or a constant value. At a pipe endcap or at a closed valve the velocity of flow at the
endcap or closed valve must be zero since the flow cannot progress through the endcap or
valve body. At an opening into a tank or pressure vessel the velocity of the flow may be any
value that is supported by the pressure differential on either side of the nozzle. A “closed”
boundary also exists where constant flows are specified. At these points incoming pressure
waves are reflected just like the system end was closed. For the hammer system above
pulling liquid from a supply tank, the node 125 will be at an open end. At this location the
pressure will be assumed to stay constant. The system will behave like a closed-open
system whose natural period is 4L/c where “L” is the length from node 5 to node 125, and “c”
is the speed of sound in the fluid. Reflections from a closed end pipe will cause a pressure
maximum and a corresponding velocity minimum. Reflections from an open end will result in
a velocity maximum and a pressure minimum. Closed-closed systems, and open-open
systems will have a fundamental natural period of 2L/c. This relationship is demonstrated in
the figure below:

When a frequency analysis is done of the water hammer solution for the system shown
above it is expected that the fundamental period displayed will be 4L/c where L is the
distance from node 5 to node 125. Note that the actual wavefront is shown as a sloped blue
line in the figure above. Actual wavefronts can be quite long as the total extent of a valve
closure wavefront is (c)(tc) where “tc” is the valve closure time. If the speed of sound (c) in
the system is 1373 m/sec, and the valve closure is 0.4 seconds, then the wave front length
will be (1373)(0.4) = 549 meters or longer than the total system length! (The actual wave
front length will be some fraction of this distance since the valve closure is not typically a
linear function through the closure time duration.) The total system length is 83 meters. The
characteristic time scale is 2L/c since the returning wave begins canceling the source before
the valve is fully closed. There will still be some unbalanced load however, and this is the
magnitude that is sought. The distance between an elbow-elbow pair would have to be
equal to or longer than 549 meters to develop the full unbalanced pressure thrust load due to
the waterhammer.
The maximum magnitude of the unbalanced waterhammer load will be smaller than that
predicted by the famous Joukowski equation: ρcdv. Waterhammer and steamhammer waves
reflect from both closed and open ends, as well as from changes in diameter. Only five
reflections are needed to produce a resonant level response in an undamped piping system.
This means that even low magnitude pressure waves can produce large displacements if
they move in the system at a speed that corresponds to a mechanical natural frequency. It
is for this reason that some systems are particularly susceptible to acoustic excitation:
1) Hot pipework supported by springs often has low natural frequencies
2) Any long runs of pipe that is hanger supported typically will often not have inherent
horizontal restraint.
3) Any system that is exposed to flow perturbations:
1) Systems attached to reciprocating equipment
2) Systems where valves open or close quickly
3) Systems where boiling or chemical reactions occur

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.4.2


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

4) Systems containing two phase flow.


Job Procedure
The recommended procedure for running the delivered job: Hammer is listed below:
1. Install the BOS Fluids program and start PRGMAPS.
2. Using the File:Old menu selection, navigate to the BOSMODELS subdirectory and
select the job HAMMER.
3. Select BOS Fluids.
6. Select Old Job.
7. Select Input
8. From the input menu enter: E-Submit for Analysis
9. The job will run through the steady state and transient analysis and will stop on the
Input|Analysis|Results|Utilities main menu.
From this menu select: Reports,
then type: Run1,
when asked for a unique character identifier. Select the reports you are most
interested in by using the mouse to click on the report title. More than one report can
be selected at any time. Hit F1 to generate the report. Results from the steady state
solution and the transient solution are accessible at the bottom of this screen.
10. Escape from the tabular reports, return to Results and select 3D Plots.
A plot of the system geometry should appear. The user should be able to produce
animation, maximum plots, etc. using the various menu options. Animation can be
produced in a static movie view or using the 3d DirectX viewer so that the user can
rotate, zoom and clip on the animating plot. The user is encouraged to experiment
with this feature as it is a very powerful visualization tool.

Results
There are three types of results available from a BOS Fluids transient run. The user should
become familiar with each one to know which is better for a particular application. The three
types are:
3D Graphical Results – A 3d model of the system is displayed. The user can overlay
animations and maximum results in color.
2D Graphical Results – A 2d X-Y plot is generated for non-pipe elements such as pumps
and valves, or for any pipe node in the system. The 2d plots can display pressures,
flowrates, velocities, cavity sizes and forces. A frequency analyzer is also available to
perform a fourier transform on the resulting time history plot so that frequency components
can be identified.
Tabular Results – These reports contain input echos, warning messages, and results for
both steady state and transient solutions. The tabular results should be studied carefully to
be sure that the calculated solution is reasonable. Any abnormalities observed by the
program during execution should be reported in the tabular results.

Discussion
When force vs. time is plotted for any point in the system using the 2D-Plot output, four data
files are automatically written that contain the force vs. time curve for a positive or negative
application. For the node 50 force vs. time plot for the HAMMER system, the four files
written are named: PP50 and MP50 for PipePlus, and PC50 and MC50 for CAESAR II.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.4.3


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Force vs. Time Plot for Node 50 in Job “Hammer”


The PP50 file is for the positive application of the force, and the MP50 file is for the negative
application of the force. For the node 50 in the Hammer job the negative force file will be
used because the force will be applied in the minus X direction initially. (The sign in general
does not make a difference in a linear dynamic analysis, but may make a difference when a
nonlinear transient dynamic analysis is performed or when loads are applied at multiple node
locations. This is one of the reasons that multiple node application is not recommended.
Good, practical solutions can most usually be produced in typical piping systems when one
loaded elbow-elbow pair is addressed at a time.) A portion of the MP50 file is shown below.

The first column contains the time in seconds, and the second contains the force in the
user's input units. For the Hammer job the units for the force are Newtons. This format can
be read by PipePlus. For CAESAR II the time should be in milliseconds. The files PC50
and MC50 contain the time column in millisecond units. Force vs. time curves can be
generated for each node-pair in the model where BOS Fluids has a computed unbalanced
loading. Unbalanced loadings are computed for node-node pairs where the transient
pressure wave can be fully reflected. This is considered to occur at all 90 degree elbows,
dead ends, and branch intersections. Each pair of nodes is contained in the Steady State
Warnings and Message report. Users can also coerce BOS Fluids to select particular node-
node pairs. See the 9-Output Exclusion/Force Lists during Input.
Computed maximum unbalanced loads can be animated using the 3d Plotted output or can

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.4.4


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

be displayed in a 2d X-Y plot. (The output PCnnn, MCnnn, etc. are only generated when
the 2d plotted output is reviewed.)

Maximum Unbalanced 3D Force Plot


Users often begin a dynamic mechanical study by investigating the effect of the few largest
waterhammer loads on the most flexible axial portions of the piping system. This is generally
where displacement or stress problems will be experienced. If the system is not flexible then
the user must only be sure that the supports providing the rigidity can sustain the maximum
value of the dynamic loading.
Once individual dynamic loads on flexible parts of the system are evaluated individually, the
user may wish to add dynamic loads, but this is seldom necessary at the design stage. Both
the PipePlus and CAESAR II transient dynamic capabilities allow for the cumulative action
of transient dynamic loads at different points in the piping system.
When multiple loads are evaluated, the user is responsible for use of the correct positive or
negative force file. The application is determined by inspection. (This is why it is
recommended to solve each force problem one at a time.)

Time Steps and Total Solution Time


The solution time step for a mechanical analysis by a pipe stress program is not the same
as the solution time step for a transient fluid analysis by BOS Fluids. The time step for the
mechanical analysis is determined by the highest piping system mechanical natural
frequency that should be included in the analysis, and by the shape of the transient loading
curve. Without friction, and full of water, the Hammer system shows to have natural
frequencies in the one Hertz and lower range that will be of interest. The modal analysis
using CAESAR once a static run is completed is performed by selecting dynamics from the
analysis menu, and then setting the analysis type to Modal as shown in the screen captured
below.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.4.5


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

All other parameters can be defaulted. The little running man is selected to continue the
dynamic analysis. The first CAESAR calculated mode shape of interest is the second mode
and is shown in the plot below.

The base mode of excitation will be due to the waterhammer wave starting at the closed
valve and moving to the open end, and then reflecting back to the valve.There will be higher
frequencies associated with secondary reflections, but this will be the major acoustic
excitation mode of interest. The acoustic frequency for the open-closed system in the
HAMMER example is 4L/c = 0.241 seconds. This corresponds to a frequency of (1)/(0.241)
= 4.15 Hz. (A frequency transform of the unbalanced load at node 50 from BOS Fluids is
shown below illustrating the 4.15 Hz. peak. BOS Fluids confirms the hand calculation of the
fundamental acoustic frequency.)

BOS Fluids Frequency Transform from node 50 Closeup


The lowest mechanical natural frequency of the Hammer system that wants to participate
with the acoustic excitation is found from the CAESAR calculation at 0.499 Hz.
The acoustic excitation is at 4.15 Hz, and the first mechanical natural frequency it would
excite is at 0.499 Hz. The dynamic amplification factor for this interaction can be found from:
1/ | [1-(w/wn)2] |
1 / | [ 1-(4.14/0.499)2] | = 0.014

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.4.6


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The first major mode of the system at 0.499 Hz. will have a dynamic amplification factor of
about 1.4%, and so not much participation is expected here. Higher modes have more
complex mode shapes and cannot be so quickly dismissed and it will be for these modes
that the analysis will be performed.
The steady state flow calculation from BOS Fluids shows the flow velocity to be about 1.817
meters per second. Joukowski’s equation can be used to estimate the pressure rise due to
an instantaneous valve closure:
dP = ρcdv = 1000kg/m3 (1373) m/s (1.817) m/s = 2,494,741 Pa (2.5 MPa)
The total load is AdP = 32,275 mm2 (2.5) N/mm2 = 80,687 N. This is a modestly significant
dynamic load in the horizontal direction on an eight inch pipe.
It is expected that 1.4% of that load will act with the first mode, but it is not known how much
of the load will interact with the other modes. From the dynamic DLF plot shown below it can
be seen that the DLF drops to about 15% at a frequency ratio of w/wn = 2.5 and reaches
essentially one at 0.2. To find the highest natural frequency of interest then that would be
excited by the 4.15Hz. excitation:
w/wn = 0.2 = 4.15/wn; wn = 4.15/0.2 = 20.75 Hz.
This is not an excessively high natural frequency for this system.

As a first approximation, the mechanical solution time step can be estimated by taking the
smaller of:
i) 0.1-to-0.2 / [high frequency of interest(hz)]
ii) (0.25)(transient loading time scale.)
For the Hammer system, the highest natural frequency of interest will be 20Hz on the high
side, so the time step for this frequency would be:
0.2 / 20 = 0.01 sec.
The transient loading time scale can be estimated by reviewing the force time history plot
generated from the 2D plot routines. A close-up view of that plot is shown below:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.4.7


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

To properly trap the first peak, at least four or more time steps should be used. The peak
occurs in 0.05 seconds. Four increments would put the time step at: 0.05/4 = 0.0125 sec.
To trap the highest frequency of initial interest a 0.01 second time step would be needed,
and to properly characterize changes in the loading a 0.0125 second time step must be
selected. 0.01 seconds will be used as the time step.
The total time must be long enough to develop the mechanical natural frequency. This
typically requires at least five passes of the pressure wave. A single period occurs in 0.241
seconds. Five periods will pass in (0.241)(5) = 1.205 seconds.
The total number of time steps would be: 1.205 / 0.01 = 120. This is no problem. Many
hundreds of thousands of time steps are a problem, (depending on the system size), 120
times steps are not a problem.
The transient in BOS Fluids will be run for 2.0 seconds and the output time step will be taken
as 0.005 seconds. BOS Fluids selects its own time step so that pressure waves can be
accurately followed. The input the time step and total time in BOS Fluids is shown below:

In general the BOS Fluids selected time step will be less than the 0.005 seconds requested
for output time. There may be three or four actual time steps in between each output time
step. The output time step of 0.005 seconds was chosen so that the loading plot for the
mechanical analysis would be sufficiently accurate. The user should check these values in
the Transient Warnings and Messages Report.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.4.8


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The total solution time and the output time step are critical variables in an accurate analysis.
These parameters should be varied and their effect on the solution evaluated. BOS Fluids
users never have to enter either the total time or the time step. Both parameters are
estimated by BOS Fluids automatically. Users should verify that values selected by BOS
Fluids adequately trap the transient response of interest.
BOS Fluids also accepts the input of the lowest system mechanical natural frequency. (See
the optional screen above.) If entered this input will be used to help determine the total
solution time required. Users should always convince themselves that BOS Fluids has
selected a suitably conservative value for the total solution time or the user should enter the
time themselves.
Knowing how long to run a transient simulation and how to select the output time steps is a
skill that comes from experience. Different fluid behaviors involve different time scales and
the user must be able to apply the basic rules of physics (some enumerated above), to be
sure that solutions fall within the realm of the reasonable.
When the output from a fluid analysis is to be used in a subsequent mechanical analysis the
total transient fluid time can be estimated from at least the greater of:
i) (5)(Fundamental Wave Period)
ii) (Longest Loading Time Scale) + (2)(1/[lowest participating mechanical natural
frequency(Hz.)]
When the times for the above are approximately equal, or if the wave period corresponds to
an important system mechanical natural period, then about five reflections of the wave are
all that are required to setup a fully resonant mechanical response in that mode. The
solution will always be more conservative if the total solution time is increased, and the
solution time step is decreased.
The damping used in a mechanical system is typically taken to be 0.03, which is 3% of
critical. Greater values have been suggested to approximate the damping effect caused by
plastic action. In this case, values of damping in the 10 to 15% range seem to more
accurately simulate dynamic behavior.

Applying BOS Fluids Transient Dynamic Loads in a CAESAR II Analysis


1. While in BOS Fluids generate a 2-D plot for force as a function of time for the node-
node pairs of interest. Generate as many plots as necessary. Typically start with only
the most significant pairs. The menu buttons used to access the 2D plot function are
shown below. The Steady State Warnings & Messages report gives each node pair
used for a force unbalance calculation. Each node along the path between the node
pairs will produce the same unbalanced force plot.

2. Inspect the piping system and decide if the positive application of the wave or the
negative application of the wave should be used, and decide where to apply the load in
the node-node pair pipe run. The most important consideration is that the node

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.4.9


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

selected should be on the axial run of pipe. Do not select nodes at 45 degree
locations on elbows! Common practice uses the node at the appropriate elbow
weldline, or the node closest to one end of the node-node pair. Typically only start with
one of the worst loadings to test the systems behavior, before applying multiple
dynamic loadings. Most systems can be evaluated for design purposes by solving for
the effect of one load at a time.
3. Enter the CAESAR II input for the job. Usually the piping model will be available
because it was used to generate the BOS Fluids input. Make sure that the static
analysis has been run, and include any friction coefficients if friction is to be included in
the dynamic analysis.
4. Enter "dynamics input" and select Time History from the dynamic option menu.
5. Select Spectrum/Time History Definitions, then Spectrum/Time History Name and Type
to point CAESAR II to the force vs. time data file on the hard disk. When the Time
History Name and Type window appears enter the following line:
#<filename> TIME FORCE LINEAR LINEAR
For the Hammer job, for the node 50, and for the "Minus" direction loading file, the user
would enter:
#MC50 TIME FORCE LINEAR LINEAR
6. Enter Force Set Editing. Add a force with a magnitude of 1.0, in the direction desired,
at the node desired, for force set number 1. For the Hammer job this input will appear:
1, X, 50, 1
The magnitude of the force is 1.0 because the actual magnitude for the load magnitude
is contained in the MC50 file.
7. Enter Spectrum/Time History Load Cases. Add a single load case for the transient
analysis. The form of this input appears:
<spectrum name>, <factor>, <direction>, <force set>
For the Hammer job the input will be:
CM50, 1, X, 1
The user is encouraged to test this loading facility on a simple model before attempting
the dynamic analysis of a complex piping system.
The 5'th through 8'th items on the CAESAR II Control Parameter data screen control
the transient solution. Anywhere between 8 Modes to 24 Modes produce essentially
the same maximums dynamic solution. A total solution time of 6.8 seconds can be
used with a time step of 0.02 sec., and a damping of 0.1.
Once these values are entered, the CAESAR II transient user will produce results that
show approximately 200mm of displacement and 100MPa of maximum code stress
due to the multiple passing of the waterhammer wave. Even though this stress is
tolerable for the reasonably low number of cycles that will probably be experienced,
200mm of displacement is often a sufficiently high value to cause concern.

Applying BOS Fluids Transient Dynamic Loads in a PipePlus Analysis


1. While in BOS Fluids generate a 2-D plot for force as a function of time for the node-
node pairs of interest. The screen combinations used to generate 2D force plots are
shown below. Generate as many plots as necessary. Typically start with only the most
significant pairs of nodes. The Steady State Warnings & Messages report gives each
node pair used for a force unbalance calculation. Each node along the path between
the node pairs will produce the same unbalanced force plot.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.4.10


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

2. Inspect the piping system and decide if the positive application of the wave or the
negative application of the wave should be used, and decide where to apply the load in
the node-node pair pipe run. The most important consideration is that the node
selected should be on the axial run of pipe. Do not select nodes at 45 degree
locations on elbows! Common practice uses the node at the appropriate elbow
weldline, or the node closest to one end of the node-node pair. Typically only start with
one of the worst loadings to test the systems behavior before applying multiple
dynamic loadings. Properly applying dynamic loads
3. Enter the PipePlus program for the job. Usually the piping model will be available
because it was used to generate the BOS Fluids input!
4. Find the node in the input data listing where the transient loading should be applied.
(For the Hammer job the BOS Fluids load file MP50 will be applied to the PipePlus
node number 50 in the X direction.) Move the cursor into the Extra Data Field for the
node and enter an "H" for History. Immediately a pop up will appear requesting:
Degree of Freedom
Arrival Time
Time Forcing Function.
Enter "TX" for the Degree of Freedom, 0.0 for the Arrival Time and MP50 for the
Time Forcing Function. The "T" in front of the X implies a "translational" loading. The
0.0 second arrival time means that the BOS Fluid plot started at time zero. (There are
options in BOS Fluids where the starting time for output can be delayed some fixed
amount of time. When this feature is used, the delay in arrival time will be entered to
properly sequence the later arriving load with earlier arriving loads.) The user is
encouraged to test this loading facility on a simple model before attempting the
dynamic analysis of a large system.
5. Using the mouse click on Load:History from the PipePlus Task Bar. Here the user will
be prompted for:
Solution Time Step
Number of Steps
Output Print Interval
Damping Ratio
These values should be entered as described above for the CAESAR calculation.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.4.11


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Things to Check in the HAMMER Job:


Once the HAMMER job has been completed as suggested above the user should have transient
and steady state output available to review. The following check list is used as a guideline to help
the user become familiar with the different reports and plots available.

When any BOS Fluids job completes control is returned to the PRGMAPS main menu:

The user should select Results: and then reports:

The first time a job completes the user will be prompted for a title to be added to each page of the
report to aid in keeping different runs of the same model separate:

Next each report will be listed in a scrollable window:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.4.12


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Each report and its purpose is described in brief below:


Input Process – A more user readable version of the input listing. Used for checking.
Job Title – An echo of the job title. User’s can enter or change the title from the output. Lines
can be added or deleted.
Input Data Echo – An exact image of the screen contents. This report tends to be too long to
print, but may be useful for checking exactly what input was used in a certain place. An
excho of each screen that contains data is included.
Steady State Warnings & Messages – A detail of notes that were generated by the steady
state fluid solver. This report also contains the elbow-elbow or node-node pairs that will be
used for computing unbalanced forces. This report should be reviewed for warning or
caution messages. If an analysis terminates inadvertently or with a red screen fatal error,
the error message, if available will be printed in the Warnings and Messages reports.
Steady Solution – Pressures, flows and velocities for the steady state solution for all pipe
elements and for all non-pipe elements, i.e. pumps, valves, etc.
2D Data Listing (SRG File) BOS Fluids performs several translations of the 3D geometry into
2D models for analysis. This reports contains the first translation. Some error messages
refer to lines in these files. When this occurs the user can go to the particular line and
check the input associated with the item described on the data line.
Optimized 2D Data Listing (SHW File) – If the ISRM is turned on this is the Intelligent System
Reduction version of the input data. Seldom are there errors related to this file.
Transient Warnings & Messages – A detail of notes that were generated by the transient fluid
solver. This report gives solution parameters, calculated time steps, etc. This report should
always be reviewed for warning or caution messages.
Pipe Parameters – Gives a detailed breakdown of the user’s system. Contains the input and
used pipe lengths and wave speeds. These two parameters are adjusted based on the
tolerances on the OPTIONAL data form to conform to the requirements of the transient
solution. Some lengths may be adjusted while others are not. If the lengths are excessively
adjusted by the program in a critical part of the system then the “wavespeed” or “length”
tolerances on the optional form should be reduced. The Colebrook-White friction factor is
also shown for each pipe element along with the number of reaches. Longer pipe elements
are broken down into shorter pipe elements called “reaches” so that the maximum
pressure and flowrates can be calculated if they occur in between nodes. When slugs or
large vapor cavities are being tracked the user can print the size of the vapor cavity at
particular times for each point along a pipe section. (See Ouput Exclusion / Force Lists in
the input.)
Maximum and Minimum Presures, Flowrates & Forces – For each user defined elements
gives the maximum pressure, flow, velocity, force and cavity size.
Sections from the tabular listing that are of most interest are included below. Annotation is
provided in blue.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.4.13


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Input Process the input process listing is where to check the input elements, etc.
Output Data Listing

Element Dimensions
From To Dx Dy Dz Pg
--------------------------------(mm.)-------(mm.)-------(mm.)-------
5 10 Bnd 914.400 0.000 0.000 1
10 15 0.000 609.600 0.000 2
15 20 0.000 381.000 0.000 3
20 25 Bnd 0.000 3657.600 0.000 4
25 26 0.000 0.000 -4572.000 5 coordinates
Element Data

Near Mid Far OD T Rough Temp Pg


---------------------------(mm.)-----(mm.)-------(mm.)----(Deg)-------
5 10 0 219.075 8.179 0.0500 60.0 1
10 15 0 219.075 8.179 0.0500 60.0 2
15 20 0 219.075 8.179 0.0500 60.0 3
20 25 0 219.075 8.179 0.0500 60.0 4
25 26 0 219.075 8.179 0.0500 60.0 5 pipe size
VALVES
Valve Bore Open
From To Type ID Cd (mm.) Exp (%) Pg
-------------------------+-----------+---------------------------------
4 5 GATE v1 0.100 198.12 0.20 100.0 1 valve ID
General/Control

ANALYSIS TYPE TRANSIENT


Initial Transient State STEADY
--------------------------+----------------------------------------
Column Separation Model CAP
Reference Temperature 21.000 deg.
Reference Pressure 0.000 bar
Void Fraction (by Volume) 0.000
--------------------------+----------------------------------------
Valve Identifier V1
Load Step Time(Sec) Percent Open(%) valve closure
1 0.0000 100.0 time
2 0.4000 0.0
--------------------------+----------------------------------------
Fluid Mesh Reduction OPTIMIZED ISRM
Initial Transient State HEAD
Diameter Visualization 3.000
Minimum Element Length 0.000 mm.
Default Pipe Roughness 0.00000 mm.
Default Datum 0.00000 mm.
--------------------------+----------------------------------------
Solution Method LINEAR
Steady State Iterations 100.000
Flow Tolerance N /cu. m. 0.001
--------------------------+----------------------------------------
Transient Simulation Time 2.00000 sec. solution control
Lowest Mech Frequency 0.000 Hz.
Maximum Time Step 0.00000 sec.
Output Time Step 0.00000 sec.
Wave Speed Tolerance (%) 10.000
Length Tolerance (%) 5.000
--------------------------+----------------------------------------
Deactivate Pressure Traps NO

EXCLUDED NODE LIST

From To
Node Node Exclude
---------------+----------

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.4.14


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

1 15 YES

FLUID BOUNDARY CONDITIONS

Piping Boundary Node : 1 Page 1 of 2


Boundary Action : FIXHEAD

Head (Pressure) : 16.000 bar


-------------------------------------------------------------
Piping Boundary Node : 125 Page 2 of 2
Boundary Action : FIXHEAD

Head (Pressure) : 17.200 bar


-------------------------------------------------------------

Steady State Warnings & Messages

General System Parameters

Number of unique nodes = 37


Number of pipe elements = 35
Number of valve elements = 1
Number of orifices = 0
Number of surge vessels = 0
Number of pumps = 0

Fluid Properties:

Name = WATER
Temperature (deg.C) = 60.000
Density (kg./cu.m.) = 983.20
Viscosity (N.sec./sq.m.) = 0.00047
Bulk Modulus (MPa) = 2280.0
Vapor Pressure (m.)[abs] = 0.2065E+01

Node-to-Node Force Calculation Path: 1 force calculation paths


From: 10 To: 10 Element: First
From: 10 To: 15 Element: 2
From: 15 To: 20 Element: 3
From: 20 To: 25 Element: 4
From: 25 To: 25 Element: Last

Node-to-Node Force Calculation Path: 2


From: 10 To: 10 Element: First
From: 5 To: 10 Element: 2
From: 4 To: 5 Element: 3
From: 1 To: 4 Element: Last

Intelligent System Reduction Mechanism is ON:


reduced number of nodes = 4
reduced number of pipes = 2

Pump Stabilizing is ON

Linear Steady State solution took 10 iterations iterations req’d SS


Steady State Solution

Pressure in : bar
Flowrate in : cu.m. /s
Velocity in : m /s

Node1 Node2 Pressure@1 Pressure@2 Flow Velocity


5 10N 17.992283 17.993033 -0.058650 1.817184
10N 10F 17.993033 17.971121 -0.058650 1.817184
10F 15 17.971121 17.942098 -0.058650 1.817184
15 20 17.942098 17.905819 -0.058650 1.817184

40 45N 17.385845 17.387721 -0.058650 1.817184

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.4.15


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

45N 45F 17.387721 17.395207 -0.058650 1.817184


45F 50 17.395207 17.397083 -0.058650 1.817184
50 55 17.397083 17.404586 -0.058650 1.817184

55 60 17.404586 17.412090 -0.058650 1.817184


60 65 17.412090 17.419594 -0.058650 1.817184
65 70 17.419594 17.427097 -0.058650 1.817184
70 75N 17.427097 17.428223 -0.058650 1.817184

2D Data Listing (SRG File)

1:ISRM
2:#
3:# BOS FLUIDS LIQUIDS AND GASES 2D INPUT FILE
4:# FORMAT BASE VER. 4.1000
5:#
6:AUTO_PUMP_STABILIZE
7:# PRESSURE IS MEASURED IN FORCE/AREA (OUTPUT UNIT IS BARS.)
8:PINBAR
9:# USERS UNIT CONVERSIONS
10:UNIT FLOW 0.1000000E+01 M.
11:UNIT PRESSURE 0.9999844E-05 BAR
12:UNIT FORCE 0.9998652E+00 N
13:UNIT LENGTH 0.1000000E+01 M
14:# BASIC PIPE DATA
15: PIPE 1 5 100010 0.6096000E+00 0.2190750E+03 0.8179000E
16: BEND 2 100010 300010 0.4787786E+00 0.2190750E+03 0.8179000E+01
25: BEND 11 100035 300035 0.4787786E+00 0.2190750E+03 0.8179000E+01
26: PIPE 12 300035 40 0.1219200E+01 0.2190750E+03 0.8179000E
27: PIPE 13 40 100045 0.1524000E+01 0.2190750E+03 0.8179000E
28: BEND 14 100045 300045 0.4787786E+00 0.2190750E+03 0.8179000E+01
29: PIPE 15 300045 50 0.1524000E+01 0.2190750E+03 0.8179000E
30: PIPE 16 50 55 0.6096000E+01 0.2190750E+03 0.8179000E
31: PIPE 17 55 60 0.6096000E+01 0.2190750E+03 0.8179000E
32: PIPE 18 60 65 0.6096000E+01 0.2190750E+03 0.8179000E
33: PIPE 19 65 70 0.6096000E+01 0.2190750E+03 0.8179000E
34: PIPE 20 70 100075 0.9144000E+00 0.2190750E+03 0.8179000E
35: BEND 21 100075 300075 0.4787786E+00 0.2190750E+03 0.8179000E+01
36: PIPE 22 300075 80 0.4267200E+01 0.2190750E+03 0.8179000E
37: PIPE 23 80 100085 0.9144000E+00 0.2190750E+03 0.8179000E

Optimized 2D Data Listing (SHW File)

1: NODE 5 JUNCTION 0.000000000000000E+000 0.00000000000000


2: NODE 125 FIXHEAD 180.690026699313 7.54380000000000
3: NODE 1 FIXHEAD 168.083745766803 0.000000000000000
4: NODE 4 JUNCTION 0.000000000000000E+000 0.00000000000000
5: PIPE 1 5 125 82.2616 219.075 8.17900 2099
6: PIPE 4 1 4 1.00000 219.075 8.17900 2099
7:# BOS FLUIDS LIQUIDS AND GASES 2D INPUT FILE
8:# FORMAT BASE VER. 4.1000
9:#

Transient Warnings & Messages

Number of internal (transient) nodes is 166

Solution time step is 0.0003695 seconds. Used time step for BOS Fluids
Solution output time interval is 0.006061 seconds.

Number of solution time steps per output time step is 16.

The total number of output cases will be 329.

The Column Separation Model used is CAP column separation model


CAP Reference Pressure is 1.01323 bar
CAP Void fraction is 0.550000E-05

Pipe Parameters

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.4.16


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Length in : m
Wavespeed in : m /sec

node1 node2 input used calc. used friction #reaches


length length wavespd wavespd factor
5 10N 0.61 0.61 1373.78 1374.35 0.0154 1
10N 10F 0.48 0.48 1373.78 1374.35 0.0154 1
10F 15 0.30 0.44 1373.78 1374.35 0.0154 1
15 20 0.38 0.44 1373.78 1374.35 0.0154 0

40 45N 1.52 1.52 1373.78 1374.35 0.0154 3


45N 45F 0.48 0.48 1373.78 1374.35 0.0154 1
45F 50 1.52 1.52 1373.78 1374.35 0.0154 3
50 55 6.10 6.10 1373.78 1374.35 0.0154 12

Maximum and Minimum Pressures, Flowrates & Forces

Pressure in : bar (Max = 28.67 Min = 0.00 ) max value shown


Flowrate in : cu.m. /sec.
Distances in : m
Cavity size in : Percent (Max = 0.0001% )
Forces in : n

From 45N max pres.= 26.577 @ 0.000 from node 45N


To 45F min pres.= 16.523 @ 0.000 from node 45N
max flowrate = 0.0010 @ 0.479 from node 45N
min flowrate = -0.0587 @ 0.000 from node 45N
max cavity = 0.0000 @ 0.000 from node 45N
max force = 16979. max force in 45 to 70 run
From 45F max pres.= 26.535 @ 0.000 from node 45F
To 50 min pres.= 16.530 @ 0.000 from node 45F
max flowrate = 0.0011 @ 1.524 from node 45F
min flowrate = -0.0587 @ 0.000 from node 45F
max cavity = 0.0000 @ 0.000 from node 45F
max force = 16979.

From 50 max pres.= 26.408 @ 0.000 from node 50


To 55 min pres.= 16.524 @ 6.096 from node 50
max flowrate = 0.0014 @ 6.096 from node 50
min flowrate = -0.0587 @ 0.000 from node 50
max cavity = 0.0000 @ 6.096 from node 50
max force = 16979.

After or while the tabular output is being reviewed the user can also simultaneously look
at either 3D or 2D output plots. Each plot driver is independent. All three output viewing
mechanisms can be used simultaneously, and the graphics windows support multiple
instances. (This means that the user could have two or more 2D plot windows open at
the same time for the same job. This is convenient when, for example, the user wants to
compare a pressure time history for node 15 with one for node 232. The plots can be
generated simultaneously on the screen and compared side-by-side.)
2D and 3D plots are accessed from the PRGMAPS main menu:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.4.17


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

2D Plots

Select 2D from the menu option, then Pipe 2D from the pull down menu to plot points in the
piping system. The following data panel will appear:

Enter any node number in the piping system in the Node Number text cell and select “what” to
plot, and the “type” of plot. The default shown above is for a pressure time history plot.
Pressure, Flowrate, Velocity or cavity size, and the unbalanced force are transient data items
stored for each point in the piping system. The user must choose between velocity or cavity
size as a part of the 2D output. This selection is made on the OPTIONAL data form in the
input. Time history or frequency transforms are available as “types” of plots. Each for the
unbalanced force at node 50 is shown below:

Time History of Unbalanced Force Frequency Transform of Unbalanced Force


The valve that is closing is assigned the ID “V1” as can be seen in the input listing. Several
properties for all nonpipe elements can also be plotted from the 2D menu selection. If the user
selects “Non-Pipe 2D Plots” instead of “2D Plots” the following data panel will be shown:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.4.18


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The Non-Pipe ID must be entered in the Item ID: text cell and the type selected. When the
“type” is selected, the list of available plotted data will change according to the selection. For
the valve the above list is shown. The opening for the Valve ID “V1” is shown below:

Note that the valve starts from the fully opened state at time = 0 and closes over the 0.4
seconds as specified in the input. Each nonpipe element in the input may be similarly
interrogated.
3D Plots
The 3D Plot selection initially shows a static 3D rendering of the system analyzed, as shown
below for HAMMER.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.4.19


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The Viewing, Setting and Controls menu selections are designed to allow the user to get the
particular view of the system that should be reviewed. For large systems the user might want to
zoom in on a particular pump station or vessel and “clip” out the remainder of the system. After
so “clipping” the viewed results will only contain scaled data for the model that remains in the
“unclipped” portion of the screen. This is also true for dynamic 3D views of the system. This
“clip” what you want method lets the user limit the displayed output range of data and is very
handy. The user is encouraged to experiment with these options. Transient output is displayed
under the Transient/Steady menu selection. The pull down menu contains the following
choices:

Each brings up a separate selection panel:


Steady State Options

The user can display the steady state solution for pressure, velocity or flowrate. If the 3d
Viewer box is checked, a 3d “rotatable” model will be generated and displayed using the Paulin
Research 3d DirectX viewer.
Transient Options

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.4.20


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The user can display any of the listed maximums from the transient solution. The transient
solution stores the maximum values of pressure, velocity, cavity size and force and can display
them in a static plot. Optionally a 3d “rotatable” model can be generated of the generated
image and viewed using the Paulin Resaerch DirectX viewer.
Animate

The animation window will show any of the four plot-able items, (pressure, velocity or cavity
size, flowrate and force) as a function of time in an animation. The animation speed can be
selected from Normal, Double Quadruple speed or DirectX. The DirectX option invokes the
directX viewer to display the animated image in a user manipulatable model. This is the most
flexible of the viewing options to invoke. From the 3D DirectX modeler the user has numerous
tools for manipulating the animating image.
Time

The time slice panel allows the user to display any of the system properties at a single point in
time during the transient. This plot is useful if the user wants to know exactly how the system
looks at a particular point in the transient.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.4.21


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Controlling Transient Output


There are several characteristics of the output that the user may wish to control. Each
characteristic is described in some detail below.

Maximum Simulation Time:


If the total time is too short, the maximum pressure, velocity, flow or cavity size during the
transient may not be calculated, further if the total time is too short, a fourier analysis of the
time history plot might not contain enough low frequency data. If the total time is too long
then runs may take too much disk space and an inordinate amount of time to post-process
or view. The longer the total time the more conservative the resulting calculation.
The Maximum Simulation Time field on the BOS Fluids "optional" data screen can be left
blank. In this case BOS Fluids will estimate a required total solution time based on its best
estimate of the system transient loading and the acoustic response. In many cases the BOS
Fluids calculated solution times will be satisfactory. In some cases it will not. It is the users
responsibility to review the transient output and assure himself that a sufficient total
calculation time has been used. A patient review of the animated pressure results along with
some quick hand calculations should provide an adequate assurance. In the animated
results the pressure wave should be seen to progress from the source of the disturbance
through the system at least two or three times. (In some long pipeline systems only a single
return of the pressure wave is required.) The total time should also be long enough for any
transient disturbance to stop, and the system to have undergone its worst possible pressure
loading scenario. As the user gains more experience with transient fluid solutions, a greater
confidence will be gained in the comfortable estimation of an optimum total solution time.
Longer times are also required if low frequency data is to be obtained from the time history
plots. The longest period plotted can only be one-half to one-fourth of the total solution time.
(The lowest frequency plotted is the inverse of the largest period plotted.)

Maximum Time Step:


The solution time step is the interval of time between each calculated transient solution. The
user cannot directly control this value, but the user can set a maximum for its value. The
time step is calculated by BOS Fluids as required for an accurate application of the method
of characteristics. The smallest element length in the fluid model and the fluid properties
control the time step. The ISRM mechanism in BOS Fluids automatically combines small
and large pipe lengths together. If the ISRM mechanism cannot eliminate certain small
lengths because of their location in the system flow geometry the user can still indirectly
control the Solution Time Step by setting the Minimum Element Length on the "optional" data
screen. The smallest time step is based on the minimum element length entered. Entry of

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.4.22


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

large values for the Minimum Element Length will speed up the transient solution but may
distort the results around controlling short elements. A typical approach to take is to perform
design runs with longer Minimum Element Lengths, checking the error in the solution at
reasonable intervals by reducing the Minimum Element Length value. In general, little
change in the solution is expected. Minimum Element Lengths up to 100 inches have been
used successfully. The user may never need to adjust this value. The first indication that
adjustment is required occurs when transient solution screens appear showing hundreds of
thousands of time steps required for solution. Normal solution time steps are on the order of
1.0 millisecond. The total solution time steps for a 10 second simulation are: (10.0)/(0.001)
= 10,000. The maximum allowed output time step may also need to be controlled by the user
if there are input actions, i.e. valve closures or tube ruptures that will occur over a very short
time interval. Tube rupture for example occurs in milliseconds. A solution time step of 10 ms
would not be suitable in this case, regardless of the element lengths involved.

Output Time Increment, Start, Stop (Sec):


This input field on the "optional" data screen is used to control the volume of information
output from the transient solution. Long transient solutions can easily produce many
megabytes and in some cases gigabytes of disk file data if some care concerning output
selection is not exercised. BOS Fluids attempts to set the Output Time Increment for the
user automatically based on system parameters and experience. This is a difficult value to
establish however without some prior knowledge of system behavior and the user’s desired
output. If the only information required from the transient solution is the maximum pressure,
or the maximum unbalanced load then the Output Time Increment can be any value,
because the maximum pressure and maximum unbalanced loads for all points in the
system are trapped during each time step. If the user wants an accurate depiction of the
pressure or load time history then a suitable Output Time Increment must be used.
Accurate time histories are needed when the time history data is to be used in a subsequent
mechanical load analysis of the system, or if a frequency decomposition of the time
waveform is needed. BOS Fluids estimates the Output Time Increment based on an
approximated fundamental acoustic mode, and several experience factors. If the force vs.
time function that is produced from BOS Fluids is to be used in a PipePlus or CAESAR II
analysis then the Output Time Increment should be no larger than 1/4 the period associated
with the highest mechanical natural frequency of interest.
The user may also specify a starting and stopping time following the Output Time Increment
entered. These inputs tell BOS Fluids when to start and stop writing time data to the output
data file. This start-stop "window" for writing time history data is used when the transient
phenomena of interest occurs several seconds or minutes into the simulation. One example
where this would be used occurs when vapor columns form over pipe bridges. In this case it
may take several minutes for the vapor column to form, and then only a few more seconds
for the bubble to collapse. There is no need to store the solution before the bubble forms.
The only interesting part of the simulation occurs when the high pressure spike appears on
collapse of the vapor column. If the output time step is 4 ms, and the vapor bubble starts
forming at 40 seconds, and collapses at 50 seconds; and if the total solution time is 1:30,
then the user might specify an Output Time Increment, Start, Stop input to be: 0.004, 40,
90. The tabular results also reflect only the transient that occurred in the time window
specified. This can be useful when a startup transient must be allowed to die-out before the
real transient to be followed is started.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.4.23


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Acoustic Frequencies Example


Introduction
The following piping system is subject to gas flow, valve transients, and changes in operating conditions at
the various inlets. There is concern that acoustic natural frequencies in the system will correspond to
mechanical natural frequencies resulting in large displacements and stresses. The acoustic natural
frequencies can be determined by providing a single impulse to the fluid system and interrogating the
solution for frequency content. The input can be found in the <Installation>\BOSMODELS subdirectory,
and is named: FLUID (FLUID.IFU). The system layout is shown below:

Acoustic waves are reflected differently from open vs. closed ends, and from fixed heads vs. fixed flow
boundary conditions. In general, the steady state flow pattern for the system should be developed and
analyzed before any transient analysis is begun. This is not necessary for this particular system because
the geometry of a Helmholz resonator is recognized. (See the Figure below.)

The fundamental acoustic resonances in these systems will be determined by the volumes of the vessels
and the interconnecting pipe. Boundary effects away from the vessels will be minor compared to the effect
of the volumes on the resonant frequencies. The acoustic frequencies of the combined Helmholtz systems
above can be determined by providing a square wave pressure excitation in the evaporator vessel
described by the nodes 20-90. The acoustic frequencies are determined by doing a Fourier decomposition
on the time waveform of the pressure in the system that results from the square wave excitation. The
square pressure wave will be applied at the node number 90. The magnitude of the square wave pressure
impact should be some recognizable multiple of the total system pressure, but the exact magnitude is not
important in a frequency analysis. The BOS Fluids input for the square pressure wave is shown below:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.5.1


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

NODE: 90
TYPE: TIMEHEAD
TIME: 0,100
HEAD: 800,800
Once this boundary condition is defined, the ACTION:ANALYZE:TRANSIENT option can be set and the job
submitted for analysis. (The file FLUID.IFU is in the <installation>\BOSMODELS subdirectory. All transient
input for the problem is prepared. The user needs only to copy FLUID.IFU into a working subdirectory,
enter input for the job, and then SUBMIT the job for analysis.)
We are interested in what happens to the gas in the pipe between the evaporator and the exchangers and
heater. The exits from the exchangers and the heater will have only a secondary effect on the main
phenomena of interest and can be ignored.
Running acoustic frequency simulations in this manner is surprisingly easy. The user must only
pay attention to the properties of the liquid or fluid and the type of boundary condition, i.e. closed
or open, because open and closed boundary conditions reflect waves differently and can influence
the resulting calculated frequency.

2D Results for Acoustic Frequencies Example


Results from an acoustic "pulse" type of analysis give pressure, flow, velocity or force as a function of time
or frequency. The main system acoustic frequencies will be observed at all points in the system. Higher,
secondary frequencies will be noticeable at points in the system where reflections of the pressure wave
occur from shorter branches, dead ends or intersections in the system.
Pressure vs. time and frequency are given below for the node number 20 and 210.

Pressure at 20 Frequency at 20

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.5.2


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Pressure at 210 Frequency at 210


The effect of reflections from the multiple changes in diameter and different volumes is more pronounced at
210, but the fundamental Helmholtz frequency is still present at about 1.56 Hz.
The plots above are accessed from the 2D Graphical Results Menu.

Select Pipe 2D Plots, then enter the node number where the transient plot is to be generated.

3D Results for Acoustic Frequencies Example


Output plots of maximum pressure and unbalanced force can be displayed. The maximum unbalanced
force shown will not be correct in magnitude because we are only scanning for frequencies, but should be
correct for location in the geometric model. The location of the maximum unbalanced forces will show
where the largest tendancy to vibrate will appear. If a mechanical natural frequency also exists in the piping
system associated with large acoustic shaking forces and in the same direction then large displacements
could be expected.

Computing unbalanced forces in a complex three-dimensional piping system can be considerably


complicated by the geometry of the system. This is because all unbalanced loads require two points – a
starting point and a terminal point. Where bends, tees, crosses and closed valves exist in a system it can
be very difficult for the program to compute the correct unbalanced loads or the node pairs that are
associated with them. Often some of the unbalanced loads in the system will be correct and others will be
missed. When the unbalanced loads are a critical part of the analysis the user should review carefully the
unbalanced load node pairs selected by the program to be used for computing unbalanced forces. If the
pairs selected by the program are not sufficient this list can be augmented by the user. See the “Output
Exclusion – Force List” input data form.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.5.3


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Discussion:
1. Note that the Transient Simulation Time and the Output Time Step are left zero or blank in the input.
BOS Fluids estimates these values from system characteristics. Even though BOS Fluids may do an
good job selecting these parameters, the user should always check the results produced by a BOS
Fluids estimation and should verify that the values selected suit the requirements of the problem
being solved. The total time and output time steps used for each run are printed in the Transient
Warnings & Messages Report.
2. The impulse boundary condition that produced the frequency study is shown below

A "TIMEHEAD" is specified at node 90. TIMEHEAD is used to apply the square wave pulse to the
fluid system. The static line pressure is specified in the static HEAD row, and the square wave is
entered in the transient HEAD row. For this example, the static head is 150psi, and the transient
head defines the square wave. When the transient solution starts from 150psi, the pressure will jump
immediately to 800psi and will remain there. This instantaneous rise in pressure from 150psi to
800psi will excite all of the computable acoustic natural frequencies in the system. Plotted pressure
results will show the undulations in pressure resulting from the amplification effect of the acoustic
resonances.
3. A portion of the Steady State Warnings & Messages for this job is shown below:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.5.4


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The nodes on the piping system at end points where no boundary condition is found are changed to
DEADEND’s. This is the same as a welded end cap. This automatic changing to DEADEND’s is
convenient because the user does not have to specify boundary conditions for most nodes in the
model. CAESAR, FE/Pipe or PipePlus models can be brought in and studied for acoustic frequencies
quickly. There are also a number of nodes where no pressure was calculated. These parts of the
system are in the relief valve exhaust piping. The relief valve stays closed during the steady state
calculation, and so the exhaust piping has no access to any other pressure in the system. If the relief
valve pops open during the transient, the user should reduce the square wave pressure peak until
this does not occur. (Unless the relief valve interaction with the system is to be studied.) The mean
line pressure may also need to be lowered if the relief system is particularly sensitive to the pressure
fluctuation. The position of the relief valve can be displayed in the 2D Graphics plots by selecting
Non-Pipe Transient Data. If the user is concerned about the resonant acoustic state that occurs
when the relief valve is opened, then the relief valve should be replaced with a comparable gate
valve, and the valve left open for both the steady state and transient calculations.
When acoustic frequency studies are done on liquid filled systems, square wave pressure
pulses are used in a similar manner. The user should deactivate the column separation option
in this case so that inadvertent vapor formation does not cloud the solution by introducing
pressure pulses of their own during vapor implosion. The user can certainly study the effect
of such implosions on pressure wave formation and on system response, but generally this is
not done. Vapor formation can produce a very nonlinear response. If pressures will be
produced at any time during a transient that are below the vapor pressure the user should be
aware of its presence. The default vapor cavity model is ON.
To run an acoustic frequency analysis on a system that has opened and closed ends the user
should provide fixed head boundary conditions at all open end sections. If the evaporator,
exchangers and heater in the system above were not modeled then the end connections to the
vessels should be modeled as fixed head ends so that an incoming pressure wave is reflected
correctly. Note however, that if the vessels were omitted, the Helmholtz resonant frequency

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.5.5


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

effect would be missed. Helmholtz resonators are not dependant on the speed of sound in the
medium, and typically can provide very low frequency response in even short piping systems.
Where a low frequency response is detected, the user might want to investigate this
possibility.
Two different acoustic phenomena have just been described, and both can be
simulated with BOS Fluids. A Helmholtz resonator exists because large gas volumes
act as springs in between a “gas mass.” The natural frequency is independent of the
speed of sound, but is a function of the volume, pressure, density and ratio of
specific heats.
An acoustic resonance in the piping is unrelated to the Helmholtz frequency, does
not depend on the volume of the vessels attached to the pipe, and is a strong
function of the speed of sound in the fluid. The resonance is determined by how long
it takes a pressure wave to travel back and forth between the system boundary
conditions.
Helmholtz frequencies are a strong function of the volume of the attached vessels. Where gas
is over water in one or more of the vessels involved, a change in liquid level will change the
gas volume in the vessel and the Helmholtz frequency will change correspondingly. The BOS
user can experiment with this phenomena.
4. A portion of the “Transient Warnings and Messages” report is included below:

The total simulation time chosen by BOS Fluids for this system was 4.375334 seconds. From the plot
shown in the 2D Results above, this looks like a good choice. The fundamental acoustic resonance
was reflected at least eight times during the solution. The solution output time interval was 0.02
seconds. This selection seemed to produce a good "smoothness" to the output plots. Higher
frequency resolution in the frequency plots can be obtained if the user wants to lower the output time
interval value on the Optional Data Screen.
5) A portion of the Steady State report is shown below. Note that the calculated flowrates are zero.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.5.6


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

This is not a problem. The acoustic behavior of Helmholtz resonator is not dependant on the flow
through it. For most systems where the acoustic resonance of a piping system is wanted,
independent of the vessels attached to it, the user must establish a reasonable steady state flow
condition. Once the steady state condition exists, the user can perturb it with a square wave change
in pressure just like was done in the Helmholtz analysis performed here.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.5.7


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

BOS Fluids Theory


BOS Fluids consists of two modules, a steady state module and a transient module. Before a
transient calculation can be started an initial solution (at time t =0.0) must to be available.
The steady state module determines the initial solution.

Fluid Flow through pipes

The fundamental equations in fluid dynamics problems come from conservation of:

mass,
energy and
momentum.

Conservation of mass states that no mass can be created or destroyed. Conservation of energy
means that energy can not be lost but can only be transferred from one form into another.
Conservation of momentum means that momentum can only change due to the action of an
(external) force.

When solving fluid flow problems through piping systems, the mass conservation
equation (better known as the continuity equation) is applied at every node in the system
stating that the total net inflow to a node must sum to zero (assuming that the flow is
incompressible, which means that the density of the fluid is constant).

The conservation of energy equation can be used to calculate the pressure drop over a
pipe section when fluid is flowing through it. A fluid flowing through a pipe contains three
forms of energy: potential energy due to its pressure and elevation and kinetic energy due to
its motion. These three combined result in the often used Bernoulli equation:

pV2
E = gz + +
ρ 2

where E is the energy per unit mass, gz is the potential energy due to elevation, p/ρ is the
potential energy due to pressure and V2/2 is the kinetic energy due to motion.

Applying the Bernoulli theorem between two points in the system gives the following
relation:

p1 V12 p V2
z1 + + − hL + h p = z2 + 2 + 2
ρg 2 g ρg 2 g

Where hL is the energy loss (in height of fluid column) due to friction and hp is the energy
added to the system by pumps.

Friction losses

The energy loss due to friction can be found from one of the following equations:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.6.1


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

L Q2
Darcy-Weisbach: hL = f
D 2 gA 2
Q1.852
Hazen-Williams: h L = 10.7 L
C 1.852 D 4.87

where f and C are friction factors (see explanation below), Q is the flow through the pipe, D
is the diameter and A is the crossectional area of the pipe. Typical values for the Hazen-
Williams friction factor, C are 140 for very smooth and straight pipe and 100 for cast iron
pipes that are already a few years in service.

The Darcy-Weisbach friction factor f can be found from :

64
Laminar flow, Re<2300: f =
Re
1  ε 2.51 
Turbulent flow,Re>2300: = −2 log 10  + 
f  3.7 D Re f 

where ε is the absolute roughness of the pipe and Re is the Reynolds number. BOS Fluids
uses the Darcy-Weisbach friction factor as calculated above. Friction factors for the transient
condition are found from the flow in the steady state condition. If the flow is zero then the
first term of the turbulent flow equation is used.

Flow through valves and fittings

In the same manner as for pipes, valves and fittings generate a pressure loss when fluid is
flowing through them. For fittings the pressure loss is normally described in terms of a so
called equivalent length. The equivalent length of a fitting is the length of a pipe of the same
diameter that generates the same pressure loss as the fitting. The table below lists a few
fittings and their equivalent length:

Fitting Equivalent length


30° elbow 10 x D
45° elbow 18 x D
90° elbow 30 x D

An equivalent length of 10xD means that the specific fitting generates a pressure drop equal
to the pressure drop generated by a straight pipe with a length of 10 times the diameter of the
fitting.

The pressure loss due to a valve is most often written as:

1 1 2
dH = Q
( C A)
d
2
2g

where Cd is the valve discharge coefficient and A is the crossectional area of the valve.

Pumps

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.6.2


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Pump are used to add energy to the system. The energy added by a pump can be described by
the following relation:
h p = a1 + a 2 Q + a 3 Q 2

where a1 ,a2 and a3 are coefficients determined from the pump characteristic curve which is
supplied by the pump manufacturer.

Steady State Solution for Complex Piping Networks

In general a system consists of a number of pipes, valves, pumps etc. All elements that are
not pipes are referred to as non-pipe elements. The pipe and non-pipe elements are connected
at nodes. The objective of the steady state analysis is to determine the flow through all the
pipes and non-pipe elements in the system and the pressure at every node. The equations for
the conservation of mass and energy are used.

The mass conservation equation (continuity equation) can be applied to every node in the
system where there is no pressure (head) specified. For every node an equation is produced
that sets the total net inflow to that node equal to zero:

n
Qnode + ∑ Qi = 0
i =1

Where “i” loops over all elements (pipes and non-pipes) connected to any single node.
Normally the number of equations found by applying mass conservation is not enough to
solve for all flows throughout the system. Additional equations can be found by applying the
conservation of energy. To calculate the pressure at every node in the system, the pressure at
at least one node needs to be given by the user. For systems where the fluid is flowing,
pressures at two nodes must be given, or a pressure and a flow at two different nodes must be
given.

Starting at one node where the pressure or flow is specified, the system can be traversed until
another node where the pressure or flow is specified is encountered. The pressure loss over
each element encountered in the path can be calculated once the path is found. The total
pressure loss is the difference in pressure between the starting end ending node. Doing this
for every node with a prescribed pressure gives the required additional equations to solve the
steady state system.

Introduction to Waterhammer Theory

Waterhammer (or surge) is introduced into the piping system as a result of a change in the
steady state conditions. A change in velocity or pressure at one point results in pressure
waves (positive or negative) emanating from that point and traveling through the system.

When a fluid is flowing in a steady manner through a valve that is closed suddenly, the fluid
flow is interrupted. The flowing fluid has a certain velocity and momentum that must be
transferred from the kinetic energy of motion into the potential energy of an increase in
pressure. Stopping the fluid means that the momentum of the fluid column flowing must be
changed, and this can only be done by a force.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.6.3


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

When a fluid column of length L is to be decelerated from its initial velocity to rest (velocity
0) the momentum to be balanced is:

momentum = mass∗ velocity = ( A∗ L∗ ρ )∗V

This momentum is balanced by the application of a pressure force ∆p*A during a finite time
interval specified as ∆t. Upon application of the pressure increment ∆p, and due to the
compressibility of the fluid and elasticity of the pipe, a pressure wave is propagated in
upstream direction at a propagation speed a. The time increment ∆t involved is equal to the
time required for the propagating pressure wave to cross the full length of the column L.
Equating force times application time to the momentum, the following equation is found:

L
A∗ L∗ ρ∗V = ∆p∗ A∗
a
∆p = a∗ ρ∗V

The result is the famous Joukowsky equation. Keeping in mind the assumptions made for the
derivation, it may be clear that the pressure increase calculated by use of the Joukowsky
equation is rather conservative.

The pressure rise calculated from the Joukowsky equation is often referred to as the potential
surge. It is the actual value of the surge pressure increase that will be found by the more
accurate BOS Fluids evaluation.

Wavespeed

The speed at which pressure disturbances travel through a pipe can be found from the
continuity equation. The continuity equation states that the net flux of mass into a control
volume must equal the time rate of increase of the mass within that volume.

L
L∗ ∆A∗ ρ + L∗ A∗ ∆ρ = ρ∗ A∗V ∗
a

where the first term is due to the pipe elasticity, the second term is due to the fluid
compressibility. Substituting the Joukowsky equation and rearranging results in:

gh
a2 =
∆ρ ∆A
+
ρ A
Recognizing that:
∆A D
=
A∆p Et
ρ∆p
=K
∆ρ
∆ρ
gh =
ρ
results in:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.6.4


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

K
ρ
a=
DK
1+
t E

where D is the pipe diameter, t is the wall thickness of the pipe, E is the modulus of elasticity
of the pipe material and K is the bulk modulus of the fluid. This equation is normally used to
calculate the speed of pressure disturbances when dealing with transient fluid flow problems.

Transient Solutions for Complex Piping Networks

The equations of motion and continuity can be written as (see any text on fluid dynamics):

f
gH x + Vt + VV = 0
2D
a2
Ht + V =0
g x

where H is the pressure head (in meters of fluid column), V is the velocity of the fluid, f is
the friction coefficient (see above) and a is the wave speed. Indices x and t refer to
differentiation with respect to x and t respectively.

The two equations may be combined using an unknown multiplier λ:

 g   a2 fV V 
λ  H x + H t  + V x λ + Vt +  =0
 λ   g 2D 

Choosing an appropriate value for λ results in simplification of the above equation. In general
the variables V and H are functions of x and t. If the independent variable x is permitted to be
a function of t then:
dH dx
= Hx + Ht
dt dt
dV dx
= Vx + Vt
dt dt
g
Substituting these equations into the combined equation shows that if λ is chosen as ± , the
a
equation simplifies to an ordinary differential equation:

g dH dV fV V
± + + =0
a dt dt 2D
dx g
Keeping in mind that = , two sets of ordinary differential equations result:
dt λ

 g dH dV fV V dx
C+  + + =0 for = +a
 a dt dt 2D dt

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.6.5


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

 g dH dV fV V dx
C − − + + =0 for = −a
 a dt dt 2D dt

dx
As the wavespeed, a is constant for a pipe section, the lines = ± a plot as straight lines in
dt
the xt plane. These lines are called characteristic lines. Along these lines the simplified
equations are valid.

Solving Transient Equations for a Pipeline

A pipeline is divided into an even number of reaches, N. Each reach is ∆x in length. A


timestep is computed from ∆t = ∆x/a. If the variables V and H are known at a certain time,
then the transient flow equations can be integrated along the positive sloped characteristic
line between the limits A and P. Similarly the for the negative sloped characteristic between
limits B and P.

P
t=∆t

C+ C-

t=0
X=L
X=0 A B

∆x

Simultaneous solution of the equations along the positive and negative sloped characteristic
lines yields the solution at point P:

a f∆x
HP − HA + (Q P − Q A ) + Q Q =0
gA 2 gDA 2 P A
a f∆x
HP − HA − ( Q P − QB ) − Q Q =0
gA 2 gDA 2 P B

Q
Where the velocity V is replaced by V = , with A the crossectional area of the pipe.
A
The above equations are the basic relations that describe the transient propagation of head
and flow in a pipe line. These equations also satisfy the steady state conditions since steady
state conditions imply that QA = QB = QP.

By introducing the pipeline characteristic impedance B and resistance coefficient R:

a
B=
gA
f∆x
R=
2 gDA 2
the equations may be written as:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.6.6


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

H P = C P − B P QP
H P = C M − B M QP

where
C P = H A + BQ A BP = B + R QA
C M = H B − BQB B M = B + R QB

Solving the two equations for HP yields:

CP B M + C M BP
HP =
BP + B M
CP − CM
QP =
BP + B M

The above formulation of the transient flow equations is well suited for solution by computer.
Point P refers to any grid location on the pipeline. At locations where different pipes join or
pipes and non-pipe elements (valves etc.) are connected, only one characteristic equation is
available. Boundary conditions and/or relations describing the pressure loss over non-pipe
elements (see section on steady state calculations) determine the pressure and flow at these
locations.
dx
Sectioning of the pipes and determination of the timestep, satisfying = a for every pipe, is
dt
automatically done by BOS Fluids. The only data required by the user is the system layout
(pipe diameter/wallthickness/length, valve data etc.) and the locations where the pressure or
flow is given. Based on the transient condition(s) the user wants to analyze (valve
closure/opening, pump start/trip) the steady state condition, and the transient minimum,
maximum and time history of the pressure and flow can be calculated for every node in the
system and reported back to the user in the form of a report and/or 2D or 3D graph. If the
user specifies the location of bends and tees the unbalanced forces can also be calculated.

These forces can be entered into a pipe stress program to analyze the response of the
mechanical system to the transient conditions. For above ground pipework, the pipe stress
program will calculate the displacement, forces and stresses at each node in the system and
the user can check whether the support system is adequate.

BOS Fluids also provides several vapor cavity models (not discussed here). If the pressure
drops to the vapor pressure of the fluid, vapor cavities are formed. The growth, destruction
and movement of these cavities is taken into account in the surge analysis. Large pressure
pulses can be developed when vapor cavities implode.

A standard reference text on the subject is “Fluid Transients in Systems,” by E. Benjamin


Wylie and Victor L. Streeter.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.6.7


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Safety-Relief Valve Notes and Examples


For the usual high-lift relief valves the lift is greater than one-fourth the nozzle diameter,
which results in a flow area between the nozzle and the valve disk greater than the nozzle
area.

Liquid relief valves reach their rated capacity at 25 percent overpressure. Overpressure is
the pressure increase over the set pressure.

Liquid relief valves rise slowly as the pressure increases, and is different from a safety
valve which attains the rated lift with little overpressure.

Safety valves are classified as “conventional” or “balanced,” depending on the effect of


back pressure on their performance.

Flow Capacity Calculations: “On discharge, the flow in the throat of the nozzle reaches a
constant value corresponding to the critical flow pressure for a given upstream pressure
so that, under adiabatic flow conditions, the flow through a safety relief valve is directly
proportional to the absolute upstream pressure. For the 3% pressure-drop limitation
suggested in 2.2 the gas discharge is evaluated by the formula:

W = CKAP1 (M/ZT)0.5

W = flow through valve in pounds per hour


C = Coefficient determined by the ratio of the specific heats
K = coefficient of discharge for the valve
A = effective discharge area of the valve
P1 = upstream pressure in psi absolute. Usually the set pressure mmultiplied by 1.10 or
1.20 depending on the permissible overpressure plus the atmospheric pressure.
M = molecular weight of the gas or vapor
Z = compressibility factor
T = absolute temperature of the inlet vapor, degF + 460.

In the event Z is not available for a gas, the value of 1.0 is used conservatively.

The critical flow pressure PCF in absolute units may be estimated by the following perfect
gas relationship:
PCR / P1 = ( [2]/[k+1] )(k/k-1)

If the downstream pressure of the valve exceeds PCR then flow will be restricted and a
readjustment will be needed. BOS Fluids can certainly be used in the full flow steady
flow state to determine the pressure drop from the vessel or process line connection to the
valve.
For balanced type safety valves where the back pressure is isolated from the valve, the
BOS model should include an intermediate FIXEDHEAD at the downstream side of the
PSV to keep the backpressure on the valve constant. The pressure downstream of the

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.7.1


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

safety valve will need to be checked in this unusual situation. Pressures downstream of a
choked safety valve can be checked against the B31.1 II-2.2.1(1) equation for pressure
downstream of the valve.

The set pressure is the pressure where the valve just begins to open. The full flow, rated
or opening pressure is the pressure where the rated flow is established. This is a function
of the manufacturers design, but typically for gases the set and opening pressure
difference is between 3 and 5%, and for liquids it is around 10%. Reseat pressure
differences range from 6 to 20%.

If the user does not enter mass and stiffness for the valve, the response time is based on
the rate of pressure rise. When the response time of the valve may be important with
respect to the response time of the event, (usually on the order of 10-to-20 ms or less),
then the user should enter the mass and stiffness associated with the valve. Simulations
can be performed with the valve model in BOS Fluids to approximate the manufacturers
test stand if approximate opening times are known.

If the spring constant is not known it can be estimated from:

K = (Prated-Pset)(1.2)(Aorifice)/[(0.27)(dorifice)]

The weight of the disk and assembly can be estimated from:

M = (2.5)(Aorifice)(0.3) = Weight in pounds, (Aorifice in inches)

The user employing the above should be cautioned that these are approximations only to
be used when actual data is not available, and that the resulting estimates should be
checked carefully before being used in any application.

The basic algorithm is as outlined below:

The bore is taken from the valve data sheet. This should be input by the user.

The maximum lift at the rated flow is estimated from:

xr = 0.27 * bore

The displacement of the valve disk is found from:

x = (xr)(TAU)(1/em)

The critical damping is calculated.

critical = (2)[(stif)(mass) ]0.5


damping = (critical)(percent_critical)

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.7.2


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

If the mass, damping or stiffness values are entered then the value of tau is found so that
the head drop across the valve is equal to:

(M)(a) + (damp)(v) + (K)(x) – Q2/[(g)Area2] + Puse

where M is the mass of the disk, (a) is the acceleration of the disk, (damp) is the entered
damping, (K) is the stiffness, Q is the flow, (g) is the acceleration due to gravity, (Area)
is the throat area of the valve, and Puse is the set pressure if the valve is opening and the
reseat pressure if the valve is closing.

If the user does not enter a value for either of the mass, stiffness or damping, then the
head loss over the valve is between the set and full flow, or reseat and full flow pressure
depending on the direction of valve movement.

The calculated flowrate through the valve is also limited by the sonic velocity in the fluid.

A schematic of the basic relief model is shown below:

The following example will illustrate its use:

The tank elevation is 35ft. This converts into a static head of: (35)(62.4)/(144) = 15.16
psig. The quick closure shutoff valve between nodes 10 and 15 can be closed in one
second. When the pressure builds up sufficiently the relief valve will lift at 250 psi. It’s
rated opening pressure (full flow pressure) difference is 315 psi, and the reseat pressure
on closing is 225 psi. When the valve shuts will the relief valve adequately redistribute
the flow. A LONGPIPE boundary condition will be used at node 25 with the rated flow of
3.3 cu.ft./sec. The LONGPIPE boundary will prevent reflections and will support any
pressure or flow presented to it downstream. This will represent a long incoming
pipeline.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.7.3


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The simulation valve at 10-to-15 will be closed in one second, from (1) –to- (2) seconds
during the transient. As a first pass, the transient will be run for 4.0 seconds to determine
if a steady state condition is reached. The model input file is TANKFILL and can be
found in the BOSMODELS folder under the installation directory.

The pressure at node 20 after the run is shown below:

Pressure at 20 (psig)
The valve is just opened, but the increased pressure supports the necessary flowrate. The
flowrate through the relief valve is plotted below:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.7.4


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Flow at Relief Valve (cu.ft./sec.)

Even though the flow during steady state conditions was 3.3 cu.ft./sec. the flow through
the relief valve is only 1.2 cu.ft./sec as the flow through the inlet line from 25 is stalled.
The flow at 25 is shown below:

Flow at 25 (cu.ft./sec.) (Note: All negative (delivery) flows shown.)

The flow at 25 is reduced when the valve is closed and the pressure rises, and then begins
to recover. The pressure wave that occurred when the valve closed is still traveling back
through the line looking for a location in the system to reflect, but this will not happen
with the LONGPIPE boundary. The reflection will occur at 2L/c, where L is the distance

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.7.5


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

from the valve to the pressure let down station, change in volume, pumping station, etc.
The pressure at the long pipe connection is shown below.

Pressure at LONGPIPE Boundary Condition (TANKFILL)

Depending on the boundary the returning pressure wave will either be a rarefaction wave
with an accompanying drop in pressure, or may be a compression wave, with an
accompanying increase in pressure. At a pressure let-down station, the boundary
condition would be more accurately simulated by a fixed flow. The pressure reflection at
a FIXFLOW node 25 is shown below. (TANKFIL2.)

FIXFLOW Boundary Condition at Node 25 (Supply) (TANKFIL2)


The FIXFLOW boundary condition acts like a closed end that imparts a velocity onto the
flow. The maximum pressure at node 20 is still not high and is shown below:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.7.6


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

FIXFLOW Condition at 25 Pressure recorded at 20. (TANKFIL2)

What happens if the flow is cutoff in 1/3 second instead of 1.0 sec? The resulting pressure at 20
is shown below:

Pressure at 20 for 1/3 second closure time. (TANKFIL3)

The faster closure time interacts with the Pressure Relief valve that is responding instantly to the
pressure difference across the valve. Pressures are dropping below the vapor pressure at node
30 (the inlet to the relief valve), and vapor is forming and collapsing producing pressure spikes.)
To produce a more meaningful answer in more rapidly acting cases a relief valve that is more
reluctant to changing state can be introduced. Using the guidelines above for computing
stiffnesses and masses for the relief valves:

K = (Prated-Pset)(1.2)(Aorifice)/[(0.27)(dorifice)]

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.7.7


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

= (315-250)(1.2)(13.9) / [(0.27)(4.2126)] = 953 lb./in.

The weight of the disk and assembly can be estimated from:

M = (2.5)(Aorifice)(0.3) = Weight in pounds, (Aorifice in inches)


= (2.5)(13.9)(0.3) = 10.4 lb.
A typically damped mechanical system has about 3% critical damping. Since vapor pockets are
being produced for a 0.3 second opening it is desired to provide added damping to the valve to
slow its travel. A 100% critically damped relief valve will be provided by pilot control. The
resulting pressure at 20 if the valve closes in 0.3 seconds is shown below. If the shutoff valve is
closed in 1.0 seconds or longer there is no requirement for a slow acting valve.

100% Critically Damped Relief Valve (TANKFIL4)

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.7.8


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

SRG and SHW File Formats


This section is intended to give the user a brief overview of the data format used in the
SRG and SHW files listed in the BOS Fluids output. These files are listed because
occasionally an error message will be printed that points to a particular line in the SRG
file usually, and the user that has a little familiarity with that file structure will be able to
trouble-shoot the problem from there.

Only truncated data lines are written to the SRG and SHW files. If the user wants to see
the full length of the data lines written then an editor like NOTEPAD should be used to
see the extended data sets. The files that contain the aforementioned data will reside in
the current working directory and will have the name <Jobname>.SRG or
<Jobname>.SHW. The SHW file only exists if the ISRM is turned on. This is the
compressed shadow file.

A typical piping system will consist of nodes, pipes and non-pipes. Non-pipes are
elements that follow some special relation between the flow through the element and the
pressure drop across it. Examples are valves, tube ruptures, pumps, equipment, etc. The
input is not case sensitive, and any line beginning with either a !, # or / character is
considered to be a comment line.

All data in the SRG file is in metric units.

The exact format of each command has been extended in most situations to include some
added plotting and geometry data, but for the most part the user will only need to be able
to identify the parts of the commands that are defined below.

TITLE

This keyword may be used to give the simulation a title. The title will be printed in the
output report:

NODE

Used to define a node. There are a number of different node types available, and they are:
FIXHEAD, JUNCTION, TIMEHEAD, TIMEDISCH, OSCILHEAD, OSCILDISCH,
DEADEND, LONGPIPE, LONG2PIPE, SONICHEAD, SLUGHEAD. The general input
fomat is:

NODE <ID> <type> <type specific data> <elevation>

If a TIMEHEAD type node is identifier, then a listing of the time vs. head should also be
provided. It’s format should be:

TIMEHEAD <ID>
<time1> <head1>

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.8.1


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

<time2> <head2>
…. …..
END

A linear interpolation will be used for values of the time that are not in the list. The head
does not change for time larger than the last time specified in the list.

TIMEDISCH, OSCILHEAD and OSCILDISCH should also have time and property lists
provided.

A JUNCTION node is used when a node is not any of the above. These are typically
nodes that are somewhere inside the piping system not at a boundary. In this case the
major function of the JUNCTION node specification is to define the elevation of the
position.

PIPE

This keyword is used to define pipe elements. It is assumed that the pipe is straight and
the elevation of intermediate points is computed by linear interpolation of the elevation of
the end points. The flow at any position along the pipe is assumed to be positive if the
direction is from the first node towards the second node. This order should be as the user
has given it in the BOS Fluids input.

PIPE <1st node> <2nd node> <length> <diameter> <wall thick> <modulus> <roughness>

CKVALVE

This keyword is used to define a variety of valve types, and in particular check valves.
The flow through the valve is related to the pressure (head) drop over the valve by the
equation: Q = (Cd)(A) SQRT( 2(g)(dH) ) ; where Q = flow through the valve, Cd = valve
discharge coefficient, A = cross sectional flow area of the valve, g = constant acceleration
due to gravity, and dH = pressure head drop across the valve. As a valve may be opened
or closed during the transient the value of Cd can vary during the transient. To effect this
in BOS Fluids a value of tau is calculated where: tau = [CdA] / [CdA @100%]. The
variation of tau with time between opening and closing steps during the transient is
calculated from: tau = taui – (taui – tauf) [ t / tc ) em , where taui is the initial value of tau,
tauf is the final value of tau over the increment, tc is the duration of the increment, and em
is the valve exponent given by the user..

CKVALVE <1st node> <2nd node> <CdA@100%> <tauSS> <direction>; where tauSS is the
value of tau to be used in the steady state solution. If the direction is 0.0 then flow
through the valve can be in either direction. (The valve is NOT a check valve.) If
direction is 1.0 then the valve is a directional check valve and the flow is only allowed to
go from the 1st node to the 2nd node. If the direction is -1.0 then the valve is a directional
check valve and the flow is only allowed to go from the 2nd node to the 1st node.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.8.2


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

ORIFICE

This keyword isused to define an orifice element. An orifice is equivalent to a valve only
if the valve cannot be opened or closed. Therefore the value of tau for an orifice is fixed.
The flow through an orifice can be in any direction.

ORIFICE <1st node> <2nd node> <CdA>

INLET/OUTLET

These keywords define inlet and outlet elements. They may be used at boundary points of
the model but do not have to be. The flow through the element should be given for the
steady state condition. From this value the head loss over the element during steady state,
a value of CdA is computed. In the transient simulation these elements are then treated as
orifice or “fixed port” valves with an equivalent and constant CdA. These elements can
be used to model fixed flow heat exchangers, etc, where the exact resistance to flow of
the heat exchanger is not known, but the flow through the exchanger is known. An
INLET element is used when the flow is positive from the 1st node to the 2nd node, and an
OTULET element is used when the flow is positive from the 2nd node to the 1st node.
The forms of the commands are:

INLET <1st node> <2nd node> <flow>

LUMPED

This keyword is used ot define lumped inertia elements. Lumped elements are pipes that
have an infinite wave celerity. This makes them useful for modeling very short pipes so
that the time steps do not have to get prohibitively small. Lumped elements may not be
used next to non-pipe elements.

LUMPED <1st node> <2nd node> <length> <diameter> <wall thick> <roughness>

VESSEL

This keyword is used to define surge or damper elements. The vessel element is
implemented as a duel node element. No friction losses are accounted for at the entrance
of the vessel. If there should be some friction losses at this point then the user should
provide them by adding the short nozzle neck into the vessel with an orifice. The air
above the liquid in the vessel model is treated as an ideal gas following the polytropic gas
law: PVk = constant. There are several forms of the VESSEL functionality. If a fixed
head is specified at the 2nd node on the vessel with a pressure, then BOS Fluids assumes
that a blowdown simulation is to be performed with the vessel element. If a 0.0 pressure
fixhead boundary condition is provided for the 2nd node on the vessel element then BOS
Fluids will calculate the internal pressure for the 2nd (gas) node from the steady state
solution and then use this as the starting gas pressure in the steady state solution. This
will simulate a nitrogen charged cylinder for example. Alternatively, the user can specify

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.8.3


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

a pipe connection to the second node and the polytropic expression will relate the
pressures between the 1st and 2nd nodes.

VESSEL <1st node> <2nd node> <Volume> <Liquid Height> <Interface Area> <Cp/Cv>

WAVETOL

This keyword specifies the tolerance for the speed of sound calculation. The method of
characteristics requires that for every element in the system (a)(dt/dx) = 1.0, where (a) is
the speed of sound, dx is the distance between two internal points on the pipe, and dt is
the time step for the transient simulation. Starting from an initial distribution of internal
points for the pipe elements, the values of (a), (dx) and (dt) are adjusted so that (a)(dt/dx)
is the same for every pipe element, but because the wave speed is a physical constant,
BOS Fluids will only change the speed within the limits specified by wavetol. It is not
advised to allow for more than a 10% change in wave celerity. WAVETOL 5.0 means
that the wave celerity of the pipes may be adjusted by the program by at most plus or
minus 5%.

LENGTHTOL

This keyword is used to specify the tolerance for the length of the pipe elements. The
default is 10%.

LIQUID

This keyword is used to specify the liquid or gas that runs through the system. The
physical properties fo the liquid are found in the liquid database file. The physical
properties listed in the database are described in Chapter 2 Section 1, but are:
1) Density in kg/cu.m
2) Bulk Modulus in MPa,
3) Vapor Pressure in meters absolute.
4) Saturation Pressure in meters absolute
The properties must be listed for each temperature for the material.

STDYTOL

This keyword specifies the tolerance for the steady state computation convergence. The
steady state calculation will be assumed converged when the maximum difference
between two successive iterations is less than stdytol. The default value is 1.0E-4.

TMAX

The maximum transient solution time.

ITMAX

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.8.4


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The maximum number of iterations allowed to find a converged steady state solution.
Typically the linear solver used in BOS Fluids converges in 10 iterations or less. The
default is for 100. If steady state solutions do not converge in 100 iterations then there is
most likely a problem with a pressure regulating valve, a check valve or a pump is way
off the pump curve. In this case the steady state warnings and messages report should be
checked.

ACTION

This keyword defines the action that the transient simulation should analyze. Available
actions are OPENVALVE and CLOSEVALVE. The form of the command is:

ACTION <type of action> <ID> <timeS> <timeE> <tauE> <em> ; where ID is the
identifier for the valve to have the action applied, timeS is the starting time for the action,
timeE is the ending time for the action, tauE is the ending value of tau for the valve, and
em is the exponent for this change of valve state.

VCM

This keyword specifies that the Vapor Cavity Model should be used during the transient
solution if the pressure drops to the liquid vapor pressure. The Vapor Cavity Model is
defined in Wylie & Streeter “Fluids Transients,” McGraw-Hill, New York, 1978. In the
case that a valve is opened or closed very quickly, the local pressure (head) may drop
below the vapor pressure and air will be released from the liquid as well as vaporization
of the liquid will occur. The Vapor Cavity Model is a simple model to account for air
present at points where the local pressure is low.

CAP

This keyword specifies that the Concentrated Air Pocket model should be used during the
transient simulation. This model is more advanced than the Vapor Cavity Model and does
support cavity movement through the pipe and large cavity formation, i.e. drainage. The
large drainage cap model must be used for slug flow simulations. The CAP model also
accounts for the “spongy” effect of free air in the flow even when the pressure does not
reach the fluid vapor pressure.

CAP <reference pressure> <void fraction of free air at reference pressure>

ISRM

This keyword indicates that an intelligent system reduction should be performed on the
2Ddata once fluid properties are calculated. This must be the first line in the file when an
ISRM is to be performed. It is only in this instance that an SHW file is produced.

PINBAR

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.8.5


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The presence of this keyword means that pressure units are input in bars instead of meters
of fluid column.

UNIT

These keywords give the conversion of the different types of units from program units in
metric to user units, which may be either English or metric.

LARGE_DRAINAGE_MODEL <n>

This keyword is used with the CAP cavity model to indicate that the program should
keep track of cavity movement and provide for the possible complete drainage of section
of the pipework.

The user can optionally add a value for the type of drainage desired. These inputs
correspond to the user’s definitions:

1-GLUG flow – The cavity creation and collapse is driven by the CAP model.
2-PLUG flow – There is more mass transfer and a liquid with gas cover.
3-SLUG flow – There may be even more mass transfer and liquid with gas cover.

TEEVORTEXTIMES <start> <stop>

The start and stop times for the tee vortex shedding to start. Tee vortex shedding will
exist in the model providing the <start> time is less than the total solution time and
<stop> time is greater than the <start> time. This feature can be used by the user to
observe the effect of tee vortex shedding on the transient solution. This variable is only
used in the transient solution. Tee vortex shedding energies exist at three fundamental
frequencies that are a function of the Strouhal number for the flow past the branch. BOS
Fluids finds all flows past side branches and applies excitation to the flow.

TEEVORTEXPARAMS <fact> <tol>

<fact> is a multiplier times the inside diameter to get the inside radii for each flow-by
branch stub. The magnitude of the pressure perturbation is a function of this internal
radius. <tol> is the tolerance on the Strouhal excitation frequencies produced. The
solution is typically evaluated at tolerances values of 0.9 and 1.1. The frequencies of
excitation are found to vary within these ranges.

CVVORTEXTIMES <start> <stop>

The start and stop times for the Control Valve vortex shedding is entered. Control valve
vortex shedding will exist in the model providing <start> time is less than the total
solution time and <stop> time is greater than <start> time. This feature can be used by
the user to observe the effect of control valve vortex shedding on the transient solution.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.8.6


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

This variable is only used in the transient solution. Control Valve excitation tends to be
flat and broadbanded.

CVVORTEXK1K2 <k1> <k2>

MVLOADMULTR <multiplier>

The unbalanced load on elbow-elbow or branch-elbow, or branch-branch pairs contains


momentum and pressure loading terms. The factor entered for this keyword is multiplied
into the momentum flux terms at each elbow-elbow string found in the model. The
momentum flux term tends to be small for steamhammer and waterhammer episodes, but
large for slug flows.

PALOADMULTR <multiplier>

The unbalanced load on elbow-elbow or branch-elbow, or branch-branch pairs contains


momentum flux and pressure loading terms. The factor entered for this keyword is
multiplied into the PA term. For certain slugging situations the PA term can be ignored
since the downstream pressure will be about equal to the upstream pressure. The user is
encouraged to investigate this effect on design unbalanced force loads and to select a
value that is safe and that encompasses all possible process and flow scenarios.

DRAINAGEK <multiplier>

STARTUP

This keyword appears when the steady state solution starts in a NOFLOW condition. This
tells the software to set the pressure based on the elevation and then to startup the
transient solution.

A partial example listing of a typical SRG file is shown below:


ISRM
# Pressure is measured in force/area (Output unit is bars.)
PINBAR
# Cavity size is output instead of velocity
3DVAPOUT
# Users Unit Conversions
UNIT FLOW 0.1000000E+01 m.
UNIT PRESSURE 0.9999844E-05 bar
UNIT FORCE 0.9998652E+00 N
UNIT LENGTH 0.1000000E+01 m
# Basic Pipe Data
pipe 1 5 100010 0.1995428E+03 0.3210000E+03 0.1900000E+02 0.2099882E+06
0.5000000E-01 0.3000000E+00 0.0000000E+00 1. 1 0.0000000E+00
0.0000000E+00 -.1995428E+06 1.0 0.2100000E+02
bend 2 100010 300010 0.7181679E+00 0.3210000E+03 0.1900000E+02 0.2099882E+06
0.5000000E-01 0.3000000E+00 0.0000000E+00 1. 1 0.0000000E+00 0.0000000E+00 -
.4572000E+03 0.0000000E+00 -.4572000E+03 -.3143311E-02 0.2100000E+02

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.8.7


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

pipe 3 300010 100015 0.5856001E+00 0.3210000E+03 0.1900000E+02 0.2099882E+06


0.5000000E-01 0.3000000E+00 0.0000000E+00 1. 1 0.0000000E+00 -
.5856001E+03 0.0000000E+00 1.0 0.2100000E+02
bend 4 100015 300015 0.7181679E+00 0.3210000E+03 0.1900000E+02 0.2099882E+06
0.5000000E-01 0.3000000E+00 0.0000000E+00 1. 1 0.0000000E+00 -.4572000E+03
0.0000000E+00 0.0000000E+00 0.3051758E-04 -.4572032E+03 0.2100000E+02
pipe 5 300015 100020 0.1408560E+02 0.3210000E+03 0.1900000E+02 0.2099882E+06
0.5000000E-01 0.3000000E+00 0.0000000E+00 1. 1 0.0000000E+00
0.0000000E+00 -.1408560E+05 1.0 0.2100000E+02
bend 6 100020 25 0.7181679E+00 0.3210000E+03 0.1900000E+02 0.2099882E+06
0.5000000E-01 0.3000000E+00 0.0000000E+00 1. 1 0.0000000E+00 0.0000000E+00 -
.4572000E+03 0.0000000E+00 0.4572000E+03 -.3143311E-02 0.2100000E+02
CKVALVE slug 25 30 0.6290175E-01 0.0000000E+00 0.0 0.0
0.2830000E+00 0.0000000E+00 0.5000000E+03 0.0000000E+00 0 0 0 0 0 0
bend 8 30 300035 0.7181679E+00 0.3210000E+03 0.1900000E+02 0.2099882E+06
0.5000000E-01 0.3000000E+00 0.0000000E+00 1. 1 0.0000000E+00 0.4572000E+03
0.0000000E+00 0.4572000E+03 0.0000000E+00 0.0000000E+00 0.2100000E+02
NODE 5 FIXHEAD 0.3500000E+02 0.0000000E+00
NODE 100010 JUNCTION 0.0 0.0000000E+00
NODE 300010 JUNCTION 0.0 -.4572000E+00
NODE 100020 JUNCTION 0.0 -.1500000E+01
NODE 25 JUNCTION 0.0 -.1042800E+01
NODE 30 JUNCTION 0.0 -.5428001E+00
NODE 300035 JUNCTION 0.0 -.8560007E-01
NODE 100040 JUNCTION 0.0 -.8560007E-01
NODE 300055 JUNCTION 0.0 -.1585600E+01
NODE 60 TIMEHEAD 0.0000000E+00 -.1585600E+01
TIMEHEAD 60
0.0000000E+00 0.0000000E+00
0.1000000E+01 0.0000000E+00
0.5000000E+01 -.9999999E+00
0.5000000E+03 -.9999999E+00
end
STARTUP
ACTION OPENVALVE slug 0.5500000E+01 0.9999990E-01 0.1000000E+01
0.2000000E+00
liquid User 0.2100000E+02
CAP 0.0000000E+00 0.5500001E-03
lmin 0.6420000E+00
itmax 0.1000000E+03
stdytol 0.9999999E-04
wavetol 0.1000000E+02
lengthtol 0.5000000E+01
# Maximum time step
dtmax 0.0000000E+00
tmax 0.1500000E+02
sprint 0.9999999E-03,0.1000000E+01, 0.5000000E+03, 1
method 1
freqprint 0.0000000E+00
LARGE_DRAINAGE_MODEL
end!

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 2.8.8


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

BOS Fluids SPLASH 2d CFD Calculation Version 2.2


Overview:
SPLASH is a 2d CFD Free Surface computer program based on the SOLA-VOF algorithm. (Ref. 1)
(SOLution Algorithm-Volume Of Fluid, Hirt and Nichols, 1981; and Nichols, Hirt, and Hotchkiss,
1980). It is designed to allow the user to simulate arbitrary geometries, and has preset models for
vertical and horizontal vessels, tanks, and pipes with elbows. Several such models are shown below:

Baffles in a Horizontal Vessel Insufficient Freeboard in 10 m. Tank

Tank Fill and Spillover Sequences Slug Impact on Elbow

Available Simulations Include:

1) Sloshing Frequency for Fluids in Tanks and Trays


2) Wave Motion Periodic Loading on Tanks or Vessels
3) Slug Loads on Elbows
4) Slug Loads on Pipe U-Bends
5) Seismic/Periodic Time History Loading
6) Falling Fluids
7) Tank Fill Simulations
8) Baffle Design
9) Slug Breaker Design
10) Loads on Flow Obstructions
11) Deceleration of Tank Cars or Containers
12) Branch Flow Splitting
13) Flow around Baffles

Features:

1) Interactive Data plotting


2) RTF Report Generation with User Selected Embedded Graphics
3) Stored movies for easy replay
4) User Solution Control
5) Preset Models
6) Frequency, Shear and Overturning Moment Calculations

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


3.1.1
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

SPLASH is written for the piping and pressure vessel engineer that needs to study free surface,
sloshing or fluid impact problems in tanks, pressure vessels, pipes and related components. Default
settings for tolerances and maximum pressure smoothing algorithms are designed to provide quick,
reasonable solutions with a minimum of user interaction. Examples are given that show the user how
to adjust tolerances and be sure that a reasonable and conservative solution is obtained.

The SPLASH user should remember that the program was developed to deal with the analytically
difficult free surface problem. If systems are modeled that are completely filled with liquid and that
have no free surface, the program will tend to look for a free surface and will create one where a free
surface is likely to exist.

The Internation Building Code 2000 Paragraph 1621.3.11 states in part that, “Seismic effects that shall
be considered in the design of a boiler or pressure vessel include …dynamic effects …and the sloshing
of liquid contents” SPLASH is designed to produce the seismic shear and overturning base
moments needed by most standard design programs to satisfy these and other Code requirements. El
Centro time history spectra may be applied easily to any model along with ship-board pitch and roll
loads, global accelerations, or velocities indicative of slugs or impinging liquid loads on vortex
breakers or diverter plates.

SPLASH may also be used to simulate the decelerating fluid in a rapidly stopping tank car or truck
and can be used to design baffles to reduce system sloshing periods and minimize loads.

SPLASH is surprisingly easy to use, but the user is encouraged to read through the examples in this
manual to get a better understanding of the program and the options and methods available that can be
used to document, verify and present the results.

Running a Calculation:
The startup sequence from the PRGMaps main menu is shown below:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


3.1.2
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The user should select one of the three model types listed in the upper left hand corner of the main
SPLASH data form. The models are:

1) Tank Sloshing – Builds a model of a tank from only the user dimensions.
2) Slug Elbow Impact – Studies the effect of slugs traveling down straight pipe and impacting
elbows.
3) User Defined Shapes – Allows the user to select from general shapes, but provides a grid editor
whereby the user can define boundary conditions, obstacles and pressure transducer locations.

Filenames are normally provided once a job has been run and the data should be saved, or when an old
model should be retrieved. If the “User Defined” model type is selected, the form changes as shown
below:

The liquid and loading data in the top and bottom frames are independent of the model type and are the
same for each of the three different types. The data in the middle frame, and additional data required
may change depending on the model. The fluid specific gravity and kinematic viscocity default to
values for water. The total solution time should be entered by the user and should be a multiple of the

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


3.1.3
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

longest period of interest. A default of 5.0 seconds for the total solution time is entered automatically if
no value is entered by the user. Earthquake solution times are usually on the order of 40-to-50 seconds,
and periods on the order of 2-to-10 seconds are not unusual.

The middle data collection frame for the user-defined model type is shown below.

The Horizontal Length is the total horizontal length of the model grid. For a horizontal vessel this is
the total tip-to-tip length of the vessel. For a tank, this is the diameter of the tank. The Vertical Length
is the total vertical height of the model grid. For a horizontal vessel this is the diameter of the vessel.
For a tank, this is the height of the tank.

The Horizontal No. of Cells is the number of computation grid points in the horizontal direction.
Usually, the ratio of the horizontal-to-vertical cells will be equal to the ratio of the horizontal-to-
vertical lengths. The user should start with a small grid and increase the grid density as the work and
need progresses. The largest available grid size is 400 –by- 400, but sizes in the range of 40-to-50
maximum in each direction are recommended. Initial grid sizes in the 20’s are typical.

Once the basic model data is entered for user defined geometries the “User Layout” may be
specified. The User Layout screen is shown below:

The “Geometry Button” should be selected next.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


3.1.4
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The dimensions for the listed model types are taken from the main form horizontal and vertical
dimensions. For example, if an 8m.vertical dimension and a 2m. horizontal dimension were entered on
the main screen, and a “Vertical Vessel with Elliptical Heads” was selected, the following textural
layout screen would be displayed automatically when the “Build It Now” button was pressed.

The text layout method is provided so that the user can edit the model as necessary to define any
particular problem of interest. An example user-modified geometry is shown below.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


3.1.5
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

In this example the user might be interested in the time it took to fill the overflow line, and the loads
generated on the spillover. (Note that “V”s and “P”s are inserted into this model by the user. These
will be discussed in greater detail later, but the entry of a “V” is for a specified inflow or outflow
boundary velocity. The “P” is for a pressure sensor point. The pressure in the model will be recorded
at the points where “P”s are placed and reported graphically.)

For many systems either a periodic or seismic excitation will be specified. Periodic and seismic
excitations are input in the bottom of the main form. This portion of the main screen is reproduced
below:

Periodic and seismic time history loads may be specified simultaneously if desired, but they are
typically mutually exclusive. For periodic side-to-side accelerations, the Period and Maximum
Periodic Acceleration would be entered. For a transient seismic loading, any combination of
horizontal and/or vertical Elcentro seismic transient load components may be entered. For the example
above, the Elcentro earthquake will be applied fully in the horizontal direction and 60% in the vertical
direction. Thirty-five seconds of simulation time will be used. The main form input for this example is
shown below:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


3.1.6
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Once this input is completed, the “RUN” button can be depressed. A screen similar to the one shown
below will appear animating the results of the simulation.

The base shear and transient pressures are shown in the plot in the bottom left hand corner of the
screen during the simulation. The CFD grid is shown scaled in the right side of the screen. In this grid
the liquid free surface is drawn as a red line, and the velocities in each of the fluid data cells are drawn
as blue lines during each rendering time step. (The magnitude of the velocity scaling is controlled on
the Advanced data form.) The user may escape from the simulation at any time by hitting the
<ESCAPE> key. If the simulation runs to the total time allotted, an FFT of the pressure will be drawn
in the upper left of the screen as shown below:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


3.1.7
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

When the simulation reaches the total solution time or the user <ESCAPES> the transient iterations
will be terminated and the user will be given the opportunity to step back through the solution using
the movie option. The movie replay mode looks very similar to the solution mode except the user has
control of the progress of results. Hitting the key “1” advances the solution and hitting the key “2”
backs-up the solution. <ESCAPE> can always be used to exit.

In movie replay mode, the instructions for replay appear in the middle of the upper left portion of the
screen and they are:

Hit the key “1” to move the movie in the forward direction:

Hit the key “2” to move the movie in the backward direction.

Using the keys “1” or “2” the user can step forward or backward through the entire transient simulation
just performed. The movie files are also stored on the hard disk in the folder where the simulation was
performed. The “movie” button on the front of the main form can be pressed to replay the latest movie,
or the movie for the model filename selected.

On final exit, the user is returned to the main form with plots and reports available. Additional buttons
appear in the upper right and bottom portions of the main screen.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


3.1.8
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Buttons for the available text and graphics reports are highlighted in the top and bottom of the right
hand side of the screen. The main plot area is in the upper right hand side of the screen. Spectrum,
Max Pressure, Average Pressure, Shear Loads or Overturning Moment plots can be generated in this
area. The Spectrum plot for the example model is shown below:

The Zoom button in the bottom of the screen allows the user to zoom on a particular range of the plot.
For example, in the above plot it appears that there are two modes of interest between 0 and 4 Hz. A
“Zoomed” plot of this area appears below:

Other plot types may be generated and “Zoomed” as necessary to extract desired information. A Rich
Text report is available that can incorporate plotted and text results. This report is accessed by clicking
on the “Report” button in the upper right of the main form. (If errors occurred during the run then
messages reporting these errors will be in this report.) To send plots immediately to the report
window, the user should open the report window by clicking on the Report button. When the report
window is shown, the user can send any generated plot to the Rich Text report window by clicking on
© Copyright 2003, Paulin Research Group & Dynaflow Engineering
3.1.9
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

either of the buttons on the bottom of the main form: . In this way the user can add
graphics and text to the report as needed. Once the reports contain the necessary information they can
be sent to the printer. Reports tailored in this fashion are not saved with the input data files.

Pressure Sensor Points

Slug and Tank Sloshing models have the maximum pressure points selected automatically. The tank
sloshing model puts a single pressure sensor on the bottom left of the tank. The “Slug Elbow Impact”
model puts pressure sensors along the elbow impact area through a distance along the radius equal to
the diameter of the pipe as shown in the figure below. The user may place pressure sensors in any user
defined geometry, but if none are found, SPLASH finds the first point in the model starting from the
bottom up and from and from left to right that contains fluid and puts a single pressure sensor there.

Five (5) different plot types that can be generated.

Spectrum – Generates a fourier transform of the time history of the average pressure plot. The
maximum frequency plotted in the spectrum transform is taken from the advanced data form. The
average pressure is the sum of the pressure in each of the pressure sensor cells divided by the total
number of cells. The location of pressure sensor points is defined in the section above. If there is only a
single pressure sensor, then the average pressure and the maximum pressure plots will be identical.

Max Pressure – The maximum pressure recorded from the defined pressure sensor cells. If there is
only a single pressure sensor cells then this is the pressure in that cell.

Average Pressure – The sum of the pressure from each of the pressure sensor cells divided by the
total number of cells.

Shear – The sum of the pressure on vertical faces on the model times the effective area of the model
exposed to the pressure. The effective width is the diameter of circular geometries and the entered
width of rectangular geometries. The width of rectangular geometries is entered on the advanced data
form. The plotted shear load can be considered the total load that must be applied from an external
source to withstand the collective fluid load on the model at any point in time. This load calculation
assumes that each obstacle in the model extends for the full width of the model.

Moment – The sum of the pressure on each vertical face in the model times the effective width of the
model times the height from the bottom of the model. The effective width is the diameter of circular
geometries and the entered width of rectangular geometries. This can be considered to be the total
overturning moment about the base of the geometry that an external source must be able to withstand
at any point in time.
© Copyright 2003, Paulin Research Group & Dynaflow Engineering
3.1.10
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Controlling the SOLA-VOF Solution:


The time stepping algorithm of the SOLA-VOF solution can encounter free surface configurations that
produce erratic pressures. The algorithm has several mechanisms to counter the errant developing
pressures during iteration, but it can not be known a priori which approach will work best for a given
problem and time stepping scheme. The user must be aware of the type of irregularities that can occur
and know how to combat them.

The default settings for convergence tolerances on the Advanced Form are set at levels intended to
provide reasonable solutions at the cost of some accuracy. The user is encouraged to review this
section and the example problems to learn how to best determine when improved accuracy is required.
In general, solutions having a high degree of accuracy are more difficult to produce and require more
time in the numerical iteration process.

In general, a pressure convergence error will manifest itself in one of two ways as illustrated in the
figures below:

The figure on the left shows a solution that has detected an uncorrectable irregularity and has
terminated. This is indicated by the “Unstable Solution” message printed in the middle of the screen.
The figure on the right shows a solution whose pressure magnitudes have grown numerically without
limit. Solutions like the one on the right can also occur if the velocity visualization scalar is set too
high, and in this case the solution may be satisfactory, and only the visualization needs adjusting.
Reducing the magnitude of the vector visualization scalar a few orders of magnitude will always
correct this problem. The user can typically tell that the solution has begun to encounter problems
when one or more of the following occurs:

1) The time stepping visibly slows down


2) The time increment grows smaller and smaller.
3) The number of iterations per solution hits the maximum value allowed.

The time stepping is something the user will notice by watching the screen as the calculation
progresses. The time increment and the iterations per increment are shown in the upper left of the
screen while the solution is processing. These values are also summarized in the text report available
when the solution completes.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


3.1.11
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The SOLA-VOF algorithm is controlled using the Advanced Data Form. This form is shown below
and arrows are drawn to those data cells that can be used to improve the convergence performance.

When the user begins to have convergence problems the following approach is recommended:

1) Terminate the existing solution by hitting the <ESCAPE> key.


2) Set the “Velocity Display Multiplier” to 0.0001 or smaller. This is to be sure that any errant
plotted velocities or pressures are not associated with graphical scaling.
3) Change the Cell Flow Imbalance to 0.4.
4) Change the Full Cell Tolerance to 0.001
5) Change the Maximum Allowed Pressure Iterations to 20. Reducing the maximum allowed
pressure iterations can have a very favorable effect on convergence and a reduction of
“noise” when the free surface motion is particularly violent. The user must only make sure
by both inspection and by sensitivity investigation that the resulting “improved” solution is
sufficiently accurate for design purposes.
6) Reduce the Minimium Allowed Time Step by one or two orders of magnitude.
7) Review the report “Solution Attributes” if available to see what part of the solution is failing.
6) Restart the solution.

The Cell Flow Imbalance determines the allowed error in the continuity expressions before a pressure
adjustment is required. Relatively large flow imbalances can typically be tolerated. If the value is set
too high, the user will notice that the volume of liquid in the solution slowly reduces to zero. Large
flow imbalances (Cell Flow Imbalance Tolerance = 0.4) will generally alter the solution slightly. Loads
and predicted frequencies will typically be smaller than the actual values. Typical “loose” solutions are
1.5 –to- 3.0 times lower than more accurate ones.

More accurate solutions unfortunately are more difficult to obtain. Reducing the Cell Flow Imbalance
will definitely produce more accurate results, but the solution will also become more sensitive to the
runaway pressure problem.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


3.1.12
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Increasing the Cell Flow Imbalance to large values like 0.4 make the solution more convergent and
allow it to run faster, but the resulting pressures, shear loads and frequencies are typically 1.5 –to- 3.0
times too low.

When the Cell Flow Imbalance is set to smaller values, like 0.001, the solution can be improved by
increasing the maximum allowed pressure iterations, and by fixing the initial time step at a small value
and disabling the program control. When the pressure iteration converges and the solution time step is
stable the user has the best solution possible within the tolerance set. If the iteration begins to show
diverging pressures than the pressure iteration count can be set to a value on the order of 50. When
this happens SPLASH will detect that the pressure iteration is failing and will smooth pressure and
velocity solutions based on a heuristic algorithm whose intention is to maintain the total system
energy. The higher the system velocities, the more unstable the solution. The lower the number of
pressure iterations the more stable the solution, but the less accurate.

The full cell tolerance should not get much larger than 1.0E-2. and values smaller than 1.0E-7 can
begin to slow the solution too much. This value is used to determine when a cell is full of liquid and
when it is empty. Cells that have a percentage less than this tolerance will eventually be assumed
empty, and cells with a percentage larger than 1.0-(tolerance) will be assumed full. Values that are too
large can result in excess amounts of fluid being added or taken away during the iteration process.

Users are encouraged to approach non-convergence problems in the following manner:

1) Try to find a solution using the approach outlined above.


2) If this works, then begin adjusting the tolerances in a direction to improve the solution accuracy
and see if the calculated value of interest, (i.e. maximum pressure, shear load or moment),
changes. If it does not change, then most likely the original solution is a good one. If the
major system variable changes, then further convergence studies are in order.
3) As a last effort – set the initial time step to a small value that can be tolerated given the problem
size and the needed duration. This is usually in the 0.0005 –to- 0.00001 sec. range. Disable the
program control of the time step, and set the number of allowed pressure iterations to 3000.
4) Changing the grid size can also affect convergence. Typically smaller grids converge better. Be
careful with grid sizes larger than 100X100 cells. Grids of this size are seldom needed.

The errant pressure problem can also manifest itself in “noise” in the pressure/load time history plots.
The user should watch the time history plot as it progresses along the left side of the screen during
solution and when “random” responses superimposed over what is an otherwise smooth solution
appear the user should be aware that the associated pressure and load values may be in error by the
magnitude of the noise observed. The plot on the left below shows a uniform shear and pressure load
plot. The plot in the middle shows a shear and pressure load plot that is suffering from pressure noise
on the right side of the plot. When this happens the user will see artificial spikes in the pressure and
shear load plots that cause the automatic plot scaler to reduce the size of the normal fluctuations to
accommodate the errant variations. This differs from the noise associated with a typical earthquake
response that is shown in the plot on the right below. A good earthquake response plot ends after the
earthquake’s duration with a usual sinusoidal response as the liquid returns to a non-excited state.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


3.1.13
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Uniform Shear & Pressure Plot Runaway Pressure Noise Seismic Shear & Pressure Plot

Intermediate Void Formation due to High Cell Flow Imbalance Tolerance

Adjustments to the cell flow imbalance may speed solutions and result in good calculations, but large
values of the cell flow tolerance can result in the appearance of internal fluid voids such as those
shown in the plot above. These voids are not necessarily associated with runaway pressures, and may
not disturb the solution. The voids can almost always be eliminated by reducing the “Cell Flow
Imbalance Tolerance”, albeit at the expense of an increased run time and possibly by the addition of
other instabilities.

The text report available on output via the Report button provides a summary of the nonlinear
convergence history that can be used to determine the stability of the solution. Sample attribute report
text is shown below:

Solution Attributes:

Max Delta T used (sec) = 0.1250E-02


Min Delta T used (sec) = 0.4258E-03

Total Solution Time Steps = 695


Number of Steps Requiring Pressure Smoothing = 10
Number of Times the Velocity Cell Size was Exceeded = 0
Maximum Iteration Used = 1000
Number of Time Steps where Iteration Limit Exceeded = 78
Number of Times Steps Reset and Shortened = 0

The minimum delta T used is smallest delta T that was required to keep the solution stable. The
maximum value used will be the largest value selected by SPLASH to speed the solution at a rate that
still produces stable solutions. In the above report, pressure smoothing was required for 10 iterations.
This means that runaway pressures were detected during the iteration process and aggressive measures
were needed to return the solution to a physically meaningful state. The user would certainly want to
further inspect the sensitivity of this solution to the convergence parameters. The velocity cell size
was never reduced due to high velocities, but the pressure iteration was exceeded 78 times. The user
can increase the allowed maximum pressure iterations to attempt to correct this problem or fix the
startup time step and disable the program control of the time step. Disabling the program control of
the time step means that the initial value will be used unless a smaller value is needed. Otherwise,

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


3.1.14
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

SPLASH may inadvertantly begin to increase the time step and potentially introduce an unstable
pressure iteration. Ideally, the user would like to see a value printed for the Maximum Iteration Used
that is less than the maximum allowed. This means that the pressure iteration converged without
difficulty and corrective measures were not employed.

Other parameters on the Advanced Data Form are Described below:

“Show Real Time Pressure Instead of Velocity” – This checkbox is used to generate pressure real
time plots instead of velocity plots. Velocity plots are generated by default for each time step. The
velocity of the center of each cell is scaled with respect to each other cell, multiplied by the velocity
scale multiplier value and then drawn. The dynamic changing shows the direction of fluid movement.
The plot below on the left shows a velocity plot during an earthquake solution as the fluid on the right
moves down and toward the center, while the fluid on the left moves in an opposite direction, up and
away from the center. The plot below on the right shows a pressure plot as the vessel begins to
accelerate due to the earthquake toward the right.

“Highest Frequency of Interest (Hz.)” – This value along with “Samples” is used to extract
appropriate data from the time history solution. These values are used to find the fixed solution time
steps that should be extracted from the SOLA-VOF data. Time steps for a frequency transformation
must be evenly spaced. The SOLA-VOF data is produced for varying time steps and so the varying
time step data must be interpolated to find values for the frequency transform.

“Maximum Frequency Plotted (Hz.)” – This is the maximum frequency that is generated for
SPLASH spectrum plots. Typically the user is interested in low frequency response.

“Grid Multiplier” – This value is used with autogenerated templates, (i.e. the Bend-Slugging model
and the Tank Sloshing model.) If the user wants more or fewer data cells in either of these models the
Grid Multiplier should be used. The Grid Multiplier is not used with user defined models since the user
has to specify the grid size explicitly.

“Velocity Display Multiplier” – This value multiplies the velocity or pressure symbols drawn in the
fluid during each iteration. The user should start a solution and then if the movement or pressure in the
fluid cannot be seen during the iterations, then the Velocity multiplier should be increased. If the
velocity or pressure vectors drawn are too large then the multiplier should be reduced.

“Movie Frames per Time Step” – Move files are stored during each solution. The default is to store
one frame per each solution time step. If particular movie files are getting too large, or take too long to
view the user can always elect to store fewer frames per time step. Ten is a common election for long
© Copyright 2003, Paulin Research Group & Dynaflow Engineering
3.1.15
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

running problems that require very small time steps. The movie files created by SPLASH are
<name>.mvi and <name>.mvj files.

“Check to Enter Model Depth and Depth” – The effective depth of the model perpendicular to the
2d section may be entered here if the default used by SPLASH is not satisfactory. Defaults are printed
in the text report. This distance will be used to adjust the magnitude of the pressure load so that it acts
over the correct vessel side area. For rectangular vessels the depth is the actual depth of the tank or
vessel. For cylindrical vessels the magnitude should be 0.785 times the diameter of the vessel. The
0.785 multiplier is automatically applied for all circular geometries constructed by the user.

“Plot Display Type” – 2DXY plots should be used in all cases.

“Send Plots to Clipboard” – Once spectrum or load plots are generated on the main form they may be
included in the rich text report prepared by SPLASH or they may be copied to the system clipboard.
The default is to include the plots in the SPLASH report, but the user can check this box and optionally
have the plots sent to the system clipboard so that they can be inserted into another document. The user
will notice that the buttons in the bottom right of the main form change when this checkbox has been
selected. If the plots from the clipboard do not show up in the document as desired, then the plot can be
pasted into the SPLASH report and then copied from the SPLASH report window using the key
sequence <cntrl><c> to copy the plot from the REPORT window onto the system clipboard in a
graphics mode.

“Disable Run Time Base Shear Plotting” – Used to turn off the interactive base shear plotting that
appears during the interactive solution display. Turning off the plot does not affect the users ability to
generate shear plots interactively once the solution has completed.

“Deceleration Loading” – This checkbox is used to change the seismic and periodic load options so
that only the maximum “g” load is used during the solution, filtered by a ramp-and-hold waveform.
This option is discussed in more detail later.

“Plan View – Disable Gravity in the Y Direction” – This checkbox is provided so that users can run
plan view simulations rather than elevation view simulations. In a plan view simulation there is no
gravity acting in the –Y direction. The example below illustrates. The simulation is “looking down” on
a header, with two cross-overs to a second header. The flow enters from the left, bypassing the first
cross-over because the velocity of the flow is only in the direction to the right. The flow continues to
the right until impacting the end cap on the far right of the header. This alters the velocity field in the
flow until the free fluid begins to flow through the first cross-over. When looking down at a model like
this, when gravity acts into the page, the “Plan View” box should be checked.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


3.1.16
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

File Handling & Units


The input for SPLASH is in metric units and uses Kilograms, Meters and Seconds. Density is given in
terms of specific gravity. The only fluid property that is unit dependant is the kinematic viscosity,
which is given in m2/sec. The kinematic viscosity of water is 0.000001 m2/sec. at standard temperature.
This value is input automatically. The kinematic viscosity is the absolute, or dynamic viscosity divided
by the density of the fluid. Output pressures are in Pascals (N/m2) unless otherwise noted.
Occasionally values in bars and psi are given for comparison, but in these cases the units used are
noted.

The model input is stored in a file with the extension 2DC, i.e. “tank.2dc.” This file contains all
advanced data form settings and any user defined model work sheets. From this file the user should be
able to recreate any output results by rerunning the job. 2DC files are designed to be upwardly
compatible so that files created in earlier versions can be read in the latest version of SPLASH. Movie
results are stored in the two files <name>.MVI and <name>.MVJ. The MVI file contains movie
indices, and the MVJ file contains data to reproduce each movie frame. The sizes of the movie files
and the number of frames per movie are controlled on the Advanced Data Form by setting the “Movie
Frames per Time Step.”

After a job finishes there are a number of intermediate files remaining. The three that contain
information that may be of interest are described below: These files are generic, and are overwritten
each run.

2dout.txt – Generic transfer file that contains the text output form the last job run in the current
subdirectory.
Fort.81 – Contains the times for each frame in the movie file.
Fort.90 – Solution time step information

Periodic and Seismic Loadings


Any of the models described herein can be exposed to combinations of periodic and seismic horizontal
and vertical loadings. Both periodic and seismic loads can be applied simultaneously, but this is not a
usual design case. The input data cells for periodic and seismic loading are shown below:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


3.1.17
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The top two data cells prescribe a horizontal periodic loading that might be used to simulate the pitch
and roll of an ocean-going vessel. The acceleration loads on the fluid are found from:

Force = (Mass of Fluid) (g) (SINE [(2)(pi)(f)(T)]

Where (g) is the maximum periodic acceleration entered by the user,


(f) is the frequency, (one over the period entered, i.e. 1/P)
(T) is the current time in the solution.

The bottom four data cells describe an earthquake acceleration of the fluid. The base earthquake is
taken from the El Centro 1940 North South Component (Peknold Version). There are 1559 time
history data points in this set, spaced at 0.02 seconds. The total duration of the event is 31.2 seconds,
and the maximum acceleration in the data set is 0.31882 g’s.

Users can specify either horizontal or vertical time and/or acceleration scale factors. The acceleration
scale factor is used to multiply the “g” value taken from the El Centro data set. For example, if the
acceleration scale factor was 2.0, then the maximum acceleration would be (0.31882)(2) = (0.63764).
The time scaling is used as a divisor into the actual time to get a table extraction time. For example, if
the time scaling factor was 2.0, then the total duration of the earthquake would be extended to 62.4
seconds. User’s are urged to use time scaling carefully as this can be used to change the frequency
content of the excitation.

The maximum periodic acceleration or the maximum seismic “g” load can be applied to the fluid in a
“ramped” manner to simulate the sudden deceleration of a tanker car or truck, or to check the shear
load and moment calculations on tanks and structures due to steady state fluid loads. The acceleration
load history takes the following form:

The ramp time, duration, and the checkbox to indicate that these values should be used with the
periodic or seismic maximum accelerations is found on the Advanced Data form.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


3.1.18
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

If the “Apply Max Seismic or Periodic g load” box is checked, the maximum g load entered for the
periodic excitation is added to the maximum g load from the seismic entry and applied to the fluid
using the time ramp shown above. To apply a constant g load instantly, the box above would be
checked, the ramp time would be 0.0, and the duration time would be set to be greater than the total
solution time. To simulate a tanker truck that breaks to a complete stop in three seconds, the ramp
time would be 3 seconds, and the duration would be 0.0 seconds. In this case the total solution time
should be at least equal to 3 seconds plus two times the lowest sloshing period of interest, and perhaps
longer.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


3.1.19
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Basic Theory
The SOLA-VOF algorithm finds solutions for the transient motion of an incompressible viscous fluid
in two spatial dimensions with any number of free surfaces. The grids used in SPLASH are comprised
of uniform cell sizes, but may be longer in one direction than in another.

The solution is obtained by applying the finite difference technique with pressure and velocity as
variables in an Eulerian grid. Irregular obstructions are simulated by stair-stepping obstacles in the
rectangular gridwork. The transient motion of interest is found as a series of short steps in time, or
computational cycles. Each cycle is broken down into three phases.

1) Approximate values for the horizontal and vertical velocity components are calculated by
applying pressure and viscous forces, by allowing fluid to flow from neighboring fluid regions,
and by applying acceleration or periodic loadings.
2) Mass continuity is enforced by solving a Poisson equation for the pressure change at each point
in the fluid and updating the velocities in response to the change in pressure. This second step
is the pressure cycle described above that can often diverge. The “Cell Flow Imbalance”
determines the accuracy of the mass continuity resolution, and the “Full Cell Tolerance”
determines when the iteration believes that a cell is completely full or completely void of
liquid. The “Maximum Allowed Pressure Iterations” is the number of iteration cycles that will
be applied until the specified “Cell Flow Imbalance Tolerance” is satisfied. Once the number of
“Maximum Allowed Pressure Iterations” is exceeded a local pressure-velocity smoothing
algorithm is employed for cells that are experiencing runaway pressure problems.
3) The Volume of Fluid (VOF) algorithm is applied after the mass continuity iteration to move the
fluid at the surface and allow for the propagation of free surfaces and to provide for interaction
with solid boundaries. Boundary conditions are imposed at appropriate times during each step
of the iteration.

The basic equations satisfied are:

dU/dx + dW/dy = 0 … continuity

Newton’s 2nd Law of Motion

dU/dt = -UdU/dx – WdU/dy – dp/dx /r + gx + v(d2U/dx2 + d2U/dy2)


dW/dt = -UdW/dx – WdW/dy – dp/dy /r + gz + v(d2W/dx2 + d2Wdy2)

x and y are the horizontal and vertical coordinates. U and W are the horizontal and vertical components
of the fluid velocity on a per cell basis. (p) is the pressure and (r) is the fluid density. (v) is the
kinematic viscosity and gx, gy are the gravitational acceleration components in the horizontal and
vertical directions respectively.

At the end of each cycle, the time step is adjusted in response to the changes in velocity and surface
location. The maximum time step that can be used for numerical stability is the smallest minimum time
required for information to convect or diffuse across any cell.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


3.1.20
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

In addition, if the pressure solution is converging in more than 200 iterations, the time step will be
reduced by the factor 0.99. If the pressure solution is converging in less than 200 iterations the time
step will be increased by the factor 1.01. This time step adjustment based on pressure iterations can
result in an increasing or decreasing of the time step during solution unless the user disables the
program control of the time step on the Advanced Data Form.

Even if the user disables program control of the time step, the time step may still be reduced by the
program when the calculated velocities exceed a value that allows fluid to pass through more than one
cell in a single time step. When the solution velocity in a single cell exceeds this limiting value, the
pressure iteration is repeated with a time step equal to half of the previous time step. The time step
will be continually reduced until the velocity limit is reached. The velocity limit is:

(cell size / time step)

An indication of an uncontrolled runaway pressure or an otherwise divergent solution can be detected


when the time step gets too small because of the adjustments described above. When the time step is
adjusted below the critical minimum value specified on the Advanced form the program stops at the
current time with a report of an unstable message both on the screen and in the text report.

The user can watch the changing time step during the solution by watching the values scrolling in the
upper left of the time history model screen. An example of this part of the screen is shown above. The
value reported for DELTA is the time step. This value will continually get smaller as the solution
looks harder and harder for a solution in a divergent situation. One can imagine the case where a
solution goes unstable due to a runaway pressure. If the pressure gets higher in an artificial manner the
pressure iteration increases velocities to accommodate the pressure. As the velocities increase the time
step required to maintain a stable solution get smaller since the cell size remains constant. At a point,
the time step is too small to permit the solution to continue. There are instances where the algorithm
may reduce the time step to get past a particularly difficult part of the solution associated with a high
velocity impact, only to increase it later once the velocities and accelerations are reduced.

Slug Elbow Impact


The automatic slug modeler is setup to let users investigate the effect of a alug of liquid impacting on a
90 degree elbow of any radius. The basic model geometry and inputs are shown below.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


3.1.21
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Slug Elbow Impact Basic Data Doubled Grid Size Single Grid Size

The grid density can modified on the Advanced data form. A standard density grid model and double
density grid are shown above. The slug wavefront in general will not be known and the user is
encouraged to experiment with different entry lengths varying from 0.1D to 2.5D to investigate the
effect of slug waveform on the impact values. A more conservative result should come from the
shorter entry length The original slug waveform is shown in the upper right figure and is a function of
the inlet length. The longer the inlet length the more bottom slope to the initial waveform. For the
example problem shown the pipe diameter is 0.5m. and the slug velocity is 10.0 m./sec. The final
model result screen is shown below:

The pressure jumps upon initially encountering the elbow and then increases almost linearly, leveling
out once the flow pattern has been established. The shear plot does not suffer as much as the pressure
plot from any local cell related pressure spikes since the shear plot represents the total shear on the
model, while the maximum pressure plot reflects a jump in pressure at only a single cell. Shear load
and overturning moment results tend to be less sensitive to local pressure spikes, and therefore will
converge to the correct total load quicker than the maximum system pressure value.

The momentum change force at the elbow can be calculated from rAV2, where r is the fluid density, A
is the flow area and V2 is the velocity squared. For the pipe modeled in this problem the calculation is:

rAV2 = (1000)kg/m3 (0.1963)m2 (10)2 m2/sec2 = 19,630 N.

The 2d model does not exactly replicate the cross sectional area of a round pipe however. The area
represented by the 2d model will be (0.5)2 = 0.25 m2, so the net load between the models will be

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


3.1.22
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

approximately different by the ratio of the areas, i.e. the 2d SPLASH model will produce a larger load
by about the ratio:
0.25 / 0.1963 = 1.2735.

The momentum change force in the 2d model will be:

(19,630) (1.2735) = 25,000 N.

The total load on the elbow in the horizontal direction is the sum of the momentum load just calculated
and the pressure load PA.

In a typical pipe stress slugging application it is assumed that the slug load develops within a time of
aR/V, where (a) is the bend angle, R is the radius, and V is the slug velocity. For this example, the
load development time would be:

aR/V = (3.14159/2) (0.75) m. / 10 (m/sec) = 0.12 seconds.

An intermediate plot of the flow result is shown below. The slug entry into the elbow looks uniform
and reasonable and reasonable both in the 2d grid and in the time history:

The initial pressure spike is artificial since the flow is being redirected at its first contact with a step
obstruction. The shear load plot is less effected and appears below:

The time to develop the loading front can be seen to be almost exactly the calculated value of 0.12
seconds. The unbalanced shear load is approximately 80,000 N. This is 80,000 – 25,000 = 55,000 N
greater than the momentum load alone. The average pressure plot on the backside of the elbow from
SPLASH is:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


3.1.23
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The pressure load is approximately 160,000 Pa. (N/m2). This will represent an unbalanced load of
approximately: PA = (160,000)(0.25) = 40,000 N. The sum of the momentum and pressure load is:

40,000 + 25,000 = 65,000 N.

To check the accuracy of the results, the tolerance parameters used to produce the solution must be
inspected. For the model shown above, the default tolerance values are:

Cell Flow Imbalance = 0.4


Full Cell Tolerance = 1.0E-3
Maximum Allowed Pressure Iterations = 50

In an effort to get a more accurate solution the “Cell Flow Imbalance” is decreased to 0.04 and the
maximum number of iterations is increased from 50 to 1000. The following, plot resulted:

The solution attributes for the above plot are shown below. There was no pressure smoothing required.
Solution Attributes:

Max Delta T used (sec) = 0.5000E-02


Min Delta T used (sec) = 0.4094E-03

Total Solution Time Steps = 661


Number of Steps Requiring Pressure Smoothing = 0
Number of Times the Velocity Cell Size was Exceeded = 0
Maximum Iteration Used = 1000
Number of Time Steps where Iteration Limit Exceeded = 5
Number of Times Steps Reset and Shortened = 0

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


3.1.24
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The solution attributes reported for a model run longer is shown below. Note how the number of time
steps requiring smoothing is equal to 10 – indicating that the pressure values were not too accurate, i.e.
70K to 80K in terms of final shear loads:

Solution Attributes:

Max Delta T used (sec) = 0.1250E-02


Min Delta T used (sec) = 0.4258E-03

Total Solution Time Steps = 695


Number of Steps Requiring Pressure Smoothing = 10
Number of Times the Velocity Cell Size was Exceeded = 0
Maximum Iteration Used = 1000
Number of Time Steps where Iteration Limit Exceeded = 78
Number of Times Steps Reset and Shortened = 0

The user should always check for erradic behavior in the plot, and adjust tolerances, or time steps as
described herein to refine or improve the model.

Tank Sloshing
The Tank Sloshing model is intended to quickly allow users to setup tank seismic or periodic
excitation solutions. Additionally, the acoustic natural frequencies of the tank at different fluid levels
can be calculated.

Using the geometry of the emergency water tank described in reference 2.

Diameter = 8.6 m.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


3.1.25
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Liquid Height = 10.8 m.


Fundamental Period = 3.066 sec.

From Reference 2, the first sloshing mode should be approximately (1/3.066) = 0.326 Hz. The default
model for this tank and its input is shown below: The fluid properties are not entered. Values default to
water. Only the four numbers shown in the figure below need to be entered to start the analysis. The
default value for the kinematic viscosity of water is 0.000001 m2/sec.

Tank Sloshing Input 2d CFD Model

When the “RUN” button is hit, the top fields are filled in and the following run is made:

The default solution time is 5.0 seconds. Since the fundamental period is 3.066 seconds, the total time
should be some multiple of this period so that the Fourier routine can effectively transform the time
waveform into the frequency domain. This small problem runs quickly, so longer run times will not be
a burden. The longer the run time the greater the frequency resolution in the low ranges. Frequency
transform data points are distributed over a frequency range limited by the frequency N/Tmax, where
N is the total number of Fourier sampling points and Tmax is the total solution time. The total number
of Fourier sampling points can be found from:

N = Tf/n

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


3.1.26
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Where (f) is the frequency of interest, and (n) is the number of sampling points in the frequency of
interest. Both (f) and (n) are given on the SPLASH Advanced data form.

To trap the lowest frequency at 0.326 Hz. the Fourier analysis must be focused in this range.

The maximum plotted frequency will be set to 1 Hz. This will limit the plot size and make it easier to
see the low frequency range. The maximum frequency of interest will be set to (0.326) Hz. The
number of samples in the frequency of interest will be set to 5.

Since the period to estimate is 3.066 seconds, at least five of the wave reflections should be included in
the calculation. The total solution time should be at least equal to (3.066)(5) = 15.33 seconds.

For five samples in 3.066 seconds, the largest time increment would be 0.61 seconds. Since most
SPLASH solutions operate with time steps on the order of milliseconds, the time step will not be a
concern for this problem. To calculate an accurate low frequency mode, the pressure and volumes
of fluid should be reasonably accurate. The cell flow imbalance will be set to 0.01 for this
calculation.

The advanced data form with the input suggested is shown below:

The resulting spectrum plot is shown below.

The plot is coarse. To increase the resolution the solution time should be increased. A plot with the
solution time increased to (3.066)(20) = 61.32 sec. is shown below:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


3.1.27
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

API 650 (ref. 3) also gives an equation for the fundamental period of large storage tanks in Appendix E

T = k(D0.5)

Where D is the tank diameter in feet, and “k” is a factor from Figure E-4 which can be approximated
by:
k = 0.6 for D/H = 2.2 or less
k = (0.6) + (0.045)[(D/H) - 2.0] for D/H greater than 2.2 and less than 8.0

Reference (4) also gives the following equations for the sloshing modes in tanks:

w = 1/[(2)(3.14)] [ L(g/R) tanh( L H/R ) ]0.5

Where L is the 1.841 for mode 1, 5.331 for mode 2 and 8.536 for mode 3.
g is the acceleration due to gravity.

Using these equations for the tank system analyzed above:

From API:
D = 8.6m (28.215 ft.)
D/H = 8.6/10.8 = 0.7963

k = 0.6 for D/H less than 2.2


T = (0.6)(28.215 0.5) = 3.18 sec.

From Ref. 4
L1 = 1.841
g/R = 32.2 / 14.11 = 2.2825
H/R = 10.8 / 4.3 = 2.512

w1 = 1 / [(2)(3.14)] [ (1.841)(2.2825) tanh[ (1.841)(2.512) ] ]0.5


= 0.3262 Hz.

w2 = 1 / [(2)(3.14)][ (5.331)(2.2825) tanh[ (5.331)(2.512) ] ]0.5


= 0.555 Hz.

Values obtained are very close to those calculated by SPLASH.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


3.1.28
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

User Defined Models


The user defined modeling capability allows users to construct their own basic geometry slices starting
with preset models. A standard rich text editor is provided that allows the user to specify initial
velocities, obstructions, and pressure sensor points. The basic data for the user defined model type is
shown in the figure below:

The specific gravity and viscosity default to values for water. The solution time will default to 5.0
seconds, but the user should estimate the solution time needed and enter this value.

The horizontal and vertical lengths define the extent of the model. The user will be given a vertical
sliced area that has the horizontal and vertical lengths requested. The number of horizontal and
vertical cells correspond to the length measures entered. In general, the ratio of horizontal to vertical
cells should correspond to the ratio of horizontal to vertical lengths. This is not required however, and
long, thin cells can be created if needed. The fluid height within the region is used to describe the
initial fluid in a specified tank or vessel. To construct a 10 m. by 10 m. sphere filled half with water,
the input would appear:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


3.1.29
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Once the above items are entered, the “User Layout” button should be selected. This button will bring
up the user geometry layout panel:

Hitting the “Geometry” button opens the basic geometry form:

From this form the user can select the basic model type to be constructed. To continue the sphere
example, check the spherical geometry radio button and hit the “Build it Now” button. The following
geometry will be shown in the rich text window.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


3.1.30
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The number of horizontal and vertical cells is shown along the top, bottom, right and left side of the
geometry. Valid model data is contained between these marked cells. X’s are used to define the
obstacle of the sphere border within the rectangular model grid. V’s, P’s and F’s may also be entered
or manipulated by the user to include velocity initiation sources, pressure sensors or fluid filled cells
respectively. If a “V” is inserted anywhere in the model space, the user should also enter the
corresponding horizontal and vertical velocity associated with that constant velocity fluid source. The
boundary types can be adjusted using the list boxes on the top of the form. The model construction
characters (V,X,P,F) are not case sensitive. To illustrate the use of these constructs, a fluid source flow
with 3 m./sec of water will be defined to enter the vessel from the top. The pressure will be measured
at a single point on the bottom of the vessel in the center, and a hole will be opened on the right side of
the sphere and a nozzle inserted so that any excess water can pour out if needed. The changes made to
the above model to effect these conditions are shown below:

1) Continuation boundary set on the right side to permit liquid to leave the vessel. (Otherwise a no
slip boundary is specified by default, and this would prevent any water from leaving the model
through the nozzle on the right side of the model.)
2) Five “V”s are placed at the top of the vessel and the vertical velocity component is set to -3.0 in
the velocity text box to indicate that water will be constantly entering the vessel from the top at
3 meters per second.
3) X’s were added to the vessel to contour the nozzle insert length, and Xs were removed along
the side wall to create the nozzle opening in the vessel.
4) A single “P” pressure point was added to the bottom center of the vessel.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


3.1.31
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

When any modifications are satisfactory, the “Finished Here” button should be hit and the user will be
returned to the main SPLASH screen.

Typically to run an analysis the user will not know beforehand what tolerances or vector plot scaling
will be needed. These values must be adjusted after a few runs are made. The first run of the model
described above is shown below.

Just Prior to Impact Final Configuration

On first inspection, it is easy to see that the velocity vectors are too large. These will be reduced by one
order of magnitude. Additionally, the solution duration will be extended from 5 seconds to 10 seconds.
The vessel is almost entirely filled in 5.0 seconds. The desire here will be to see if any significant flow
development occurs after 5.0 seconds. If none is seen then the total solution time will be reduced back
to 5 or 6 seconds. The above changes resulted in the following model plots.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


3.1.32
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

With reduced velocity vector scaling the flow condition is much easier to visualize. The extended time
shows that at 6.25 secoonds the last of the void is evacuated from the vessel.

The 2d slice of the sphere is representing a uniform extension of the sphere geometry normal to the
plane shown. This is illustrated schematically below:

It is straightforward to compute the effective volume of the 2d model of the vessel as:

Volume = (3.14)R2 (R)(2) = (3.14159)(5)2 (5)(2) = 785 cu.m.

For a liquid flowing through the zone in the top with a velocity of 3 m./sec, the total flowrate would
be:

Flowrate = Q = (V)(A) = (3)(10)(5/20)(10) = 75 cu.m./sec.

If the volume is divided by the flowrate:

Time to Fill an Empty Vessel = 785 / 75 = 10.466 sec.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


3.1.33
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Since the vessel starts about half full, it can be seen that the 5.0 seconds to fill the remaining volume as
calculated is correct. The above relationship will be used later to compute the loads on slug breakers
and other obstructions.
All other controls and loads apply to user defined models. This includes tolerances, seismic and
periodic loadings.

The falling droplet problem can be easily simulated in a user defined model. Any size zone of fluid
dropped through 20 meters should come in contact with the base in (2y/g)0.5 = [(2)(20) / (9.81) ]0.5 =
2.019 sec. The model simulating this experiment with two distinctly different fluid sizes is shown
below:

Several of the time slices from this solution are shown below:

Two Fluid Masses at Rest Half-Way in 20 m. Free Fall At 2.0 sec.

Flow Obstructions

Loads on Obstructions in the Flow Path can also be simulated by initiating a flow in a rectangular
gridwork of cells that are fully filled with the obstruction in the middle. Several examples are shown
below. One objective in this type of analysis is typically to compare the resulting drag forces on
different shapes exposed to a flow condition. In the models below a rectangular shape is compared to a
tapered shape. A velocity field is established on the entire right or left side boundary and an
appropriate initial velocity specified. An initial transient period will occur while the velocity field
spreads across the fluid and exits from the right hand side that must be ignored. The drag forces can be
estimated once the transient setup period has passed. Users should be particularly careful with
tolerance effects in this type of model, and should run several simulations, reducing tolerances to be
© Copyright 2003, Paulin Research Group & Dynaflow Engineering
3.1.34
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

sure that a converged solution is obtained. When evaluating the accuracy of drag forces, the user
should note that both a pressure component and a moment component exist. For simplified analyses of
the drag on piping systems usually only the momentum component is evaluated. The momentum can
be calculated from:
rSAV2

where (r) – fluid density


(S) – Shape factor
(A) – Exposed Area
(V) – Mean flow velocity

The pressure can be estimated by making two runs. The first run places the pressure sensors along the
front face of the obstruction to get the high side pressure, and the second run places the pressure
sensors along the back face of the obstruction to get the low side pressure. The two pressure sensor
placement diagrams are shown below:

Sensors on Left Side of Obstruction Sensors on Right Side of Obstruction

Notes:
1) The vertical line of “v” down the left hand side of each diagram is used to describe the left-to-
right velocity field. The V’s are entered in a data cell where the velocity is known. The velocity
values in the horizontal and vertical directions are entered in text cells in the top of the User
Layout form. A single “v” prescribes the velocity in five cells. The vertical velocity is set in the
current cell and the cells immediately above and below the current cell. A horizontal velocity
is set in the current cell and the cells immediately to the left and right of the current cell. In this
way the acceleration of the fluid in the center of the zone is zero. Where a V is specified the
cells in its zone are taken to be full of liquid throughout the simulation.
2) The “P” characters indicating the pressure sensor locations are pointed to by arrows. Upper
case or lower case v’s, p’s, or x’s may be used. The P values are assumed to be filled with fluid
on startup of the transient but may be drained. Both an average and a maximum pressure value
are stored for the pressure sensor cells. The average pressure is used to produce the spectrum
analysis. Only a single P may be specified in which case the average and the maximum
pressure values will be the same. The pressure due to the head of fluid is included in the
pressure reported at each sensor node.
3) The right and left boundaries are set to be “continuation” boundaries for the obstruction
analysis. This is important so that fluid can enter one side and exit from the other.
© Copyright 2003, Paulin Research Group & Dynaflow Engineering
3.1.35
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

4) The user should consider running a “Plan View” model for these geometries. Use the checkbox
on the “Advanced” form.

Flow Patters around Tapered and Rectangular Obstructions

Seismic Test Model


It is typical to investigate the effect a seismic event has on the fluid in a large tank. A typical problem
input and model is shown below:

The total solution time was entered as 50 sec. This number was selected to be greater than the total El
Centro earthquake shaking duration of 31.2 seconds. After 31.2 seconds there will no further external
acceleration forces on the fluid and the free surface will continue moving only because of the prior

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


3.1.36
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

shaking of the earthquake and will eventually come to rest. The movement of the free surface after
31.2 seconds will take on a uniform and sinusoidal appearance. The Step Height is left 0.0 since there
should not be any initial step introduced into the fluid free surface since the loading will be provided
by the seismic excitation. Steps are only used when the sloshing natural frequency should be
calculated and when there are no other loads on the fluid, i.e. periodic or seismic. The final fluid state
is shown in the figure in the upper right. The shear forces and the frequency spectrum for this run are
shown below:

The solution attributes in the report for this run are shown below:
Max Recorded Pressure (Pa.) 69500.548
Avg Recorded Pressure (Pa.) 69500.548
No. of data points: 948.000

Max Shear Force (N.) 1145666.195


Max Overturning Moment (N.m.) 3091799.067

Frequency at Max Amplitude (Hz) 0.100

Solution Attributes:

Max Delta T used (sec) = 0.1347E+00


Min Delta T used (sec) = 0.5000E-02

Total Solution Time Steps = 947


Number of Steps Requiring Pressure Smoothing = 0
Number of Times the Velocity Cell Size was Exceeded = 0
Maximum Iteration Used = 42
Number of Time Steps where Iteration Limit Exceeded = 0
Number of Times Steps Reset and Shortened = 0

The solution is very stable for the seismic loads selected. Tolerances will now be checked to determine
the sensitivity of the solution to the present tolerances:

“Cell Flow Imbalance” Changed from default of 0.2 to 0.01.


“Total Solution Time” Changed from 50 sec. to 100 seconds to improve spectrum resolution.
“Max Allowed Pressure Iterations” changed from 100 to 1000.

Solution parameters from the run with the improved tolerances are given below:

Max Recorded Pressure (Pa.) 81122.154


Avg Recorded Pressure (Pa.) 81122.154
No. of data points: 1436.000

Max Shear Force (N.) 3054676.290


Max Overturning Moment (N.m.) 8727171.292

Frequency at Max Amplitude (Hz) 0.170

Solution Attributes:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


3.1.37
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Max Delta T used (sec) = 0.1638E+00


Min Delta T used (sec) = 0.5000E-02

Total Solution Time Steps = 1435


Number of Steps Requiring Pressure Smoothing = 0
Number of Times the Velocity Cell Size was Exceeded = 1
Maximum Iteration Used = 942
Number of Time Steps where Iteration Limit Exceeded = 0
Number of Times Steps Reset and Shortened = 1

The maximum overturning shear force increased from 1,145 kN, to 3,054 kN. The frequency at the
maximum amplitude often gives the lowest natural frequency of the fluid. In this case, going from the
approximate to the improved solution increased the natural frequency from 0.1 to 0.17 Hz.

This is a typical approach taken to determine the accuracy of a solution. Further reductions in
tolerances or increases in the run times result in little change to the solution.

The API equation for the fundamental period can be used to check the calculated frequencies:

T = k D0.5

D/H = 20 / 10 = 2.0
k = 0.6 when D/H <= 2.2.
D = 20 m. (1000/25.4/12) = 65.61 ft.

T = (0.6) (65.61)0.5 = 4.86 sec.

The frequency associated with this period is 1/T = 1 / 4.86 = 0.205 Hz.

The Reference (4) equation for the sloshing modes in tanks is:

w = 1/[(2)(3.14)] [ L(g/R) tanh( L H/R ) ]0.5

Where L is the 1.841 for mode 1, 5.331 for mode 2 and 8.536 for mode 3.
g is the acceleration due to gravity.

Using this equation for the tank system analyzed above:


R = (10 m.)(1000 /25.4/12) = 32.81 ft.

From Ref. 4
L1 = 1.841
g/R = 32.2 / 32.81 = 0.98146
H/R = 7 / 10 = 0.7

w1 = 1 / [(2)(3.14)] [ (1.841)(0.98146) tanh[ (1.841)(0.7) ] ]0.5


= 0.198 Hz.

w2 = 1 / [(2)(3.14)][ (5.331)(0.98146) tanh[ (5.331)(0.7) ] ]0.5


= 0.3636 Hz.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


3.1.38
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The maximum “g” load from the El Centro Earthquake can be checked by applying the maximum
seismic acceleration in a slowly ramped manner and then checking the horizontal load on the tank. For
the model above, the total 2d modeled fluid volume is (7)(20)(20) = 2,800 cu.m. The maximum
acceleration from the El Centro earthquake is 0.31882 g’s. Applied this “g” load statically to the fluid
in the horizontal direction should produce a maximum horizontal load of:

(1000)kg/cu.m (2,800)cu.m. (9.81) m/s. (0.31882) = 8,757,347 N.

The ramped constant load capability will be used to ramp the load over a time of 40 seconds. A 50
second solution will be used. This input on the advanced data screen appears:

The maximum shear load plot and final liquid confuration is shown below: The horizontal ramped
acceleration moves the fluid to the left side of the tank and closes the freeboard.

The solutions from SPLASH are exactly equal to the hand calculated value of the net shear load.

Sloshing Natural Frequencies of Horizontal Tanks & Designing Baffles


Baffle design involves first finding if baffles are needed, and then if they are, selecting their number
and making sure that they are designed to absorb the specified hydrodynamic loads.

There are two predominant natural frequencies of horizontal vessels:

1) longitudinal component
2) horizontal or side-to-side component.

The approach to this problem may be summarized as follows:

1) Build two models of the horizontal vessel, one that describes the longitudinal cross section and
the other that describes the circumferential cross section, (i.e. this section would look like a
sphere.)
2) For the longitudinal model be sure that the “check to enter model depth” checkbox is selected
in the advanced form and enter the internal diameter of the vessel in the depth text box.
3) For the circumferential model – also select the “check to enter model depth” checkbox but
enter the longitudinal dimension of the vessel in the “tank depth” text box.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


3.1.39
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

4) Expose the circumferential model to the design periodic pitch and roll shipboard loads. Make
sure that the solution has converged and then add the maximum shear loads printed by
SPLASH to the empty vessel periodic loads to design the saddle supports.
5) Expose the longitudinal model to the design periodic pitch and roll shipboard loads. Make sure
that the solution has converged and then add the maximum shear loads printed by SPLASH to
the empty vessel periodic loads to design the saddle supports.

The above approach will find loads due to the maximum periodic movement of the platform or vessel.
Unfortunately, these may not be the largest sloshing loads the vessel will be exposed to. Lower
frequency, (longer period) movement of the platform may correspond with a natural frequency of the
fluid and may produce larger design loads than those caused by a higher frequency, higher acceleration
load. Typically, (but not always,) the lowest fluid sloshing frequency will provide the most significant
sloshing response. To evaluate the design sloshing load due to an excitation at the sloshing natural
frequency take the following approach:

1) Estimate the first sloshing natural frequency using either a step function in the surface wave
or by exposing the model to a seismic event. Seismic events typically contain enough
frequency content to excite the lowest sloshing mode of typical piping or pressure vessel
components.
2) If the lowest sloshing natural frequency is higher than the design frequency then no further
analysis is required.
3) If the lowest sloshing natural frequency is lower than the design frequency then excite the
model using the period associated with the just found natural frequency and the design “g”
load. Lower “g” load values may be available for the reduced frequency.

Baffles may be used to increase the axial sloshing natural frequencies. Typically the effect of baffles is
evaluated by shortening the axial length of the vessel to the distance between baffles. The net shear
load predicted by SPLASH can be conservatively used to evaluate the loads on the baffles for weld and
plate thickness design.

References:
1) “In-Tank Fluid Sloshing Impact Effects During Earthquakds: A Preliminary Computational
Simulation,” James E. Park, Mohamad A. Rezvani, PVP-Vol. 314, ASME 1995.
2) “A Seismic Analysis of the Emergency Water Tank”, Hu Yong Tao, Gu Tang Yu, PVP-Vol
314, ASME 1995.
3) “Welded Steel Tanks for Oil Storage”, API Standard 650 9th Edition, 1993, American
Petroleum Institute
4) “Guidelines for the Seismic Design of Oil and Gas Pipeline Systems”, ASCE Technical
Council on Lifeline Earthquake Engineering, ASCE, 1984.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


3.1.40
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

BOS Fluids Reciprocating Compressor Pulsation and Mechanical Analysis


Reciprocating compressors produce pulsations in the suction and discharge piping that can be
damaging both to the piping and to the machine. API 618 gives allowable pulsation levels and
directives for the type of analysis required based on compressor power and discharge pressure.
(See API 618 Figure 4.) Even though API requirements are satisfied, piping systems may still
vibrate when reciprocating compressors are running. Pipe vibration is a function of the strength
and frequency of the pulsation, and the natural frequency of the pipe and supporting structure.
BOS Fluids makes it easy to check the mechanical shaking forces and the API 618 allowables to
be sure that the compressor installation runs smoothly.

A typical reciprocating compressor cross section is shown in Figure 1, and several installations
are shown in Figure 2.

Figure 1 – From Oil-Machinery Maintenance Online – Recip Compressor Cross Section

Figure 2 – Typical Compressor Installations

The piston displacement function for all reciprocating compressors takes the following general
shape:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


4.1.1
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Figure 3 – General Reciprocating Compressor Piston Displacement

When looking at the discharge side of the cylinder a sawtooth gas flow wave is observed due the
sequence of piston movement and valve opening and closing.

A sawtooth flowrate function appears at both the suction and discharge side of the cylinder. The
shape of the sawtooth is determined by the rotational speed of the compressor, the geometry of
the cylinder and the pressure ratio.

When a discrete Fourier analysis is performed on the sawtooth function, the strength of the
individual flow harmonics is determined:

Figure 4 – Discrete (left) and Continuous (right) Fourier Transform of Saw tooth

The discrete Fourier analysis of the flow saw tooth indicates which harmonics of the compressor
running speed produce the largest pressure amplitude.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


4.1.2
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Fundamental Reciprocating Compressor Frequency = f1 = RPM / 60


First Harmonic of the Running Speed = (2) (f1) = (2) (RPM) / 60
Second Harmonic of the Running Speed = (3) (f1) = (3) (RPM) / 60

The pressure amplitude for usually the first ten harmonics are used with the acoustic model of
the piping and bottle system to predict the pulsations and unbalanced forces. Any pulsations
must be smaller than the limits given in API 618 and the unbalanced forces must not produce a
fatigue failure of the piping system.

The bottle sizes for most reciprocating compressor suction and discharge piping systems are
sized according to API 618 3.9.2.2.2. For larger machines, usually some multiple of this bottle
size.

A BOS Fluids analysis starts with the basic geometry of the piping system:

Figure 5 – East Side Suction and Bottle Shown for Acoustic Study

A FIXHEAD or SONICHEAD boundary condition is specified at the suction source, or at the


discharge vessel or long pipe connection. (Suction and discharge lines are analyzed separately.)

The bottle is modeled as a large pipe with a surge vessel located at the center of the pipe to
incorporate gas compression effects.

A small surge vessel is also located at the compressor suction inlets and at the discharge outlets
of the compressor to simulate the gas compression effects of the valve manifold.

A RECIP boundary condition is also placed at the suction inlets or discharge outlets to produce
the exact timehistory of the sawtooth waveform that is exciting the system. The input required
for the BOS Fluids RECIP boundary condition is:

1) Stroke (m.)
2) Piston Area (m2)
3) Compressor RPM
4) Cylinder Dead Space (m3)

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


4.1.3
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

5) Phase Shift (deg) (for multiple cylinders)


6) Ratio of gas specific heats
7) Compression Ratio (Pdischarge / Psuction)
8) Double Acting Factor

The double acting factor gives the ratio of the “second” pulse in a cylinder that compresses gas
on both ends of its stroke. The second pulse is smaller than the first pulse because the piston rod
reduces the volume of the cylinder area. Basic cylinder geometries are shown in the figure
below.

Figure 6 – Compressor Cylinder Types

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


4.1.4
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Output from the BOS Fluids analysis is the allowed API 618 pulsation report and unbalanced
forces and moments in the piping system. A portion of an API 618 pulsation report is shown
below:

Figure 7 – BOS Fluids API 618 Pulsation Report

The report shows if any harmonic component pressure amplitudes have been exceeded.

BOS Fluids also generates unbalanced force plots. These plots can be animated in the 3D viewer,
or can be viewed statically. An example maximum unbalanced force plot is shown below.

Figure 8 – Maximum Unbalanced Force Plot

2D plots of the unbalanced loads, pressure, flowrate and velocity can also be generated. An
example unbalanced force plot in both the time and frequency domains are shown below:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


4.1.5
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Figure 9 – Time History (left) and Frequency Domain (Right) Plots of Unbalanced Loads

Either harmonic loads or the exact unbalanced transient loads can be input to the pipe stress
program to perform a mechanical analysis. Time history ascii data files are generated by BOS
Fluids automatically whenever a 2D plot is displayed. A portion of the transient loading input
for the CAESAR II pipe stress program is shown below.

214.3 862.3
215.9 403.9
217.5 202.5
218.3 7.9
219.1 -38.6
219.9 -240.4
221.5 -193.1

If excessive pulsations are found in BOS Fluids or excessive stresses are found from the pipe
stress analysis, orifices can be inserted into the system at points of high velocity, (typically at the
inlets or outlets of the compressor bottles). Structural steel should often be included in the
mechanical analysis since the natural frequencies and the mode shapes of the pipe in the
harmonic area of interest are strongly influenced by the mass and stiffness of the structure.

Figure 10 – Example CAESAR II Models and Dynamic Displacements

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


4.1.6
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Bos Fluids & API 618

1. Introduction
Most problems with reciprocating equipment occur because the major running speed, or an
harmonic of the major running speed is coincident with an acoustic or mechanical natural
frequency. API 618 for compressors and API 674 for reciprocating pumps are intended to control
pulsation levels such that resonance and stress problems do not occur. It is the user’s responsibility
however, to investigate all potentially dangerous acoustic and mechanical resonances. It is a
detective’s game, and not all field problems can be discovered at the design stage. The designer’s
goal, however, is to prevent obvious problems associated with the fundamental pulsation
frequencies, and to include sufficiently complex mechanical models so that side-to-side piping
loads not normally experienced can be tolerated without excessive displacements.

In paragraph 3.9 of API 618 the pulsation and vibration control requirements for compressor and
connected piping for Reciprocating Compressors for Petroleum Chemical and Gas industry
Services are addressed. Three design approaches are different in their requirement of mechanical
and acoustical simulation effort to be made by the designer. The selection of the proper design
approaches depends among others on compressor properties and is clearly defined in the API
Standard.

In appendix M of this API standard the various steps to be taken during the different design
approaches are given. In all approaches two different aspects are to be distinguished: the acoustical
simulation and the mechanical response.

In the acoustical simulation the level of the pressure pulsations in the various components of the
compressor package (cylinder passage, bottles, piping) is analyzed and in the design phase
measures are taken to limit the level of the pressure pulsations. API Standard 618 specifies limits to
the acceptable pulsation levels. For instance, equation (8) of paragraph 3.9.2.7 specifies a rule to
determine acceptable limits to the pulsation levels in the piping connected to the compressor
battery. The allowable pulsation level is depending on the frequency and the pipe diameter and it is
also depending on the average line pressure. In fact the allowable pulsation level is inversely
proportional to the square root of the applicable pipe ID and the pulsation frequency and directly
proportional to the square root of the line pressure.

The API allowable acoustic level is specified on a per frequency basis and so each frequency
contribution to the pulsation must be evaluated separately. There are several ways that these
analyses have been historically performed. BOS Fluids uses the time history approach and
automatically decomposes the solution pressure pulsations producing an API 618 pulsation
compliance report for all sections of the piping system much like a stress analysis program reports
the stresses and allowables for the B31 code.

User’s that have never performed a pulsation study previously are encouraged to apply the method
described below to a system that has been previously analyzed to be sure that they can produce
similar pulsation levels in the different parts of the system.

A few lines from an n example report is shown below:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.2.1


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

BOS Fluids also allows the user to perform a separate harmonic analysis for each compressor
loading component if desired and lets the user build a complete compressor suction and/or
discharge model so that the suction and discharge saw-tooth functions can be generated, but these
methods, while more similar to existing approaches, are not required when the time history BOS
Fluids method is employed.

Even acceptable pressure pulsations generate unbalanced forces in the compressor suction and
discharge piping. The effect of these unbalanced forces is also calculated and may be plotted and
used in a mechanical response study.

Compressor studies are usually undertaken either at the design phase or because a problem has
been experienced in the field. Field problems are usually the easiest to solve because the analyst
knows there is a problem and because numerous measurements have been taken to help identify the
exact nature of the pulsation. Field problems can also be easier to solve because models of the gas
and the compressor excitation can generally be much simpler than in the design phase. “Design
Phase” solutions require a complete set of information about the compressor, anticipated gas
composition changes, climate variations, etc.

BOS Fluids compressor models can include the mechanical properties of the valves, expansion
chambers and local compressibility, but these are sophisticated models and require an additional
level of detail and machine understanding. Models of this type are discussed further in Chapter 4,
Section 2.

The majority of field problems are due to either:

1) Incorrect or non-existent pulsation suppression, or


2) Poor supporting,
3) Matching of mechanical and acoustic natural frequencies at a machine harmonic.

In any of these cases, the measured significant response (displacement of the pipe system) occurs at
one of the harmonics of the system running speed, and the mode of vibration is easy to predict with
a computer program that calculates the natural frequencies and mode shapes of pipe and structural
systems.

“Design Phase” analysis is typically more difficult because the analyst must make sure that all
possible scenarios are simulated. All potential excitation frequencies and gas conditions must be
investigated, and all the applicable rules of API 618 must be satisfied. Excessive pulsations can
prevent the effective running of the machine, can damage valves and other mechanical parts, and

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.2.2


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

can produce large shaking forces in the piping. Climate and local ambient air changes can often
result in variations in the speed of sound of up to 15%
Typically, the user will analyze the variety of gas and ambient conditions as separate load cases.
The worst of these will then be used for the mechanical response study.

The example included in Chapter 4, Section 1 is focused on the field vibration approach to solving
reciprocating compressor problems since these are typically easier to understand and follow. More
sophisticated reciprocating compressor models are discussed in Chapter 4, Section 2, and are only
referred to in this Section. The interested reader is encouraged to read both sections thoroughly.

Several examples of compressor parts (Dresser Rand) and installations are shown below:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.2.3


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Typically a model of the piping system is retrieved from the pipe stress analysis of the system.
CAESAR, FE/Pipe, PipePlus and Triflex models of the piping system can be imported directly into
BOS Fluids.

It is often most convenient to prepare the piping model first for the fluid calculation in the pipe
stress program by removing beam elements that do not transmit gas. These include beam sections
that are rigid elements with zero weights, simulation of expansion joint hinges and directional
members, centerline extensions for branches, etc.

Once the beam model is prepared it should be read into BOS Fluids. The translator includes the
routing, pipe diameters and wall thicknesses. All gas related data and boundary conditions are
added in BOS Fluids.

The following items should be noted when preparing the compressor acoustic model:

1) It is not uncommon for compressor pulsations to appear at the inlet to 1-to-2 km long pipelines.
Straight pipelines transmit the traveling pressure wave very well. The user should investigate the
sensitivity of the long pipeline length on the solution, and use the LONGPIPE boundary condition
when the solution is not affected by this simplification. Interaction effects between plants or
stations 10-to-15 km apart have been observed.

2) The compressor boundary condition should be modeled with the RECIP specification. Large
volumes, such as offgas supply columns, knock-vessels, etc. can be modeled as fixed pressure
sources.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.2.4


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

3) The valve manifold volume should be included in the evaluation to obtain a proper estimate of
the pressure pulsations. The user should add this volume attenuation mechanism at the RECIP
boundary condition. If the valve manifold volume is not known, then 3-to-4 times the cylinder dead
space can be used as an approximation. If this volume becomes a critical item in the design, then
the actual volume can be verified by the manufacturer or determined from test. Omitting this small
volume will result in higher pulsations and more conservative results.

4) The surge bottle should be modeled as a large pipe whose length is the actual length of the surge
vessel. Extra elements should be added outside of the nozzle attachment locations to be sure that
the actual length of the bottle is properly defined. Low frequency and lower harmonics should not
be significantly affected by the actual bottle length, but for higher frequency or harmonic
responses, the bottle length can contribute to the frequency response.

5) The total bottle volume should be placed at a node at the middle of the surge vessel. Liquid
levels in the bottle should be zero. Vertical surge vessel elements should be used.

6) Control valves or pressure let down valves can also serve as compressor piping acoustic
boundary conditions. When the pressure ratio is greater than about 1.9 sonic conditions will exist
across the valve and a SONICHEAD boundary condition can be used to simplify modeling of this
boundary condition.

7) When modeling back to the well-head – the fixed pressure is often taken at the xmas tree. Since
the formation volume for the present purpose acts as a large volume of constant pressure.

8) When selecting bottle sizes, in particular for larger units, it is not uncommon to double the
minimum bottle size recommended by API 618. When the actual bottle size approaches the
minimum recommended bottle size in conformance with API 618 for larger machines the user can
expect larger shaking forces, and be expected to pay more attention to the pulsation design of the
system.

9) Orifices that drop the pressure about 1% can be placed at typically the entrance or exit of the
bottles to help attenuate pulsations. These are typically large bore orifice plates, i.e. about 90% of
the flow area, and can provide the equivalent pressure drop on the order of a valve or
approximately one-hundred diameters of pipe. The normal orifice discharge coefficient (Cd) is
about 0.6, but orifices lose their efficiency at higher frequencies, and in these cases the user might
consider multiple bore orifice plates. Most orifices have a clearly marked flow direction and
should be installed carefully for this reason. The flow direction is always from the small opening
end to the large opening end. This may be particularly important for compressor piping systems
where the downstream vortices provide the majority of the attenuation. Orifices placed in the
middle of long runs will often miss attenuating certain modes and should be applied with care. The
placement of the orifice should be at locations of the acoustical mode where the velocity is high.

10) Variable speed machines should be analyzed very carefully. Usually an analysis is performed
for six or seven different speeds covering the full speed range, but the operator should be consulted
to determine if any single delivery condition is thought to predominate. Acoustic and mechanical
natural frequencies should be studied carefully in these cases.

11) As a rule of thumb, 360 RPM machines typically produce large pressures but smaller
pulsations, 600 RPM machines can be considered intermediate machines, and machines running in
the 1000 RPM and higher range are considered fast running machines, and can produce the

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.2.5


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

pulsations over the broadest frequency range. API 618 Paragraph 3.9.2.6 Figure 4 should be
consulted for the level of attention that should be paid to the pulsation study.

12) Many suppliers recommend starting the compressor and driver in the unloaded condition. When
this is the case, the startup of the compressor is essentially instantaneous. This is the startup
simulated in the BOS Fluids analysis. Since most compressor simulations are harmonic, they
cannot consider startup conditions. Some transient vibration is expected during most startup
debottlenecking and problems related purely to startup pulsations are rare. These cases can be
analyzed by BOS Fluids, but the startup transient is generally ignored. The API start and stop
times in the input are included so that the user can ignore the startup transient. (A part of this screen
is shown below.) The startup time is typically taken some time after any startup perturbation has
passed. Usually 5-to-10 seconds after the start of the compressor. The user must experiment with
longer startup times to be sure that the transient is not interfering with the transient calculation.
This is usually an easy thing to do.

For the example above, the RPM’s are 960 and the start time is 10 seconds. This means that the
main frequency for a single acting cylinder is 16 Hz. and for a double acting cylinder is 32 Hz. In
10 seconds a single acting discharge cylinder would have discharged (10)(16) = 160 times, and the
double acting discharge cylinder would have discharged (10)(32) = 320 times.

13) The no-load startup of a compressor can be simulated with BOS Fluids by setting the start time
to 0.0 seconds. If some other startup procedure is expected and needs to be analyzed the reader is
encouraged to contact the Paulin Research Group.

14) Shaking forces (unbalanced loads generated in BOS Fluids in the 2d modeler) on the order of
25 lb. peak-to-peak per inch of diameter are generally of interest and may warrant a mechanical
dynamic analysis. This is a rule-of-thumb only and is clearly a function of the pipe routing and its
supports.

15) The API 618 Start Time must only be long enough to trap the lowest frequency system
response to the compressor pulsations. The start time should normally be set to:

API 618 Start Time = (100) / (RPM/60) sec.

16) The API 618 Stop Time should trap at least 10 of the lowest running speed of the machine. The
stop time is usually set to:

API 618 Stop Time = Start Time + 600/RPM <or>


= Start Time + 5 Sec
… whichever is larger

17) The BOS Fluids total output transient time and time step must also be controlled. These inputs
are specified on the OPTIONAL data form shown in part below:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.2.6


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The Maximum Simulation Time should be at least equal to the API 618 Stop Time. The Maximum
Time Step should be set to trap the highest harmonic of interest. The highest harmonic of interest
is usually the 10th harmonic, but is a function of the RPM, pressure magnitude, etc. The maximum
time step can usually be estimated as:

Maximum Timestep = 0.6 / RPM sec.

Two approaches to the reciprocating compressor problem will be presented below. The first
method provides a quick and straight forward approach. The second method allows the user
to perform a detailed study of the system harmonically, is the more historically representative
method, and provides a greater understanding of what is being done. The second method
also probably takes three-to-four times longer to perform.

Direct Approach
The Direct Approach can be broken down into the following steps.

1) Generate the basic piping configuration. This can be imported from a pipe stress program or
built in BOS Fluids.
2) Be sure that each compressor bottle has a node point at its middle.
3) Model to the discharge or suction flange of the compressor.
4) Suction and discharge systems can be analyzed separately providing there is no open flow path
between them.
5) Model away from the compressor to the first large change in volume that will reflect all
incoming pulsations. This might be:

a) Knockout drum
b) Offgas column connection
c) Pressure letdown valve
d) Well head
e) Long Pipeline (Must usually model into the pipeline for some distance before a LONGPIPE
boundary condition can be used. The distance to be modeled into the pipeline is a function of
what the user desires to study. The interaction of compressors at facilities more than 5km apart
have been witnessed, but typically only the piping in the vicinity of the compressor is of
concern. The user must make this determination when long pipelines provide either the suction
or discharge boundary condition.

6) Collect the data for the suction or discharge compressor cylinders that should be analyzed. This
data is used in the BOS Fluids RECIP boundary condition and includes:

1) Stroke (m.)
2) Piston Area (m2)
3) Compressor RPM
4) Cylinder Dead Space (m3)
5) Phase Shift (deg) (for multiple cylinders)
6) Ratio of gas specific heats
7) Compression Ratio (Pdischarge / Psuction)
8) Double Acting Factor

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.2.7


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The double acting factor is nonzero if the cylinder is utilized on both ends of its stroke. The double
acting factor is the difference between the outboard side and the inboard side effective volume.
Every other waveform sawtooth is multiplied by the double acting factor for double acting
cylinders. A double acting cylinder on a 1000 RPM machine introduces the primary pulsation
harmonic at twice the machine running speed since the cylinder discharges or intakes gas twice per
shaft revolution. Some component of pulsation also exists at one times the running speed. This
component is produced by the difference in the sawtooth waveforms.

7) Use one RECIP boundary condition for each suction or discharge flange modeled. Be sure to use
the phase shift to describe the relationship between the cylinders on the crankshaft.

8) At the boundary condition removed from the compressor apply either a FIXHEAD, or
SONICHEAD boundary. SONICHEAD would be used at most well heads and at pressure let down
stations or control valves where the ratio of the pressure drop across the valve is greater than 1.9.

9) At the suction or discharge flange add an empty surge vessel whose total volume is equal to the
valve manifold volume. Not including this volume will provide more conservative results. If the
valve manifold volume is not known and cannot be obtained from the manufacturer then a value
equal to 3-to-4 times the cylinder dead space can be used. The isentropic exponent should be the
value for the gas.

10) At the middle node on the compressor bottle an empty surge vessel element should be added
whose total volume is equal to the total volume of the bottle. The isentropic exponent should be the
value for the gas.

11) Enter a USER fluid whose gas properties are based on the average line pressure.

12) On the BOS Fluids ACTION form – enter API618 for the analysis type, and then specify the
five API 618 parameters in the middle of the screen as shown below and discussed above in items
15-through-17.

13) On the OPTIONAL data form be sure to enter the Maximum Simulation Time and the
Maximum Timestep as described in item 17 above. (The Maximum Simulation Time will be
increased automatically if necessary to produce the required API 618 Stop Time.)

14) All previously entered data should be thoroughly checked, and then the job submitted for
analysis. The steady state and transient solutions should run through without difficulty. The user
should then enter, REPORTS: RESULTS and should see

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.2.8


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The Steady State Warnings & Messages should be reviewed carefully, and the Transient Warnings
& Messages should be reviewed carefully so that all assumptions and calculations made by the
program are correct.

A short excerpt from the API 618 Pulsation Compliance Report is shown below:

For each pipe section in the model the maximum pressure amplitude is printed along with the
allowable per API 618. Where the calculated pulsation exceeds the allowable the label

** Excess **

is added to the right as shown above.

In those sections where the pulsation amplitudes are high the user should inspect the unbalanced
forces also. The unbalanced forces, flow and pressure results are generated by accessing the 2D
Plots. When asked to make a selection, the user should select the Window menu option 2D:

From the 2D option the user should select Pipe 2D Plots to produce plots for sections of pipe.
Non-Pipe 2D Plots are used to generate plotted output for valves, pumps, surge vessels, etc. To
generate a 2D plot the user should select the type of plot desired and enter the node number where
the acoustic data should be extracted. Example input for a time history pressure plot at node 100 is
shown below:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.2.9


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The resulting time history pressure plot is shown below:

There are several interesting features demonstrated on this plot. The first is that the startup
transient can be seen to occur for approximately four seconds. This confirms that a five second
start time is sufficient. Secondly the pressure amplitude can be visually checked. After the startup
transient has died out the pressure amplitude can be seen to be ( 23.6 – 22.0 ) / 2 = 0.8 bar.

A frequency decomposition of this time waveform can be produced by selecting the radio button:
Frequency Transform. The input and resulting plot of the pressure frequency transform are shown
below:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.2.10


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The 32 Hz and 64 Hz harmonics are clearly shown in the plots. The frequency amplitude plotted
above has no quantitative significance in a continuous frequency transform. It is only useful in the
sense that it can be compared to other harmonics of the running speed.

Unbalanced force time history plots are generated by selecting the Unbalanced Force and Time
History radio buttons:

The 25 lb. peak-to-peak per inch of pipe diameter suggested in item 14) above translates into
(25)(8) = 200 lb. for the eight inch pipe size at node 100. This is (200)(4.448) = 889.6 Newtons.
From the above plot, the shaking force at node 100 after the startup transient has died out can be
seen to be about (1100 – (-1500) ) = 2600 Newtons. Either the section of pipe that includes the
node 100 must be able to absorb this load at 16 Hz and 32 Hz. or additional restraints should be
added.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.2.11


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Once the 2D time history plot is generated, BOS Fluids automatically generates an ascii data file
that contains the time history output to be read into a pipe stress program. For the node 100, this
file is named PC100 for the CAESAR style of transient input.

User’s that have never performed compressor simulations should read the Systematic Approach
section described below to get a better understanding of what is required when performing a
pulsation study. The details of the analysis can be obscured when using the Direct Approach since
much of the method is obscured by the automation of the process.

Systematic Approach
The method described below gives a more systematic approach to the reciprocating compressor
problem and describes the basic analytical tools available in BOS Fluids to address this problem in
an harmonic manner, where each compressor harmonic is evaluated individually. The method
described below gives one a more thorough understanding of the basic compressor pulsation
problem, but is somewhat more tedious to implement. The knowledgeable user will most likely find
himself using parts of both analyses in the final evaluation of the system.

1.1.Acoustical natural frequencies

High pressure pulsation levels can result from a number of causes. One possible cause is acoustical
resonance. Acoustical resonance occurs if the compressor excitation frequency or one of its higher
harmonics coincides with one of the acoustical natural frequencies of the system. By means of a
BOS Fluids analysis the user is able to determine the acoustical natural frequencies of the piping
system and make sure that sufficient separation exists between the acoustical natural frequencies
and compressor harmonics. The acoustical natural frequencies are related to the reflection of waves
in the pipe system. The wave reflection period is proportional to the ratio of pipe section length and
wave speed. Therefore the natural frequency values can be shifted by changing the unfavorable
pipe section lengths. If shifting the frequencies is impossible, additional damping (orifice plate)
may be introduced to limit the pulsation amplitude at a (near) resonance condition.

1.2 Pulsation study


The pulsation time history generated by a reciprocating compressor is periodic with a period related
to compressor RPM. However the pulsation time-history is generally not perfectly harmonic. The
periodic pulsation time history also contains components of higher harmonic frequencies above the
basic compressor frequency corresponding to the running speed: RPM/60. All other contributions
in the periodic pulsation can be simulated by the addition of higher harmonic terms. These higher
harmonics are some multiple of the basic compressor frequency.

The allowable pulsation levels as defined in paragraph 3.9.2.7 of API 618 are given on a per
frequency basis. Therefore the pressure pulsation assessment has to be performed on a per
harmonic frequency basis. BOS Fluids provides a module by which the user can decompose the
compressor pulsation time-history signal into harmonic components by means of Fourier series
decomposition. (Usually only the first ten to twenty components are taken.) Most often the exact
compressor suction or discharge flow pulsations must be evaluated. BOS Fluids provides a RECIP
boundary node type which requires the following data:

1) Node type, i.e. suction or discharge


2) Stroke (m.)
3) Piston Area (m2)
4) Compressor RPM
5) Cylinder Dead Space (m3)

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.2.12


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

6) Phase Shift (deg) (for multiple cylinders)


7) Ratio of gas specific heats
8) Compression Ratio (Pdischarge / Psuction)
9) Double Acting Factor

The RECIP boundary condition generates the intermittent saw-tooth periodic waveform for either a
single-acting or double-acting compressor. (See Figure Below)

Figure A. Intermediate Sawtooth Flow Discharge

Once a model for the compressor is developed, a discrete Fourier analysis is made of the sawtooth
discharge at a convenient point in the piping system. (The reciprocating compressor model is
described in greater detail in Chapter 4 Section 2.) It is also possible to produce a simple
compressor suction or discharge model by the use of a harmonic flow pattern and check valves.
The simple model will be included in the example below, and the more sophisticated model will be
described in Chapter 4 Section 2.

For each of the discrete Fourier components another pulsation analysis is made and the pressure
pulsation levels compared to the allowable pulsation level on a pipe section basis per API 618.
BOS Fluids provides a module to assist in this comparison. Each of these modules are discussed
separately in Section 4.2 and are described in the example problem below.

1.3 Mechanical response analysis


One output from the pulsation analysis for each pipe section is the magnitude (amplitude) of the
unbalanced forces. The unbalanced force components are used in the mechanical response analysis.
This mechanical analysis is performed by means of a pipe stress package with dynamic harmonic
capabilities. CAESAR II, PipePlus and FE/Pipe each have sufficient harmonic capability to
perform this calculation. Both CAESAR II and FE/Pipe can include the structural framework which
is recommended for most compressor systems. FE/Pipe can also include the structural steel and
shell models of compressor and bottle components, which is sometimes needed in tight off-shore
piping configurations. The calculated support/nozzle loads and stresses are assessed by comparison
to their allowable levels per API 618. For the stresses guidance on the allowable stress levels is
given in paragraph 3.9.2.2.1 of API 618, which basically limits peak stresses to the endurance limit
for the material.

2. Implementation
In this section the user is guided thru the different steps that are taken during a typical analysis
process. For illustration purposes reference is made to the example case with file name SUCTION
included in the <installation>\BOSMODELS directory. The example model is a truncated beam
model of the suction piping connected to the compressor. Most BOS Fluids models are extracted
from mechanical models built in either CAESAR II or FE/Pipe. These mechanical models often
have supports, rigid element constructions, etc. that are not part of the gas containment system.
These extraneous parts of the model must be removed before the system is analyzed in BOS Fluids.
The pipe model of the bottle has been included to its attachment flange. All sections of the system
that are closed off by valves or blinds, and all springs, supports, struts and structural steel have been

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.2.13


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

removed. See Figure 1 below. Of course, the BOS Fluids model can always be built from scratch
in the BOS Fluids user interface, but this is not typically done since a mechanical thermal and
dynamic analysis must usually be performed in addition to the acoustic analysis, and it is desired to
enter the basic geometry of the model only once.

Figure 1. Plot of the example model (Jobname=SUCTION).

Once the model is available, the next step is to input the gas properties. For that purpose the user
selects a user defined fluid as shown in Figure 2 below.

Figure 2. First Pipe Card, “User” defined fluid/gas properties

In the fluid definition screen the user fills in the applicable gas properties similar to the way it has
been performed in Figure 3 below. At the “Design Phase” the user must be sure to consider all
possible variations of gas properties and temperatures as acoustic natural frequencies depend
strongly on these values.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.2.14


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Figure 3. Definition of gas properties.

2.1. Step 1; Extraction of the acoustical natural frequencies.


In order to extract the acoustical natural frequencies the piping model should be excited by a
pressure step disturbance at the compressor flange or at a far end boundary. In the example problem
(compressor suction piping) the excitation has been applied to the upstream boundary condition
(node 10). See figure 4 for the boundary condition example input. For the compressor in the
example the average suction pressure is 23.4 Barg and a 2 Barg step has been introduced.

Figure 4. Transient input condition for acoustical natural frequency extraction.

The acoustical natural frequencies are determined from the calculated pressure time history data.
The highest natural frequency that will be extracted is related to the sampling time increment, i.e.
the time interval between two output time steps. The shorter the sampling time increment the higher
the frequency that can be extracted. Aliasing of the high end frequencies is avoided by cutting the
frequency spectrum off below the Nyquist frequency. (The Nyquist period is half of the total
sample period.) The “total simulation duration” influences the lowest value available on the
frequency plot. The longer the duration, the lower the frequency that can be plotted. Selecting short
sampling periods and long simulation times may create large data-files. These aspects have to be

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.2.15


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

considered when entering values in the “optional” screen that control the analysis. (See Figure 5
below for the values used in the example problem.)

Figure 5. “Optional” Screen parameters to be filled in

The maximum simulation time is 20 seconds. This means that the lowest frequency viewable will
be have a period of approximately 20/4 = 5 sec. which corresponds to a frequency (f = 1/p) = (1/5)
= 0.2 Hz. The maximum solution time step is limited to 0.001 seconds or less and the output time
increment is the same as the maximum allowed output time step. This guarantees that solution
output will be generated at a rate of at least once per millisecond. (If this is too much output to
postprocess, the user can limit the parts of the system are processed – see the Force/Exclusion
lists.) The Nyquist period is half of the sampling rate, and so it is (0.001)(2) = 0.002 sec. The
Nyquist frequency is the inverse of the period and is equal to (1/0.002) = 500 Hz. The expected
frequency resolution should therefore be within the 0.2 –to- 500 Hz range based on the control
provided on the OUTPUT screen. The Maximum Pressure Trapping has also been deactivated.
This guarantees that pressures are for only the node reported. Otherwise BOS Fluids will look
anywhere in the pipe segment adjacent to a node for the maximum pressure. This is generally
desired when the system must be designed for the maximum pressure, but is NOT desired when a
precise acoustic waveform is sought.

Once the optional screen has been filled-in a BOS Fluids transient run is made. The transient option
is activated in the ACTION screen as shown below. The only transient perturbation is the pressure
time history entered on the boundary condition screen, i.e. there is no pump or valve transient (see
figure 6). The user is also encouraged to deactivate column separation when compressor models are
run in case low pressures are encountered. The default column separation model is CAP. This
should be changed to NONE.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.2.16


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Figure 6. Action screen

The job is submitted for analysis by activating option F from the main BOS Fluids job menu. The
user will see a steady-state solution screen flash quickly, and then the transient screen count down
the total time requested for solution. (See Figure 5.) When the run has been completed the user is
returned to the PRGMAPs main menu and the output is ready to be processed.

A suitable node in the piping system must be selected for output. Preferentially this node is in the
center of the pipe system. Since the selected node could incidentally be on an acoustical node
for one of the acoustical mode shapes (a location where the standing wave pattern shows zero
pressure fluctuation) the output process described herein should be repeated for a few other
nodes in the system in order to be sure that a complete picture of the pulsations is obtained.
For illustration purposes the output processing is show for one node only. Node 170 is elected
which is at the Tee-connection where the suction line branches to the two connected compressors
(see figure 7)

Figure 7. Node distribution over part of the system.

The acoustical natural frequencies should be extracted from the time history response for node 170.
In order to access the time history response, the PRGMAPs screen the option 2-D plots is selected.
(See Figure 8.)

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.2.17


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Figure 8. Access to 2D plots

In the 2D plot screen the pressure time history for node 170 is selected (see figure 9)

Figure 9. 2-D plots selection screen

The generated output shows the pressure time history at node 170 for the duration of the selected
simulation time and as a result of a 2 Barg step-function applied at the suction inlet. These time
history plots typically look like the one shown in figure 10 below.

Figure 10. Pressure time history at node 170.

From the time history picture it is obvious that the pressure is oscillating. It is less clear, but
nevertheless a fact, that the pressure signal also contains a few additional periodic components and
one dominating frequency. Typically the single dominating frequency is related to the pressure
wave traveling up and down the complete system from one boundary condition to the other.
Reflections from dead-ends (stubs), bends and tees are also present. These secondary reflections are
relatively weak, but they do have an influence on the frequency content of the response signal.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.2.18


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Moreover, it also clear that damping is present and that the initial violent oscillation dies out over
time.

The exact frequency content is difficult to assess from observation alone. In order to get a better
view of the frequency content the pressure time history is Fourier Analyzed. This is done from the
plots menu (see figure 9) by activating the plot-type: “frequency transform.” The frequency plot
appears as shown below in Figure 11. Frequency replaces time along the horizontal axis, and
amplitude replaces pressure along the vertical axis.

Low Frequency End

Figure 11. Frequency content of the pressure time history at node 170.

From the above frequency plot it is clear that the pressure time history contains large frequency
components at the low frequency band of the spectrum while there is hardly any contribution at the
high frequency end (> 100 Hz.). This is typical for compressor piping, which justifies the use of
larger output time steps. Recall that 1 millisecond was used in the example to trap frequencies up to
approximately 500 Hz.. The BOS Fluids spectrum analyzer only produces frequency results up to
300 Hz., and so is even a little more conservative. For liquids or very high pressures (such as in
ethylene hyper-compressors), there may be high frequency harmonic components. These
frequencies are often associated with valve mechanical frequencies, but may resonate with
mechanical or acoustic frequencies in the shorter piping runs or manifolds. Measurements are often
required to demonstrate that higher frequency components exist, and these are only typically taken
when problems are witnessed. (In general, detailed models of the valves are not constructed at the
design phase as it is difficult to get the necessary data, and at the valve performance level the
acoustic problem becomes a three-dimensional problem, only approximately handled by the one-
dimensional simulation. When there are problems the higher frequency components will be found
to correspond to acoustic natural frequencies in the shorter pipe runs, for example between the
suction and/or discharge compressor flanges and the end of the manifold. For a particularly critical
system the user is encouraged to construct these valve models using

In order to have a clearer picture on the frequency content two steps may be taken. From the view
option in the pull-down menu the option zoom can be selected and a zoomed-in plot can be
generated (see figure 12)

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.2.19


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Figure 12. Plot of detailed section in figure 11.

In the detailed plot several peaks are visible and each peak implies an acoustical natural frequency
of some kind. An even clearer picture of the frequency content is obtained if the plotted data are
read in into an Excel sheet and plotted. Immediately after the frequency plot is generated in BOS
Fluids the data is available in a file with the name “modgen._i” in the working directory. The result
of an Excel plot of the same data can be massaged to look like the plot in figure 13 below.

Acoust i c Nat ur al Fr equenci es Suct i on Li ne ( a = 328 m/ s)

250

200

150

100

50

0
10. 00 20. 00 30. 00 40. 00 50. 00 60. 00 70. 00 80. 00 90. 00 100. 00

F r e q u e n c y ( Hz )

Figure 13. Pressure at node 170, frequency domain plot.

In Figure 13 the frequency peaks are clearly distinguishable. Each peak represents an acoustical
natural frequency of which there are several. It is clear that the dominant resonant peak appears at
about 18 Hz. This frequency corresponds to waves reflecting between suction inlet and suction
bottle connection.

Also included (in purple) are vertical lines at the compressor harmonics (the basic frequency and
the higher harmonics). These are added in Excel. In the ideal situation the peaks and the purple
lines should be well separated. “Separation” is a qualitative measure, but 20% is often used as a
quantitative measure of separation. In the higher frequency range the peaks are typically sharper
and less separation is required. The dynamic load factor (DLF) can also be used as a measure of
separation. A DLF of 2.5 is often used in the less than 100Hz. range and a DLF of 5.0 in the
frequency range over 100 Hz. DLF = ABS( 1 / [ 1- (w/wn)2 ] ). A good separation between
acoustical natural frequencies and compressor harmonics implies elimination of acoustical
resonance. In the present example resonant conditions are to be expected around the second
harmonic (33 Hz.)

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.2.20


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

For a given piping layout the acoustical natural frequencies are dependent on the speed of sound
inside the piping system. The speed of sound in the gas is depending on the temperature. In general
the operation envelope of the compressor shows a variation of the suction and discharge gas
temperatures. In order to examine the sensitivity of the results for temperature variations the
process has to be repeated at other temperatures (other speeds of sound). In figure in 14 and 15 the
results of analyses performed at different speeds of sound for the same pipe layout (the SUCTION
example file) are presented.

From an examination of these results it might be concluded that for the present layout if variations
in the temperature are considered acoustical resonant conditions will be difficult to avoid. In the
example case if layout changes would not be feasible introduction of an orifice plate is strongly
recommended. An orifice plates introduces additional damping by means of the generation of
vortices in an oscillating velocity field. In that way the high peaks at acoustical resonance are
suppressed. The optimum location for an orifice plate is at the location of maximum velocity. In the
present configuration this would be the inlet nozzle to the suction bottle. It should however be
considered that introduction of an orifice plate and the corresponding generation of vortices will
also introduce a pressure loss. This pressure drop can be translated into a reduction of compressor
efficiency and hence costs. These costs are to be weighted against the costs of other modifications
such as rerouting of the pipe
Acoust i c Nat ur al Fr equenci es Suct i on Li ne ( a = 344 m/ s)

250

200

150

100

50

0
10. 00 20. 00 30. 00 40. 00 50. 00 60. 00 70. 00 80. 00 90. 00 100. 00

F r e q u e n c y ( Hz )

Figure 14. Pressure at node 170, frequency domain plot at increased temperature.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.2.21


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Acoust i c Nat ur al Fr equenci es Suct i on Li ne ( a = 311 m/ s)

250

200

150

100

50

0
10. 00 20. 00 30. 00 40. 00 50. 00 60. 00 70. 00 80. 00 90. 00 100. 00

Fr e q u e n c y ( Hz )

Figure 15. Pressure at node 170, frequency domain plot at reduced temperature.

2.2. Step2; Pulsation simulation.


Once the acoustical natural frequencies have been extracted and measures have been taken to avoid
or suppress obvious acoustical resonances, the pulsation analysis can be performed. Since API 618
specifies the allowable pulsation levels on a per frequency basis the pulsation analysis is also to be
performed per harmonic component basis. For that purpose first the periodic pressure oscillation at
the compressor suction or discharge nozzle is decomposed into its harmonic components. This is
done by the BOS Fluids DFA (Discrete Fourier Analyzer) module. The Fourier Analyzer is
available from the PRGMAPs main menu under the BOS Miscellaneous Button:

2.2.1. Step2a; Discrete Fourier Analysis


The first step requires construction of a BOS Fluids model for the piston-cylinder-valve-bottle
geometry. This can be done in a pipe stress package like Caesar II or the user could build it directly
in BOS Fluids. An example for such a model is given in Figure 16 (file name:TEST002. This
model is available in the <installation>\BOSMODELS subdirectory.).

From node 10 to node 20 the cylinder space is modeled. Node 20-30 is a short pipe section to the
check valve between node 30 and 40. Further downstream the bottle is modeled as well as a
elatively long pipe section representing the connected pipe work.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.2.22


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Figure 16. Simplified model for cylinder/bottle/ configuration.

For this analysis the boundary condition applied is on the upstream node 10 and is an oscillating
flow simulating the movement of the piston. (See figure 17)

Figure 17. Harmonic Boundary condition at compressor side

The compressor frequency as well as the amplitude of the pulsation is indicated. The flow
amplitude is calculated from the piston diameter, the piston stroke and the compressor frequency.
From the piston stroke “S” and the compressor frequency “f” the piston speed amplitude “V” is
calculated:
V = π * S * RPM/60

In the example case where S = 135 mm and the RPM = 960 the piston speed amplitude
V = 6785 mm/s. For a piston with a diameter of D=240 mm this results in a flow rate amplitude of
0.30 m3/s. For an estimated average flow rate of 0.05 m3/s this results in an amplitude of 600% of
the average flow.

At the downstream end (node 130 – not shown) the boundary condition is represented by an
infinitely long pipe. Here the program allows for two options. The boundary condition that is first
order accurate is LONGPIPE and the boundary condition that is second order accurate is
LONG2PIPE. Also the downstream pressure must be specified. In the example problem the
downstream pressure is 60 Barg. (See figure 18 below.) The LONGPIPE boundary conditions do
not reflect pressure waves. The user of a LONGPIPE boundary condition should inspect the Pipe
Parameters report to be sure that there are at least two reaches in the pipe section that has a
LONGPIPE boundary condition at its end.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.2.23


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Figure 18. Downstream Boundary condition.

Figure 18a – Undamped Check Valve

An undamped check valve is used to simulate the cylinder suction valve as shown in Figure 18a
above. Note that the Closure Exponent is set to 0.0 to be sure that an undamped valve is used. The
check valve is used to produce the interrupted saw-tooth waveform that is characteristic of both the
compressor discharge and suction sides. The exact time for the opening of the check valve will not
be predicted as accurately as the RECIP boundary condition model will predict it, but less data is
also required.

Once the compressor model is established a transient analysis is performed. The acoustical
simulation of the compressor, bottle and pipe system focuses only on steady state compressor
rotation, and NOT on compressor startup characteristics. During the initial stage of the transient
analysis the time history effects of the harmonic startup will be visible. (See Figure 18b below.)

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.2.24


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Figure 18b – Transient Startup Effects

Friction will fade these startup perturbations with time, but they should not be included in the
evaluation of the steady state harmonic pulsation. The startup transient effects can be ignored by
specifying a window of interest that is past the startup perturbations. This time “window of
interest” is specified on the OPTIONAL screen. See Figure 19 below.

Figure 19. Selection of a 1 second (16 periods) output window from 9 to 10 seconds.

The user must pay the same attention to sample time and duration in the compressor model to be
sure to trap the frequency ranges of interest. When the compressor run is completed the pressure
output must be examined. Comparison of the frequency transform of the flow-rate and the pressure

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.2.25


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

at node 10 shows that the valve action introduces higher harmonic components in the pressure time
history. (See figures 20 and 21 below.)

Figure 20. Frequency transform of 1 second of flowrate time history at node 10.

Figure 21. Frequency transform of 1 second of pressure time history at node 10.

Figure 21a. Time History of Flow Downstream of Check Valve

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.2.26


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Figure 21a shows the time history of the flow discharge from the check valve. This should be
compared to Figure A from the RECIP boundary condition model. Figure 21a requires less data to
build. Figure A is more accurate. A further comparison of the methods is described in Chapter 4,
Section 2.

The frequency content (the distribution of peaks in the frequency spectrum) of the pressure time
history depends on the location of the “measurement.” Compare for instance the frequency content
at node 100 (downstream of the bottle) and node 50 (at the compressor discharge flange) (See
Figures 22 and 23 below.)

Figure 22. Frequency transform of pressure time history at node 100


(piping downstream of bottle)

Figure 23. Frequency transform of pressure time history at node 50


(compressor flange)

In order to enable the user to perform the acoustical pulsation analysis on a per harmonic
component basis as required by API 618, the pressure time history at the compressor flange (node
50) will be processed by the Discrete Fourier Analyzer (DFA).

The Discrete Fourier Analyzer is started from the BOS Fluids Miscellaneous button from the

PRGMAPs screen:

The Reciprocating Compressor Data form is shown in Figure 24 below.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.2.27


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Figure 24. Miscellaneous calculations selection screen.

After clicking the Reciprocating Compressors button the following form appears. (See Figure 25.)
The current directory in the window handle on the top of the form should point to the folder where
the compressor model was just analyzed. This is how the program finds the output data to be
analyzed.

Figure 25. DFA Compressor Pulsations

The required user input for the Discrete Fourier Analyzer is:

a. The file name where the data is to be read,


b. The compressor RPM
c. The number of harmonics to be generated by the DFA
d. The compressor nozzle node number ( in essence the node number for which the DFA has
to be performed )
e. The frequency cut off for the Fourier Transform Plot

When the “Generate Discrete Fourier Pressure Spectrum” button is pressed both the discrete
Fourier analysis results and the Fourier transform of the time history data is plotted as shown
above.
© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.2.28
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

DFA Output results.


The DFA module decomposes the continuous but periodic pressure time history at the compressor
flange into a series of harmonic contributions with an increasing harmonic frequency. The output
format is both graphical and numerical. In the numerical output the magnitude of the amplitude of
each harmonic component that contributes in the pressure oscillation is printed (in the pressure
units used in the model – Barg for the example problem). The harmonic number corresponds to a
multiple of the compressor base frequency.

In the final step of the acoustical simulation the user must perform a series of harmonic analyses for
the complete piping system downstream of the compressor flange. One analysis is required for each
amplitude/frequency pair from the DFA analysis. Typically ten harmonic analyses are used. This is
not as cumbersome as it sounds since any pulsation problems found during the analysis must be
solved on a per individual frequency basis also. The individual harmonic pulsation analysis is
described in the next chapter.

2.2.2. Step2b; Pulsation Analysis per Harmonic Component.


For this analysis the full piping model up to the compressor nozzle is used again. (The same model
that was used to extract the acoustical natural frequencies). For the purpose of the pulsation
analysis the model has been copied to the file name SUCTION2. The node connected to the
compressor is 10600.

The output of the Discrete Fourier Analysis is a series of pressure pulsation harmonic amplitudes
and frequencies. A separate analysis must be performed for each amplitude/frequency pair of the
set. The reason for the multiple analyses is two-fold. The API 618 standard specifies the allowable
pulsation levels in the piping on a per frequency basis. “Excesses” over the API allowables are also
“repaired” on a per frequency basis. At this stage of the analysis violations of the allowable API
618 pulsation levels are usually resolved by adding additional damping to the flow path (orifice
plates). The most efficient location for an orifice plate is at the location of maximum velocity. The
location of maximum velocity depends on the acoustical mode shape, therefore it is convenient to
have the results available on a per mode shape basis.

For illustration purposes only one of the harmonic pulsation analyses will be performed. Also for
illustration purposes it will be assumed that one of the outputs from the Discrete Fourier Analysis is
a pressure amplitude of 0.264 Barg at the basic compressor harmonic of 16 Hz. The average line
pressure is 26.4 Barg. The input for the harmonic boundary condition at the compressor nozzle is
shown in Figure 26 below.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.2.29


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Figure 26. Boundary condition at the compressor interface node 10600.

The boundary condition at the upstream end, in the present example is removed and an infinite pipe
is connected with a supply suction pressure of 26.4 barg. (See figure 27.)

Figure 27. Upstream Boundary condition.

Again, the influence of start-up of the transient must be eliminated from output and so an output
time window is selected in the optional screen. (See figure 28.)

Figure 28. Selection of an output window.

Once the run has been made the results are used several ways. The pressure at “each” node must be
compared to the API 618 allowable pressures, and the calculated unbalanced forces must be used in
the API 618 mechanical response analysis. Since the present analysis is an harmonic analysis the
output could be examined both graphically and numerically.

For instance consider the pressure at node 160. (See Figure 29). The graphical output shows a
regular harmonic pattern with the pressure varying between about 26.43 Barg and 26.36 Barg. The
resulting estimate for the pressure amplitude is 0.035 Barg. The frequency content of the pressure
oscillation is confirmed by the frequency transform. (See figure 30). A small contribution is made

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.2.30


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

by the second harmonic which is still left from start-up. A longer simulation time would eliminate
the second peak at 32 Hz, but this is considered unnecessary. The user could additionally ramp up
the harmonic analysis from the steady state condition to minimize the startup effects.
Experimenting with these techniques will be left to the user.

Figure 29. Pressure fluctuation response at node 160.

Figure 30. Frequency transformation of pressure oscillations indicated in figure 33.

The unbalanced forces as a result of the pressure fluctuations can be represented graphically (See
figure 31).

Figure 31. Timehistory of unbalance forces in pipe section at node 160.

The unbalanced force amplitude is slightly over 200 N. This is a typical order of magnitude for the
shaking forces in compressor piping. The output results are also available in numerical format in
the Maximum and Minimum Pressures, Flowrates and Forces report. See for instance Figure 32 for
a survey of the results at node 160.
© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.2.31
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Figure 32. Numerical Output.

From the maximum and minimum pressure report the pressure fluctuation amplitude peak to peak
can be found by subtracted the minimum value from the maximum value. For node 160 the
resulting pressure difference is: 0.071 Barg The amplitude of the unbalanced force is also printed
in the tabular report and can be seen above as 223 Newtons.

2.2.3.Step 2c. API assessment.


The API allowable pressure fluctuation is given by the equation:

3.971 ∗ Pl
Pallowable peak − peak [ B arg] =
Di ∗ f
where:

Pl = Average line pressure [Barg]


Di = Internal Pipe diameter [mm]
f = Applicable pulsation frequency

The internal diameter of the line at node 160 is 202.7 mm, the applicable frequency is 16 Hz and
the average line pressure is 26.4 Barg. The allowable pressure pulsation peak to peak per the API
allowable pulsation equation above is: 0.358 Barg. The pressure pulsation level at this frequency is
20% of the API allowable. BOS Fluids provides an option to automatically calculate the API
allowable pressure levels. Entry to this feature is also found under the Miscellaneous Calculations
button.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.2.32


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Figure 33. Miscellaneous Calculations Selection Menu


When the API 618 Allowables screen is displayed it contains a number of empty text boxes as
shown in Figure 34 below.

Figure 34. API Allowable calculation menu.

The API 618 Allowable Form allows the user to:

a. Select SI or English units,


b Specify the average line pressure (taken from the numerical output screen).
c. Indicate the compressor RPM and the number of harmonics for which the allowable level has to
be calculated.
d. Specify the various inside diameters of the pipes that are part of the compressor piping
arrangement. <or>
e. The user can check the “File Option” and fill in the filename and harmonic #. When the “File
Option” box is checked the diameters and pressures from the model file are read and used in the
API 618 compliance automatically. The user is recommended to employ this option as checking
for all pressures for each reach is done automatically. System peak-to-peak pressures for each
reach in the model are stored in the file <jobname>.API after each run.

After entry of the form data click on the “Find Allowables” button to generate the output both in
graphical and numerical format. For the example model SUCTION2 job, the output is given below
in Figure 35.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.2.33


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Figure 35. Allowable Pressure Calculations per API 618

The maximum peak-to-peak pressure differential for each pipe segment is reported and the
percentage allowable printed. Where the pulsations are in excess of the allowable the notice: “**
OVER **” is printed next to the data line. An example output is shown below.

From To dP %Allow Diameter


10 170 .0592 .1687417 202.7
170 220 .176 .4975906 202.7
170 1220 .1601 .4562823 202.7
220 280 .1602 .4413489 193.7
280 285 .0478 .1310044 193.7
280 10040 .1879 .5113589 193.7
1220 1230 .1649 .4594104 193.7
10010 10020 .2689 1.000788 363.6 ** OVER **
10020 10040 .2611 .9717582 363.6
10020 10200 .2752 .6496702 146.3
10040 10060 .1768 .6580117 363.6
10060 10070 .0314 .116859 363.6
10060 10600 .535 1.260315 146.3 ** OVER **

Use the report with care and make sure that parts of the system that show to be overstressed are not
artificial compressor elements representing the cylinder. (The RECIP boundary condition model
produces a more uniform pressure increase and tends not to suffer as much for local perturbations
close to the machine.) Any text parts of the report can he highlighted and copied using the
<Ctrl><C> keys. When <Ctrl><C> is used, the selected text is copied to the system clipboard.

The plots on the right hand side of the screen in Figure 35 show the allowable curve for each
diameter as a function of diameter. The user will note that as the diameter gets larger the plots
converge.

When a pipe section is over the allowable the user can interrogate the <jobname>.API file to get a
full list of the pipe segment and its reaches. Viewing this file the user can see just how much of the
pipe segment experienced excess pressures. A section of the <jobname>.API file is shown below:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.2.34


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

13
10 170 127
0.0000 23.7097 23.7298 23.6873 -1.6785 -1.6932 0.2027
0.2620 23.7223 23.7425 23.6999 -1.6785 -1.6933 0.2027
0.5239 23.7344 23.7547 23.7120 -1.6785 -1.6933 0.2027
0.7859 23.7465 23.7668 23.7240 -1.6784 -1.6933 0.2027
1.0479 23.7586 23.7790 23.7360 -1.6784 -1.6933 0.2027

The first number is the total number of pipe segments in the file. The next three numbers give the
from and to node numbers represented by the reach data that follows and then the total number of
reaches in the segment. For the example snippet above there are 127 reaches in the pipe segment
between the nodes 10 and 170. The data lines give:

1) Distance along the element from 10 to 170.


2) Average Pressure
3) Maximum Pressure
4) Minimum Pressure
5) Maximum Flowrate
6) Minimum Flowrate
7) Pipe Inside Diameter

The API 618 compliance report is run after each frequency harmonic analysis is made to check
against the API allowables. The API 618 allowables are a function of the harmonic being analyzed.

3. Step3.Mechanical response analysis.


The performance of a mechanical response analysis is part of the API 618 assessment. The user has
several options.

3.1.Option 1.

The users can use the unbalanced forces from step 2 and apply these results in his pipe stress
program. By means of the pipe stress program a dynamic mechanical analysis is performed. The
dynamic analysis consists of two stages.

3.1.1.First stage.

The first stage is the modal analysis. Similar to the acoustical natural frequency calculation
performed in step 1 above the purpose of the modal analysis is to verify the separation between
mechanical natural frequencies and the compressor harmonics. From the analysis performed in step
2 it is clear that the compressor will produce shaking forces not only at the basic compressor
frequency but also at the higher harmonics. In order to avoid mechanical resonance sufficient
separation has to exist between mechanical natural frequencies and the compressor harmonics.
Separation of acoustic and mechanical frequencies should be maintained on the same order as
separation of acoustic frequencies, i.e. about 20%, with less for higher frequencies. “Separation” is
a qualitative measure. The user may prefer less separation, but a more secure supporting structure.
Lower pulsation levels often mean lower efficiencies, and so higher levels of pulsation are often
tolerated, and the designer must be able to accommodate these loads. The “separation” tradeoff is a
“design-level” engineering decision. In the higher frequency range the peaks are typically sharper
and less separation is required. The dynamic load factor (DLF) can also be used as a measure of
separation. A DLF of 2.5 is often used in the less than 100Hz. range and a DLF of 5.0 in the
frequency range over 100 Hz. DLF = ABS( 1 / [ 1- (f/fn)2 ] ). A good separation between
mechanical natural frequencies and compressor harmonics implies elimination of large mechanical
acoustic loads. The mechanical natural frequencies depend on the support stiffness. For an accurate
analysis it is important to incorporate the support structure or it’s stiffness into the model. See for

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.2.35


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

instance Figure 36 below for the model and structure used in the example problem. The plot is
taken from Caesar II.

Figure 36. Model of piping system including structures ready for mechanical analysis.

If natural frequencies and compressor harmonics are insufficiently separated, additional support
stiffness must be created. For instance in the support structure shown above additional bracing and
beams can be applied to increase the stiffness. Compare the original structure in Figure 36, to the
improved structure in Figure 37 below. In the example case the natural frequencies obtained after
reinforcement of the structure (figure 37) showed sufficient separation with the compressor
harmonics to continue to the second stage of the mechanical response.

Figure 37. Reinforced Support Structure.

3.1.2.Second stage.
In the second stage of the mechanical response analysis the shaking forces from step 2 of the
acoustical analysis are used to perform a standard harmonic mechanical analysis. The resulting pipe
and structural stresses are compared to the fatigue curves (preferably the endurance limits) for the
© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.2.36
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

applicable materials. Since BOS Fluids produces three dimensional plots of the unbalanced force
distribution it is typically easy for the user to inspect those plots and decide which of the force
components should be used in the mechanical harmonic analysis. The procedure is as follows:

1) Inspect the BOS Fluids 3D Plot of the Unbalanced Forces.


2) Select which of the forces should be applied in the harmonic dynamic analysis.
3) Select the load and the axial pipe direction for the load. The load magnitude can be read
from the plot or taken from the tabular BOS Fluids listing. Be sure that the loads are meaningful.
Some geometries, (closed valves), can produce unrealistic unbalanced loads in piping systems. The
loads are typically easy to identify because they are reflective of a main flow path in the
compressor piping system.
4) Apply the loads and directions into the mechanical piping analysis. Apply each load in a
direction that will produce the overall worst net effect on the system. If the user is unsure which
direction this is, then the run should be repeated with individual load directions reversed and the
worst result used. Each load should be applied simultaneously and the harmonic variation applied
without phase shifting between loads.
5.) The resulting dynamic stress from most pipe stress programs that have harmonic
simulators will be the amplitude of the B31 Code stress and can be compared directly to the value
of Sa from ASME Section VIII Division 2, Appendix 5. (This is because the peak stress from the
B31 Appendix D “i” factor calculation is half of the actual peak stress value, therefore the resulting
stress can be compared to the alternating allowable from Appendix 5. For carbon steel, the allowed
B31 harmonic alternating stress should be compared to 13,000 psi. (90 MPa).

3.2.Option 2
Option 2 is similar to option 1 however this time the BOS Fluids model is exported to FE-pipe and
the structural cycle as described above is performed by means of FE-Pipe. In order to export the
model to FE-Pipe the user enters the interface option in the BOS Fluids main-menu (see figure 38)

Figure 38. Option Interfaces.

Upon selection of a suitable transfer file name the interface “OUTPUT” option is activated. (See
figure 39 below.)

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.2.37


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Figure 39. Transfer parameters.

Upon activation of the export button an export file is written. In the next step the exported file is
read in into FE/Pipe. For that purpose FE/Pipe is started using a new jobname. (In the example the
name “mechani2” is used.) The beam element model is selected (see figure 40)

Figure 40. The beam type modeler.

Once the Beam Model menu appears the interface option is activated (See figure 41) and the
transfer filename is entered (this time the source is “INPUT”). The model is read by activating the
“import key.” See figure 41. Once the model has been imported the material properties must be
entered as they are not a part of the BOS Fluids analysis. The original model was a BOS Fluids
model that did not contain material details or supports. The completion of the material input must
only be done once on the first element form. FE/Pipe performs the downward copying of the data to
all subsequent elements. The user must only change the material data if it is different on any
subsequent elements after the first.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.2.38


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Figure 41. FE-Pipe Beam-model interface screen.

The BOS Fluids model only contains pipe elements. So the FE/Pipe model initially will only
contains pipe elements too. For many compressor installations an accurate dynamic result can only
be obtained if the effect of the support structures is considered. This means that either the support
structure mass and stiffness has to be incorporated in the pipe/beam model restraint functions or the
support elements have to be incorporated fully. The FE/Pipe beam modeler allows direct addition
of structural beam elements into the piping file. (See figure 42 below.)

Figure 42. FE/Pipe Beam-Type Pipe Model with Structural Beam Elements.

In the execution of the mechanical response analysis by means of FE/Pipe the same stages are
followed as for the pipe stress package (see option 1 above). In the figures 43a & b samples of the
results obtained during the modal analysis are shown. Examination of the deformed shapes is
essential for determination of the appropriate reinforcement if required.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.2.39


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Figure 43a. Beam model, piping + structural elements one of the mode shapes.

Figure 43b. Beam model, piping + structural elements one of the mode shapes.

Step 4. Measurements
Although vibration measurements are not a part of API 618 in many cases it is a useful and
instructive tool in the analysis of compressor vibration problems. For instance for the example used
in this example vibration measurements have been performed. The results of these measurements
are included in Figure 44 for information.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.2.40


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

100. 00

90. 00

80. 00

70. 00

60. 00

50. 00

40. 00

30. 00

20. 00

10. 00

0. 00
0. 0 10. 0 20. 0 30. 0 40. 0 50. 0 60. 0 70. 0 80. 0 90. 0 100. 0

F r e q u e n c y ( Hz )

Figure 44. Vibration measurements for Node 160.

In Figure 44 the vibration measurements at node 160 of the suction piping system analyzed in this
example are shown. Also shown (the dotted lines) are the mechanical natural frequencies
calculated by means of the pipe stress program Caesar II. High peaks in the measurements indicate
high vibration levels. When high vibration levels are recorded in a field compressor installation
they are typically caused by one of the following sources.

a. Mechanical “near” resonance condition.


b. Large unbalanced pulsation forces.

Large shaking forces can be caused by:

a. Large pulsation levels


b. Acoustical “near” resonance condition.

Inspection of the location of the peaks from low to high frequency in Figure 44 is instructive.

1. The first peak at around 16 Hz. does not exactly coincide with one of the mechanical natural
frequencies. The applicable frequency is however recognized as the compressor basic
harmonic frequency. Two of the calculated mechanical natural frequencies are close. The
relatively high vibration levels can be the result of small differences between the
mechanical model and the real world system. Either stiffness, mass, or both may be slightly
in error. A small shift in one of the natural frequencies could create resonance. The
sensitivity of the mechanical results for small stiffness and mass variation is to be analyzed
more closely. Additionally the relatively high vibration levels may be the result of high
pulsation forces which could be the result of acoustical resonances. The effect could be
analyzed in more depth and a statement of the acceptability of the vibration level can be
made based on the result of that analysis.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.2.41


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

2. The second peak in figure 44 is at about 33 Hz and indeed coincides with a mechanical
natural frequency of the system. This problem can be solved by shifting the mechanical
natural frequencies. Reinforcement (stiffening of the support) is used here. The same
applies for the peak at 72 Hz and 80 Hz.

3. The peaks at about 48 and 64 Hz are isolated from mechanical natural frequencies and must
be the result of high acoustical pulsation forces. If the stresses corresponding to these levels
are not acceptable installation of additional damping could be considered.

Additional details regarding compressor modeling and API 618 allowables can be found in Chapter
4, Section 2.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.2.42


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

API 618 Compliance for Reciprocating Compressors - Summary


The following outline is recommended for satisfying API 618:

Step 1:
Generate a mechanical model of the suction and discharge piping systems in either CAESAR II or
FE/Pipe. Include only the pipe at this time and do not include any supports or varying material
properties or temperature.

Step 2:
Write a neutral file of the piping geometries that can be read into BOS Fluids.

Step 3:
Read the pipe geometries into BOS Fluids and perform an acoustic natural frequency study as
described in Chapter 4 Section 1. Make sure that there is adequate separation between the fundamental
running speed of the compressor and its harmonics and the acoustic natural frequencies of the system.

Make sure that Maximum Pressure Trapping is deactivated on the Optional Form
and turn off Column Separation for all BOS Runs.
Step 4:
Perform a sensitivity study of the Step 3 solution varying the properties of the gas based on the
potential high and low gas temperatures for both the suction and discharge systems. Make sure that
there is adequate separation between the excitation frequencies and the acoustic natural frequencies of
the piping systems. If there is not adequate separation then either the pipe must be rerouted, or orifices
should be installed in the system.

Step 5:
Build a suction side model of the compressor and a discharge side model of the compressor as outlined
in Chapter 4 Section 1 or as described in below. Run each model for the variety of gas temperatures
anticipated.

Step 6:
Once the compressor models are run perform a discrete Fourier analysis of the compressor suction and
discharge waveforms at a convenient point on the piping system, usually at the entrance to the bottle
on the compressor side. The Discrete Fourier Analysis (DFA) is performed by starting the
program under the Miscellaneous BOS Fluids options. Point this program to both the suction and
discharge results as shown below and perform the discrete Fourier analysis of the compressor’s
interaction with the piping system on both the suction and discharge sides. (Each side is analyzed
separately.) A typical result from a suction side analysis is shown below:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.3.1


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The text window in the bottom left of the form is a standard Rich Text edit window. The mouse and
cursor can be used to select text or to scroll the window. The Harmonic frequencies and flow
amplitudes computed from the compressor model file results are generally the output of interest. This
data can be highlighted in the window and the key sequence <Ctrl><C> used to copy the highlighted
text to the system clipboard. The information generally wanted is shown below.
BOS Fluids 4.109 Reciprocating Compressor
Discrete Fourier Frequency Analyzer

Jobname = test002 7
Rpm used = 960.0000
Number of Harmonics = 10
Node Number = 50
Output Time (sec) = 1.0010010E-03

Number of Elements = 13
Number of Time Steps = 999

Nyquist Frequency = 0.5000000


Nyquist Frequency Index = 1

Compressor First Harmonic Period (sec) = 6.2500000E-02


Solution Output Time Step (sec) = 9.9999990E-05

Decomposition
Rotation (deg.) = 0.0000
Flow Constant Amplitude (L^3) = 60.9100

Harmonic# Frequency(cps) Flow (L^3) Phase(Deg)

1 16.00 0.221157372 -76.0178


2 32.00 0.177808750 105.0337
3 48.00 0.346948244 -30.0415
4 64.00 0.214652484 106.5193
5 80.00 0.209381604 -122.0343
6 96.00 0.445300389 -148.2110
7 112.00 0.329254734 178.3065
8 128.00 0.032132370 69.1450

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.3.2


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

9 144.00 0.204642760 -64.4362


10 160.00 0.192430549 124.3279

Step 7:
The 10 harmonic frequencies from the compressor suction side model should be imposed on the
suction piping system and the 10 harmonic frequencies from the compressor discharge side model
should be imposed on the discharge piping system. A single BOS Fluids run must be made with each
frequency, for each model. This is true because the API 618 allowables give the permitted peak-to-
peak pressure pulsations on a per frequency basis.

The procedure is not so tedious because any violation of the API allowables must be resolved on a per
frequency basis anyway, usually by the addition of orifices. The recommended steps are given below:

API Pulsation Check Procedure:


1) Establish a boundary condition at the compressor bottle inlet or outlet flange that will provide harmonic excitation
to the suction or discharge pipework. (Do one at a time.) Set the frequency of the excitation to the first frequency to be
analyzed. Use the mean gas flow and the percentage of the mean gas flow that will give the flow amplitude for each
harmonic being analyzed. The Discrete Fourier Analysis gives the amplitude of the flowrate.
2) Calculate the solution output parameters. The maximum duration of the transient should be found from (1/f1)(40)
where f1 is the running speed of the compressor, i.e. RPM/60. The time step should be found from: (1/f10)(0.1), where
f10 is the highest frequency of interest. (The intent here is to establish a set of solution parameters that can be used for
each frequency of excitation without the need to be changed.) Set the time “window” for output processing to be from
(1/f1)(36) –to– (1/f1)(40). These are general guidelines that may be used as a starting point for conducting the
analysis.
3) Run the first transient analysis. Inspect the pressure output at several locations and be sure that the response in the
output window selected does not contain visible remnants of a transient. The wave form should be clearly periodic
and should not show any dissipating elements.

4) Run the API 618 allowables program: from the BOS Fluids Miscellaneous Button. Be sure that the
working directory that contains the just completed analysis files appears on the top of the PRGMAPs screen before
hitting the BOS Fluids Miscellaneous button. A blank screen appears when the program is first started as shown
below:

Generally, the user will want the program to make the comparisons to API 618 allowables automatically. To do this,
the following procedure should be followed:
1) Select the Units used in the analysis.
2) Enter the Compressor RPM and the number of harmonics used in the Discrete Fourier Analysis, (usually 10).

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.3.3


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

3) Check the “Check File Output” box.


4) Enter the BOS Fluids filename. (No suffixes.)
5) Enter the harmonic analyzed in the BOS Fluids file. (This should be a number between 1 and the total
number of discrete Fourier components – but should be the harmonic associated with the anlysis just run.)
6) Click on the light green API 618 Check button. A screen and plot similar to the following should be
produced if the current subdirectory is set correctly in the PRGMAPs screen.

The text window in the middle of the screen is scrollable. The user can put the mouse in the window and then scroll
the window using the cursor keys. The shift key can be used with the scrolling to highlight parts of the text.
<Ctrl><C> will copy selected parts of the text to the system clipboard. This text can then be pasted into another
report.
The API 618 comparsion for pressure pulsation is done for each reach in the model but only the largest pulsation
magnitude is printed. The words ** OVER ** are printed if the pulsation exceeds the API 618 allowables. The BOS
Fluids node numbers associated with each section of pipe are printed first on the left. (See the text printout above.)
The plot on the right shows the allowable peak-to-peak pulsation magnitude for each diameter plotted as a function of
frequency. This plot is for information purposes.
A file is written by BOS Fluids that contains the pulsation data for processing. The file name is <name>.API. The
interested reader can view this file using Notepad and see the type of data available. The file contains the following
columns:
1) Distance from the “from” node down the pipe.
2) The average pressure
3) The maximum pressure
4) The minimum pressure
5) The maximum flowrate
6) The minimum flowrate
7) The diameter of the section.

Be sure that any high pulsations are in valid system components and are not in artificial components used to model
compressor parts, etc. If pulsations exceed the allowables then the user must decide if orifices should be added or if
the pipe routing should be changed. Changes in pipe routing are less expensive in the long term because they don’t
effect the day-to-day running of the installation. Adding orifices does not affect the line routing but does affect the
day-to-day efficiency of the compressor operation because of the added pressure drop. The user must decide which
approach is the most cost-effective. Changes in the line routing require the re-running of all prior
analyses. The addition of orifices does NOT require that any earlier analyses be remade.
If orifices are added the user should inspect the 3D plotted results so that the maximum velocity in the system can be
spotted. This is typically the best place to add an orifice. The guidelines in ISO 5167-1 can be followed to select the

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.3.4


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

orifice size and to make accurate calculations for flow coefficients. Once the orifice is placed in the system the
harmonic analysis should be rerun and the new pulsation levels checked. If the pressures are still over allowables then
the approach described above should be repeated.
5) Repeat the procedure for each frequency in the discrete Fourier analysis for both the discharge and suction sides of
the piping system. If any piping is rerouted to minimize pulsations, then the API Pulsation checking procedure should
be rerun for all frequency ranges.

The allowable peak-to-peak pulsation levels are plotted for each line size for each frequency in the
allowables plot. If the user does not want to use the output from a data file, but would rather check the
pulsation levels by hand, the user can enter the mean line pressure and enter the pipe inside diameters
that are to be checked. The API 618 peak-to-peak pulsation allowable is based on the line size and the
discrete frequency of interest. An example output from a manual API 618 allowables report is shown
below.

API 618 Peak-to-Peak Pressure Allowables

(Units used - SI)


(Pressure in bars, length in mm.)

Inside Diameter:101.6

16. Hz. Allowed P(p-p) = .5061


32. Hz. Allowed P(p-p) = .3578
48. Hz. Allowed P(p-p) = .2922
64. Hz. Allowed P(p-p) = .253
80. Hz. Allowed P(p-p) = .2263
96. Hz. Allowed P(p-p) = .2066
112. Hz. Allowed P(p-p) = .1913
128. Hz. Allowed P(p-p) = .1789
144. Hz. Allowed P(p-p) = .1687
160. Hz. Allowed P(p-p) = .16

Inside Diameter:156.44

16. Hz. Allowed P(p-p) = .4078


32. Hz. Allowed P(p-p) = .2884
48. Hz. Allowed P(p-p) = .2355
64. Hz. Allowed P(p-p) = .2039
80. Hz. Allowed P(p-p) = .1824
96. Hz. Allowed P(p-p) = .1665
112. Hz. Allowed P(p-p) = .1541
128. Hz. Allowed P(p-p) = .1442
144. Hz. Allowed P(p-p) = .1359
160. Hz. Allowed P(p-p) = .129

Step 8
After the pressure pulsations are within API allowables for all frequencies for both the suction and
discharge piping sides of the compressor, the mechanical pulsation study should be conducted. There
are typically two steps to the mechanical analysis:

Step 1: Mechanical and Acoustic Frequency Separation


Step 2: Harmonic Mechanical Analysis at Discrete Fourier Frequencies.

Step 1: Mechanical and Acoustic Frequency Separation


The step involves performing a modal analysis of the piping and structural support system. The
purpose of the modal analysis is to verify the separation between mechanical natural frequencies
© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.3.5
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

and the compressor harmonics. The compressor will produce shaking forces at its fundamental
running speed but also at the higher harmonics. In order to avoid mechanical resonance sufficient
separation has to exist between mechanical natural frequencies and the compressor harmonics.
Usually separation of mechanical modes is done for the major acoustic modes that are identified
by the acoustic analyses just performed. Separation of acoustic and mechanical frequencies should
be maintained on the same order as separation of acoustic frequencies, i.e. about 20%, with less
for higher frequencies. “Separation” is a qualitative measure. The user may prefer less separation,
but a more secure supporting structure. Lower pulsation levels often mean lower efficiencies, and
so higher levels of pulsation are often tolerated, and the designer must be able to accommodate
these loads. The “separation” tradeoff is a “design-level” engineering decision. In the higher
frequency range the peaks are typically sharper and less separation is required. The dynamic load
factor (DLF) can also be used as a measure of separation. A DLF of 2.5 is often used in the less
than 100Hz. range and a DLF of 5.0 in the frequency range over 100 Hz. DLF = ABS( 1 / [ 1-
(w/wn)2 ] ). A good separation between mechanical natural frequencies and compressor harmonics
implies elimination of large mechanical acoustic loads. The mechanical natural frequencies
depend on the support stiffness. For an accurate analysis it is important to incorporate the support
structure or it’s stiffness into the model.

If natural frequencies and compressor harmonics are insufficiently separated, additional support
stiffness must be created. The majority of compressor/piping vibration problems are solved by
adding structural stiffness or supports to a system that is too flexible. The analyst must also be
sure not to produce thermal overload problems when adding restraints to piping, and must make
sure that the supports added are suitably designed for vibration-type loading.

Step 2: Harmonic Mechanical Analysis at Discrete Fourier Frequencies.


The second step of the mechanical response analysis applies the shaking forces from the discrete
Fourier acoustical analysis to the mechanical piping/structural system. The calculated pipe and
structural stresses are compared to the fatigue curves (preferably the endurance limits) for the
applicable materials. Since BOS Fluids produces three dimensional plots of the unbalanced force
distribution it is typically easy for the user to inspect those plots and decide which of the force
components should be used in the mechanical harmonic analysis. The procedure is as follows:

1) Inspect the BOS Fluids 3D Plot of the Unbalanced Forces.


2) Select which of the forces should be applied in the harmonic dynamic analysis.
3) Select the load and the axial pipe direction for the load. The load magnitude can be read
from the plot or taken from the tabular BOS Fluids listing. Be sure that the loads are meaningful.
Some geometries, (closed valves), can produce unrealistic unbalanced loads in piping systems.
The loads are typically easy to identify because they are reflective of a main flow path in the
compressor piping system.
4) Apply the loads and directions into the mechanical piping analysis. Apply each load in a
direction that will produce the overall worst net effect on the system. If the user is unsure which
direction this is, then the run should be repeated with individual load directions reversed and the
worst result used. Each load should be applied simultaneously and the harmonic variation applied
without phase shifting between loads.
5.) The resulting dynamic stress from most pipe stress programs that have harmonic
simulators will be the amplitude of the B31 Code stress and can be compared directly to the value
of Sa from ASME Section VIII Division 2, Appendix 5. (This is because the peak stress from the
B31 Appendix D “i” factor calculation is half of the actual peak stress value, therefore the

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.3.6


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

resulting stress can be compared to the alternating allowable from Appendix 5. For carbon steel,
the allowed B31 harmonic alternating stress should be compared to 13,000 psi. (90 MPa).

Reciprocating Compressor Modeling


There are typically three different models of compressors that may be used in BOS Fluids. They range
in difficulty and accuracy, the easiest model to build being the least accurate. The three model types
can be described as:

1) Base Compressor Model – Fixed Harmonic Flow, Undamped Check Valve


2) Boundary Condition Reciprocating Model – RECIP Boundary Type
3) Complex Compressor Model – Fixed Harmonic Flow, Damped Check Valve and Cylinder Model

Base Compressor Model:


The basic compressor is the model used in the example in Chapter 4 Section 1, and involves a cylinder
and undamped check valve model. This model is probably the easiest to build and requires the least
amount of data.

Boundary Condition Reciprocating Model


The boundary condition model uses a RECIP boundary condition to simulate the cylinder, valve inlet
and valve. The BOS Fluids boundary condition input screen is shown below:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.3.7


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The RECIP box contains 10 items for compressors going from left to right: Each dimensional unit is
given in metric units.

1 – (1) Reciprocating Compressor instead of Pump


2 – (1) for Suction (2) for Discharge
3 – Stroke (m.)
4 – Piston Area (m2)
5 – RPM
6 – Cylinder Dead Space (m3)
7 – Phase Shift (deg)
8 – Cp/Cv
9 – Compression Ratio (Pdischarge / Psuction)
10- Double Acting Cylinder Factor

The flowrate at 40 (the bottle inlet) from the two compressor models above are shown below:

Flowrate @40 RECIP Model Flowrate @40 Base Model

Several observations may be made comparing the two plots:


1) The fundamental waveform is the same in both cases. The models show different signs because the
basic construction around node 40 is different. They both, however represent a suction side compressor
nozzle.
2) The “base model” waveform is more irregular than the RECIP boundary condition model. This is
because the undamped check valve is interacting with the small elements and volume changes inherent
in the “base model.” There is not sufficient detail in the “base model” to assume that the local,
irregular peaks and valleys are due to actual compressor behavior, although actual valve inlet and
outlet port local pressures will probably look more like the “base model” shape than the RECIP shape.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.3.8


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

3) The RECIP shape assumes that the pulsations external to the cylinder do not affect the compressor
behavior. RECIP includes the effect of the cylinder dead space and compression ratio, and more
accurately represents the isentropic pressure-volume relationship seen on a typical compressor card.

General Notes for the RECIP model:


1) Different compressor parameters can be given for both suction and discharge sides to include the
possible effects of clearance pockets, etc.
2) The double acting cylinder factor, if greater than zero is a multiplier on the second suction or
discharge pulse in the same crank rotation period. This is used to simulate the effect of double acting
cylinders and valves.
3) Any number of RECIP boundary conditions can be used to simulate multiple stages on the same
machine (common shaft) shifted by a phase angle.

The general RECIP piston displacement model is shown below.

The volume above the piston (between the top of the piston and the valve ports), is equal to:

V = Vo + AS/2 – AS/2 sin( 2πft ) Eq. 4.2.1

where: V = Volume above piston


Vo = Cylinder Dead Space
A = Piston Area
S = Stroke
f = RPM/60
t = time

Isentropic compression and expansion are good approximations for relatively fast compression and so
for the discharge zone:
PdVdk = PsVsk Eq. 4.2.2

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.3.9


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

To find the opening time for the discharge valve Equation 4.2.2 is arranged to find Vd, and then (t) is
found by inserting Vd for (V) in Equation 4.2.1. To find the opening time for the suction valve,
Equation 4.2.2 is arranged to find Vs, and then (t) is found by inserting Vs for (V) in Equation 4.2.1.

where: Pd = Discharge Pressure


Vd = Discharge Gas Volume
Ps = Suction Pressure
Vs = Suction Volume
k = Ratio of specific heats

Pd, Vd , Ps and Vs can be found on the compressor card:

The flowrates at the opening times can be found by differentiating Equation 4.2.1 with respect to time
and then inserting the time sought in the resulting expression for dV/dt.

Complex Compressor Modeling


Complex compressor modeling involves the use of inline surge bottles to include the local effect of gas
compression in the cylinder space and in the flow passages and collection manifolds. Users can employ
the damped check valve model to include the effect of the valve mass and spring stiffness, and
damping effects. The user of this model must have detailed information covering the mechanical
characteristics of the compressor, flow passages and valves. The schematic below is intended to
provide basic guidance.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.3.10


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.3.11


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

BOS Fluids API 674 Compliance (Reciprocating Pumps)


Most problems with reciprocating equipment occur because the major running speed, or an
harmonic of the major running speed is coincident with an acoustic or mechanical natural
frequency. API 618 for compressors and API 674 for reciprocating pumps are intended to control
pulsation levels such that resonance and stress problems do not occur. It is the user’s
responsibility however, to investigate all potentially dangerous acoustic and mechanical
resonances. It is a detective’s game, and not all field problems can be discovered at the design
stage. The designer’s goal, however, is to prevent obvious problems associated with the
fundamental pulsation frequencies, and to include sufficiently complex mechanical models so that
side-to-side piping loads not normally experienced can be tolerated without excessive
displacements.

Compliance with API 674 for Positive Displacement Reciprocating Pumps is similar to
compliance with API 618 for compressors. BOS Fluids generates the API 674 analysis
automatically in exactly the same way it does for the API 618 analysis. The user can also analyze
each harmonic separately and this approach is described below as it is the most informative.

The BOS Fluids API 674 user is encouraged to review the Chapter 4 Sections 1 and 2 for the most
straightforward method to produce API compliance.

Pulsations can cause operational problems with both pumps and compressors, but suction side
pulsation problems can cause cavitation in pumps and bring gas out of solution further
complicating pulsation and cavitation problems. Suction side pulsations should not produce
pressures below the NPSH. There should be adequate separation between the lowest pulsating
pressure and the fluid vapor pressure. Even though this is checked on a frequency basis in API 674
the user should also be careful that the proper phasing of frequency responses can add effects.

API 674 Compliance requires a discrete Fourier analysis of the suction or discharge waveform.
Pump waveforms are often less sensitive to piping configurations around the machine and so a
discrete Fourier analysis can be performed on the suction or discharge side flow signals without
including any attached piping or manifolds. For higher pressures, however, and where a manifold
collects the output from each cylinder, the triplex pump acts more like three single reciprocating
pumps, each, phase shifted by 120 degrees. In higher pressure applications the discharge
waveforms may have considerable high frequency content which necessitates a more accurate
construction of the pump model. If desired, the user can build single or multiple pump models
with common manifolds and suppression devices and perform the discrete Fourier analysis on the
BOS Fluids output signal at any point that can be considered an “input” to the system under study.

The reciprocating pump model will either evaluate a waveform from a BOS Fluids calculation or
will generate an “n”-plex waveform and perform the harmonic analysis of the collected signal.
Where the distance between the valve and manifold exit nozzle is not significant, or the
frequencies associated with these lengths are too high to be of concern the waveform model can be
used. Where higher frequencies can cause potential problems the user is advised to build the most
comprehensive model possible – and this would be the phase shifted models. These approaches
are illustrated by reviewing the schematic figure of the triplex pump below. For large, high
pressure units the suction and discharge manifolds might be quite large and removed from the

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.4.1


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

main pump body. In these cases, the acoustic frequency associated with the distance between the
valve and the end of the manifold can produce high acoustic pulsation resonances. For smaller,
lower pressure units, the manifold is part of the pump body and the acoustic length between the
valves and the manifold closure can be neglected. In the figure below, if the acoustic frequency
associated with the manifold should be studied, then the pump discharge should be at “B.” The
user should model three single pumps with a phase shift of 120 degrees. If the manifold is small,
or if the user and manufacturer already know that the manifold-to-valve acoustic periods are not
coincident, then the pump discharge at point “A” can be the first point studied, and the acoustics
between A and B ignored. (Note that accumulators can be placed almost anywhere in the
discharge or suction system, and these should definitely be modeled to simulate the proper
behavior on the side of the accumulator away from the pump. Accumulators in BOS Fluids should
be modeled as surge vessels, as an accumulator is little more than a gas charged volume.)

The user is encouraged to review the API 618 compliance sections described in Chapter 4,
Sections 1 and 2. There are similarities between the methods that can help create an understanding
of the solution.

The RECIP Boundary condition is used in API 618 to generate the output interrupted sawtooth for
the compressor suction or discharge sides and and is used in API 674 to generate the peculiar
output for a multi-cylinder reciprocating pump. The output time waveform for a triplex pump is
shown in the sketch below for example:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.4.2


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Both the RECIP boundary condition and the Reciprocating Pump module produce
exactly the same time waveforms.

Several types of reciprocating pumps are shown below and were taken from the different
manufacturer’s or vendor’s sites on the Internet.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.4.3


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.4.4


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The two major solution approaches are shown in Figure 4.3-1 below.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.4.5


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Figure 4.3-1

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.4.6


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The user of high pressure systems can also evaluate the mechanical natural frequency of the valves
and systems either by simple one degree of freedom models, or by constructing more complex
mechanical models. In most high pressure cases it is important that the acoustic natural
frequencies associated with the valve to manifold reflection lengths are not coincident. The valve
models can also be constructed using check valves. The open/close spring-loaded ball-type check
valves used with many reciprocating pumps can be easily modeled using the damped check valve
model in BOS Fluids.

Data input for the reciprocating pump models in BOS Fluids requires the radius of the crank arm
and the con rod length as shown in the figure below.

If these dimensions are not know, the radius of the crank arm can be taken as half of the stroke,
and the con rod length can be assumed to be five times the con rod length.

The volume coefficient is a value greater than 1.0 that reflects the volume in the cylinder between
the valves. The relationship is:

Volume of fluid between valves = (volume coefficient)(stroke)(piston area)

The volume coefficient is found when the piston is as far into the cylinder as it will go, i.e. at the
furthest extent of the discharge stroke. Volume coefficients of 2-to-4 are reasonable.

The remainder of input for reciprocating pumps is self-explanatory. The user must be careful to
input the dimensions in metric units. The units are given in the help. The pressure difference
required is given in meters of fluid head. The user can find this from the equation below:

Meters of fluid head = (43.89) (pressure difference in psi) / (density in lb/cu.ft)

Wylie recommends, “With low discharge pressures, for example, the triplex flow is well
represented by 4 to 8 terms in a harmonic series, but with high pressures as many as 24 harmonics
are needed to describe the flow.”

Now the user must decide which of the modeling approaches is most suited to his particular
situation. If the frequencies around the machine are not so important, and the vibration of the pipe
is more important, then the leftmost path in Figure 4.3-1 should be used. If the vibration of the
pump and the manifolds are important then the more detailed model shown on the right side of
Figure 4.3-1 should be used. The user can start with the simple model and add complexity as

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.4.7


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

problems are addressed. In this case the RECIP boundary condition would be used to produce the
pulsation flow waveform and the harmonic components. Each of the harmonic components would
be used with a HARMONICFLOW boundary condition at the exit of the manifold and API 674
checked for each harmonic.

Once API 674 is satisfied for each harmonic then the user can optionally go back and construct the
individual cylinder models, again using the RECIP boundary condition but for a single cylinder.
All three cylinders would be modeled using three RECIP single cylinder boundary conditions
shifted from each other by 120 degrees. Once the pressure waveforms are generated the user
would due a discrete Fourier pressure decomposition for any point of concern in the manifold
system and compare the peak-to-peak pressure values to the allowables from API 674. This is a
more complicated procedure to follow, but can easily be done using the tools available in BOS
Fluids and from the BOS Miscellaneous menus.

Note that nitrogen charged accumulators, or “pulsation stabilizers” are placed at the suction and
discharge manifolds as close as possible to the machines. These bottles are typically charged
when the system is not pressurized to a value selected by the manufacturer or user. The
accumulators should be entered in BOS Fluids using the Surge Vessel model. A surge vessel input
form is shown below:

The surge vessel node should be attached to the branch pipe end that extends from the header pipe
as shown in the figure below:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.4.8


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Effective Type A Bladder


Accumulators can take a variety of shapes:

The two general types of accumulators are shown above as A and B. The Type A accumulator
liquid level varies as the bladder is compressed. The Type B accumulator liquid level stays
essentially constant as the bladder is compressed since the bladder sides move inward to
accommodate the increasing pressure. The type B accumulator can also be designed to absorb
greater pressure swings by over pressurizing the bladder. The approach taken to design each is
given below:

Type A Modeling
The bladder is initially charged to a pressure a little lower than the minimum working pressure. At
the minimum working pressure there should be a small amount of liquid in the accumulator. The
maximum allowable working pressure to the gas precharge pressure is typically about 4:1. The
recommended precharge pressures are between 60-to-90% of the mean line pressure. The

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.4.9


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

precharge pressure will completely fill the bottle volume. From the steady state analysis the user
will know what the steady-state pressure will be at the bottle inlet. The peak-to-peak pressure (P-
P) allowed per API 674 is 100 / [ (ID)(f) ]0.5 ( in psi) or 3500 / [(ID)(f) ]0.5 (in KPa). Select the
precharge pressure (Po) as (0.8)[Pm-(P-P)/2]. Make sure that the bladder won’t completely fill the
accumulator on the low pressure side. The liquid level then will be found from Vm where:

PoVok = PmVmk

where Po – is the precharge pressure,


Vo – is the total volume of the bottle
Pm – is the steady state pressure found from the BOS Fluids run
Vm – is the volume of the bladder when the system is operating in steady state conditions.

The height of fluid can then be approximated from:

HL = (Vo – Vm ) / A

where Vm = gas volume input to BOS Fluids = Vo(Po/Pm)1/k


A = 0.785 d2
d = inside diameter of the accumulator
HL = liquid height

The liquid height in the accumulator (surge vessel) is plotted as an output quantity, and the user
should check this to be sure that the accumulator does not “bottom out” during pulsations as this
will significantly shorten the life of the accumulator.

When using in systems above 3000 psi, some adjustment to the polytropic expressions should be
employed as general effect is to reduce the effective volume.

Type B Modeling
The type B model is intended to keep the fluctuation in head small, (although this should not be a
major contributor to the accumulators response to the system.) These accumulators can have initial
pressure charges greater than the mean line pressure if the surge event requires this type of
behavior. The diameter of the “effective” surge bottle should be entered as:

d = [ (4)( BL – Bd ) Bd ]0.5

where d = inside diameter of the “effective” surge bottle.


BL = Bladder length
Bd = Bladder inside diameter.

The liquid height should be entered as:

HL = BL / 2

The entered gas volume should be

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.4.10


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Vm = (pm/pc)1/k Vo

where pm = mean line pressure,


pc = charge pressure (often = pm or may be greater than pm.)
Vo = (BL)(0.785)Bd2

The gas pressure should be left blank for the surge vessel modeling for both
Type A and Type B accumulators. The gas (air) pressure input is only used for
blowdown simulations.

Both Type A and Type B accumulators should be described with VERT


orientations. In this case the total volume of the accumulator is not used.
The input for a reciprocating pump from the Wylie and Streeter example is given below:

Pump Speed (RPM) = 520.


Piston Area (m2) = 0.002564
Radius of Crank Arm (m) = 0.0381
Length of Con Rod(m) = 0.1968
Volume Coefficient = 2.0000
Average Pump Suction to
Discharge Pressure Drop (m) = 87782.4
Bulk Modulus = 206.8E7
Time Steps per Period = 65
Number of Harmonics = 5
Pump # of Cylinders = 3

A typical reciprocating triplex pump RECIP boundary condition output is shown above. Notice
how the waveform is truncated during the first period to help attenuate startup transients.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering 4.4.11


BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Reciprocating Equipment Notes (Contributed by Users & from Referenced


Papers)
Reciprocating Compressors are utilized in all manufacturing industries. Because these machines are
capable of providing high pressure along with variable loading, they are favored for many gas process
applications. The total quantity of positive displacement reciprocating engines, pumps and
compressors far exceed the number of centrifugal units.

Past studies within the Hydrocarbon Processing Industry (HPI) indicate that the maintenance costs for
reciprocating equipment are approximately 3.5 times that of centrifugal equipment. Substantial
savings in maintenance costs and an increase in run time may be achieved through basic monitoring
of basic vibration levels.

Frame Vibration

The most important vibration parameter of a successful monitoring program is Frame Vibration. When properly
applied, monitoring Frame Vibration will help prevent catastrophic failures. In the event of a failure, the damage to
a reciprocating machine can be reduced.

For the greatest benefit, a Frame Vibration Monitoring system should be wired to an automatic machine trip.

Rod Drop

The vast majority of Reciprocating Compressors are designed with horizontal Cylinders and Pistons. This is
primarily due to foundation requirements and the popularity of opposed-balanced machine designs.

The force of gravity causes the Piston to "RIDE" more in the bottom of the Cylinder than in the top. In turn, this
causes the Piston to wear more in the "DOWN" direction. Machine manufactures provide wear or rider rings to
provide a replaceable wearing surface. For lubricated Cylinders, glass embedded Teflon may be used. For non-
lubricated Cylinders, Teflon may be used.

The wear or rider rings are allowed to wear sacrificially. They are rotated or replaced before damage to the
Cylinder lining occurs. There are several methods used to determine when to replace or rotate the rings. One
method is to operate a new machine for a given number of hours or days. Then a valve is removed, and the wear
is measured by using a feeler gauge. A calculation is then performed with this information. The results determine
the length of time the machine can be safely operated with periodic inspections of the rings. Obviously, this is a
very frustrating method of performing preventative maintenance.

Currently, one popular safety device for detecting Rod Drop is a unit mounted under the rod at a gap determined
by the allowable wear of the wear ring. When the rod contacts the safety unit white metal is worn through allowing
instrument air to escape. This in turn causes a pneumatic flag on the control panel to change status.

Rod Run Out

Whereas Rod Drop is a measurement of rod position, Rod Run Out is a measurement of the rod's actual dynamic
motion as it travels back and forth on its stroke. Another term for this measurement is Rod Deflection.

One method to make this measurement is to mount a dial indicator in the distance piece riding on the piston rod.
The machine is then barred through a complete cycle. Indicator readings are taken in both the vertical and
horizontal directions during the machine's cycle.

The amount of Rod Run Out is highly dependent on the cylinder alignment with the Crosshead. Due to inherent
looseness in the Crosshead and thermal growth of the machine, higher readings of Rod Run Out are allowed in

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


4.5.1
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

the vertical direction. The horizontal direction allowances are much less and high readings are attributed to
misalignment. Typical Rod Run Out allowances are 3.5 to 6.0 mils Pk-Pk in the vertical direction and 1.5 to 2.0 mils
Pk-Pk in the horizontal direction.

Crosshead Vibration

The Crosshead of a Reciprocating Machine is made up of several major components: Crosshead Bed, Crosshead,
Slippers and Crosshead Pin. The purpose of the Crosshead is to transform the circular motion of the crankshaft
into linear motion for the rod and piston.

The Crosshead slides on a lubricated babbitted surface much like a standard journal bearing. However, the
Crosshead slides back and forth instead of in a circular motion like a shaft. Clearance between the Crosshead and
the babbitt surface may be in the range of 10 to 25 mils. As crankshaft rotates, the Crosshead is driven to slide on
either the upper or lower babbitt surface. As the clearance between the Crosshead and babbitt surface increases,
the Crosshead vibration increases.

Main Bearing Vibration

Main Bearing Vibration has not been proven to be a popular approach for monitoring Reciprocating Compressors.
Several end users have had problems with broken crankshafts, which they thought were caused by unusual
bending of the crankshaft. In one documented case, machinists had over tightened the drive belts powering the
cooling fan on a reciprocating engine. This caused unnecessary bending of the crankshaft.

Valve Temperature

According to industry studies, valve failures account for 41% of the problems associated with reciprocating
machinery.

In a Reciprocating Compressor, the valves are a pressure actuated "Poppet" variety. Every machine manufacturer
has favorite types of valves for different applications. These valves operate utilizing a delta or differential pressure
technique. The opening and closing of a valve occurs when the delta pressure is less than the force of their return
springs.

When a valve begins to fail, it usually begins to leak the process gas. This causes the process gas to be re-
compressed, further heating the gas. This higher temperature process gas can be detected using a temperature
transducer. This transducer can be mounted through the valve cover plate measuring the gas temperature near
the valve. In some installations, the transducer is simply imbedded in the valve cover plate, or within a valve cover
plate bolt. This mounting method is preferred when the process gas is explosive, which is usually the case.

The measured temperature of the process gas is then compared to the measured temperature of the process gas
at the same type valve, suction or discharge, and the same stage of compression. Measured temperature
differences of 4 to 20o F can indicate a problem with a valve.

Most cooling systems in use today rely on reciprocating piston-type compressors. Reciprocating compressors
are manufactured in three types:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


4.5.2
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

As the piston nears the bottom of its stroke within the cylinder, the intake valve opens and the refrigerant
vapor enters. As the piston rises, the increased pressure closes the intake valve. Then as the piston nears
the top of its stroke, the exhaust valve opens permitting the vapor at the higher pressure to exit.
Reciprocating compressor capacity is a function of the bore and stroke of the piston-cylinder configuration as
well as the speed of the machine, and the clearance tolerances. Compressor capacity is also related to the
compression ratio.

The mechanical design is rugged and reliable but has one significant limitation. Reciprocating compressors
are designed to handle vapors, not liquids. When liquid enters the cylinder on the intake stroke, it tends to
damage the valves on the compression stroke and possibly the compressor itself. This is why chillers
incorporate liquid-to-suction heat exchangers, which assure some level of vapor superheat at the
compressor suction. Capacity is controlled by multiple staging of smaller compressors or in large multiple
cylinder reciprocating compressors by unloading banks of cylinders on the compressor. This tends to make
the machine most efficient at full load. Therefore, for maximum efficiency recips should generally be
operated at full load. This is the reason small compressors are cycled on and off in most residential and
small commercial applications.

The mechanical design is rugged and reliable but has one significant limitation. Reciprocating compressors
are designed to handle vapors, not liquids. When liquid enters the cylinder on the intake stroke, it tends to
damage the valves on the compression stroke and possibly the compressor itself. This is why chillers

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


4.5.3
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

incorporate liquid-to-suction heat exchangers, which assure some level of vapor superheat at the
compressor suction.

Acoustical Resonances

When an acoustical resonance is encountered, pressure pulsations can be reduced by eliminating the
resonance or by attenuating the amplitudes through the addition of a resistive element, i.e an orifice, (or
partially closed valve). A typical orifice diameter ratio to start with is 0.4. This is the inside diameter of the
orifice to the inside diameter of the pipe.

The user must also be careful to keep the suction and discharge pressures sufficiently above the charge
pressures on accumulators that can “bottom out.”

When high frequency vibrations or pulsations are experienced, these can often be eliminated by inline
volume and choke tube type Helmholtz low pass filters.

The volume-choke-volume, or volume-choke-filter configuration is governed by Appendix O in API 618.


Equations taken from Appendix O are given below:

The Helmholtz Frequency (the lowest acoustical resonant frequency of the filter) is found from:

fH = {(c)/[(2)(π)] }{ u/V1 + u/V2 }0.5

where fH = Helmholtz frequency in hertz

c = speed of sound in the gas

V1 = Volume of the cylinder bottle

V2 = Volume of filter bottle

U = Acoustical conductivity = A / [Lc + 1.2Dc]

A = Internal cross-sectional area of the choke

Lc = Length of choke

Dc = Diameter of choke

The filter cut-off frequency above which pulsation attenuation is achieved is usually defined as:

fCo = (1.414)fH

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


4.5.4
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The preferred Helmholtz frequency is:

fH = RPM / 85

where RPM is the compressor speed in revolutions per minute.

Only when conditions are such that it is uneconomical, or physically impractical, should a higher Helmholtz
frequency be considered: that is, only when pressure drop is very critical – as in the case of low suction
pressure, or when space is limited by the compressor system layout. In that instance, a higher Helmholtz
frequency may be chosen. Generally the Helmholtz frequency should not be higher than: RPM / 45 unless
the acoustic simulation proves otherwise. For compressor speeds above 500 RPM, the Helmholtz frequency
should not exceed: RPM / 85.

Filters are effective in pumped liquid systems just like in compressor systems, but for pumps a cut-off
frequency is usually taken to be one-half of the plunger frequency.

Single volume, chokes can also be used as shown in the figure below. The choke tube connects directly to
the larger suction or discharge piping via a reducer. This configuration is not as effective in reducing higher
frequency pulsations as immediately as the volume-choke-volume filter, but it may be very effective in liquid
systems where higher pressure drops can be tolerated better.

Valve behavior can be quite complex and can be influenced by three dimensional flow phenomena, viscous
“sticking” effects with lubricant, lag, and chatter. At the design stage, one must assume that the equipment
manufacturer will provide a dependable machine that does not require special considerations. If this is the
case, then the standard procedures outlined herein will generally provide a sufficiently trouble-and-
maintenance free installation. In special situations, or when machines are put into different temperature or
material service, other problems may arise that the designer and operator must be aware of.

Since cavitation at pump suctions can produce considerable pulsation problems, API 674 provides that some
separation must exist between the fluid vapor pressure and the lowest pulsation pressure. As recommended
in, “The Effect of Pulsations on Cavitation in Reciprocating Pump Systems,” by Wachel, Tison, & Price of
Engineering Dynamics Incorporated, the cavitation must exist for some significant part of the total time
before significant problems arise. From the paper, “Based on the Author’s experience, a CPN value above
25% is required before the pump would have significant cavitation. Values above 50 are typically required to
effect a 3 percent reduction in flow capacity.” This says that the suction side pressure pulsation would need
to be at the vapor pressure at least 50% of the time. This seems to be somewhat excessive and is left for
the user to further investigate.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


4.5.5
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

When a straight pipe terminates in an opening, there is an “equivalent length” that is slightly longer than the
actual length. This extra length compensates for the mass of gas near the open end which oscillates with the
gas in the pipe. For an open pipe, various empirical expressions have been given for the equivalent length.
One of the more common is:

Le = L + D(0.85 – 1.06 D/Do)

where D is the inside diameter of the pipe,

L is the actual length, and

Do is the inside diameter of the vessel into which the flowing pipe empties.

Kinsler and Frey give the following approximations:

End correction for a flange = L + 0.8D

End correction for an unflanged pipe: = L + 0.6D

Quarter wave stubs can be used to detune main runs that experience resonance effects since the stub will
short circuit the pressure peaks. For the best results the side branch diameter should have an area at least
equal to ½ or more of the main pipe. Quarter wave stubs should be put at points of pressure maxima in the
main line to detune it.

For open-open or closed-closed resonances, the resonant frequencies are half-wave frequencies and fall at:

F = (n)(c)/[(2)(L)] where n = 1,2,3,…

For open-closed resonances the quarter wave frequencies are:

F = (2n-1) (c) / [(4)(L)]

In normal pipe systems an open end resonance can be sustained if the pipe size suddenly increases to twice
the flow area or more. Conversely, if the area suddenly decreases by a factor of two or more, sufficient
reflection will result to permit closed end resonances.

Magnification of pressure waves on the order of 30 can exist in typical piping systems that experience
excitation at an acoustic resonance. This means that if the pressure source variation is 10 psi, the resonant
pressure in the line could be (30)(10) = 300 psi.

API 618 N.2.1 “When resonant lengths of pipe are used, and the resonant frequency matches compressing
frequency, one can expect major changes to the system as a result of the acoustical simulation analysis. The
resonant length of various piping configurations is given …. It is recommended that lengths of these
configurations be avoided in a plus-or-minus 10 percent band for the first four harmonics of compressor
speed. The piping areas where this is most important are the sections of piping between the first major
volume on the suction side and the first major volume on the discharge side.”

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


4.5.6
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Slugs and Flashing


BOS Fluids provides a number of independent tools for evaluating slug flow and the mechanical loads
on piping systems that result. Slug flow is different from waterhammer and steamhammer loads in
several important ways. Flashing is primarily a process and thermodynamic event that results in erratic
volumetric flows downstream of the point where flashing occurs. The local velocity changes in the
fluid can produce forces at changes in directions such as bends and intersections. BOS Fluids permits
the user to incite these local velocity changes using a FLASH large cavity model. Verification of the
thermodynamic state is done after the calculation is performed by evaluating the total inflow and
outflow volumes from the reaction. Flashing is described at the bottom of this section.

1) Slug flow loads develop because of the momentum change at elbows or intersections. Waterhammer
and steamhammer loads develop because of a differential pressure between elbow-elbow, or elbow-
intersection pairs. For waterhammer and steamhammer, the momentum change is negligible.

2) Slug loads are proportional to (ρ)(A)(V2), where (ρ) is the density of the slug liquid, (A) is the
inside area of the pipe cross-section and (V) is the velocity of the moving slug. Waterhammer and
steamhammer loads are proportional to (ρ)(c)(dV), where (c) is the speed of sound in the fluid, and
(dV) is the velocity change of the flow. Since the speed of sound (c) is much larger than the flow
velocity, waterhammer and steamhammer loads are much more rapid events, often occurring on a
millisecond time scale. Slug loads develop over a longer period, and occur typically on a second time
scale.

3) Slugs travel at flow velocities while waterhammer and steamhammer pressure waves travel at the
much higher speed of sound.

4) Slug loads typically move through a piping system like waterhammer and steamhammer loads but
do not reflect from end boundary conditions. Slug loading on an individual elbow, therefore, is
generally only a single application load that can be readily treated in mechanical piping system
analysis by the application of a Dynamic Load Factor, usually equal to two. Waterhammer and
steamhammer loads are repeated loadings that can easily excite system acoustic and mechanical natural
frequencies producing dynamic load factors equal to ten or more.

The velocity of the slug is probably the most important mechanical design parameter.
Overly conservative values produce excessively conservative forces as a function of
the slug velocity squared. For this reason an accurate prediction of the slug velocity is
important.

The slug and flashing facility in BOS Fluids is a new feature that was prepared for
version 4.110. User’s are cautioned to employ it carefully, and to verify the results
using experience, expectation, and other calculations.

The BOS Fluids Slug calculation introduces the pipe contents into an essentially
empty (voided) pipe. This documentation should be studied carefully before
attempting to apply voided pipe applications, (i.e. slugs and flashing).

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.1
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Properly evaluating the mechanical effect of slugs in piping systems involves:

1) Determining if a slug of liquid will exist in the gas system.


a) Can liquid collect in low areas of a gas system?
b) Can a two-phase gas/liquid system produce a stable slug?
2) Finding the slug velocity and the part of the system that is likely to be affected.
3) Including the effect of changes elevation, pipe size and branches.

Slugs can exist due to a number of conditions, a few of which are:


1) Steam condensate collection in low areas, and/or trap malfunction
2) Two-phase flow lines
3) Firewater pumping systems. (Firewater is pumped into an empty system.)
4) Emergency systems where liquid is evacuated rapidly into an empty line.
5) Liquid products injected into a vapor line or vice-versa.
6) Control or operation errors.
7) Incorrect shutdown and startup procedures.
8) Flare systems collecting multiple process flows.

There are three methods available in BOS Fluids to investigate the effect and quality of slugs.

1) Slug Calculator ( )
a) Predicts possible maximum and minimum slug flow velocities.
b) Predicts the tendency for stable horizontal and vertical slugs to exist.
c) Predicts the maximum horizontal length required to produce a maximum slug velocity.
d) Estimates free flow velocity of the gas (very crude estimate of slug velocity)

2) LARGE CAVITY – SLUGFLOW BOS Method of Characteristics Approach.


a) Follows a design slug through a 3-dimensional BOS Fluids model of the piping system. Slug
types include:
SLUG – Impact Unstable Slugs
GLUG – Non-Impact Stable Slugs
PLUG – Intermittent Unstable Slug Flow.
FOAM – Uniform Low Density Flow
FLASH – Unstable Variable Flow Expansion

3) SPLASH – Effects of Slugs on Elbows and Other Flow Obstructions.


a) Shoots a 2D model of the design slug at an elbow or pipe intersection.
b) Effects of reducers, orifices, or other 2D flow obstructions can be studied.
c) Shoots a 2D model of the slug at impingement plates inside slug catchers.

The major goals in the BOS Fluids slug evaluation process are:

1) To predict the tendency for a particular flow state or geometry to experience slugging, and
2) To accurately predict the unbalanced loads that act on the piping system at elbows and
changes in direction due to the slug impact.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.2
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Once the user has a feel for the probability of slug development and the maximum velocities involved
a three dimensional analysis can be run to determine how the slug loads will be distributed throughout
the piping system as the slug travels.

Each of the slug problems described above can be put into one of the following categories:

1) Maximum Slug Velocity equal to Free Flow gas velocity


2) Maximum Slug Velocity determined by the pressure differential accelerating the slug and the nature
of the slug.

The primary factors that affect the stability of the slug flow are:
1) Line orientation and geometry
2) Liquid and Gas volume fractions
3) Tendency of the liquid to flash
4) Viscosity and surface tension of the fluid

Two phase flow problems, and problems where liquid flows along the pipe with a gas, fall into the
category where the slug is traveling at essentially the free flow velocity of the gas. In these problems,
the velocity of the slug can usually be easily predicted by the process engineer. Low-point collection
problems, where the liquid slug is accelerated from a zero velocity fall into the category where the
maximum velocity of the slug must be determined.

These conditions are described graphically in the figures below:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.3
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

In the top drawing, liquid has collected at a low point during one phase of a process due to shutdown, a
process temperature change, an error in the operating procedure, altered feed stock quality, etc. When
gas flow is redirected through the pipe section where the liquid has collected without the pipe being
drained, a slug can be propelled down the horizontal run as shown in the figure. In this case the initial
slug velocity is zero, and the slug is propelled to its maximum velocity by the build-up of pressure
behind the slug. The buildup of pressure must overcome the fluid head associated with the relative
vertical rise, friction, and the slug’s resistance to acceleration. The deliverable maximum gas flowrate
can also affect the maximum velocity the slug can achieve at some point further on in the pipe.

The middle figure shows the collection of liquid at a change in elevation in a flowing two phase
mixture. Liquid can collect in the downstream vertical leg and form a slug which will be transported
down the pipe.

In both of the top two instances the slugs start with an essentially zero velocity and are accelerated to a
potential maximum velocity. The distance over which the slug must be accelerated to reach the
maximum velocity is often of the order of several hundred feet. When these distances are not available
in the downstream piping the momentum change loads produced by the slug can be significantly
smaller than those predicted by a slug traveling at the free flow velocity of the gas.

For a slug starting at a zero velocity to reach the free flow velocity of the gas instantly, its mass must
be zero, friction acting against the liquid must be zero, and pressure buildup in the gas upstream must
be instantaneous. The slugging liquid does have a mass that must be accelerated. Pressure buildup in
the gas flow is not instantaneous, (it is a function of the initial volume of gas in the pipe and the

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.4
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

pressure source, i.e. constant flow or constant pressure), and friction against the slug is not zero, but
may become a controlling factor as the slug is accelerated to velocities approaching the free flow
velocity of the gas.

A conservative calculation can be made in these cases by always assuming that any stable slug travels
at the free flow velocity of the gas. Unfortunately, where this is not true overestimates are a function of
the velocity squared and can be excessively large, especially where horizontal dynamic loads exist
high-up in a structure not originally designed to absorb these loads.

The BOS Fluids calculation tools help the user estimate slug loads in piping systems and design
economical, safe supporting systems. The procedure below takes the reader through several typical
slug design load scenarios. The user is encouraged to read through each, to be sure that a thorough
understanding of the physical events, and the software’s simulation of the physical events, is obtained.

The bottom two figures in the diagram above show horizontal and vertical sections of pipe that contain
two-phase flow. When two-phase flow exists, slugging may occur when the flow is comprised of the
correct proportion of liquid and gas flowrates. The slug calculator helps the user make these
determinations.

Two types of slug creation have been described:

1) Collection of fluid in a low spot, or direct injection of fluid in a gas line. (Slug starts from zero
velocity.)
2) Two-phase flows, where mutually flowing liquid and gas concentrations produce a fully closed pipe
section. (Slip between two phase flows is usually small in practical applications where slugging is a
mechanical concern. It can be assumed that the liquid and gas flow at the same velocity. Slip can be
significant both when the pipe section is uniformly almost full or uniformly almost empty. The user is
cautioned to examine the uniformly almost full condition carefully to be sure that a rapid buildup of
pressure cannot occur when a slug is formed.) (Slugs, in this case, start from essentially the free flow
velocity of the mixture.)

Whereas, it can be assumed that the liquid and gas portions of a two phase flow travel at the same
velocity, slugs created by the collection of fluid in a low spot do not travel at the free flow velocity of
the gas. Assuming the slug immediately achieves the free flow gas velocity may be overly conservative
for many plant piping systems that do not contain long downstream pipe sections that permit an
unimpeded slug acceleration to the free flow velocity. Where horizontal slugs turn downward, an
increase in the slug velocity can be expected. Where small lines and highly viscous slugs collect in low
pressure gas lines, the slugs can be accelerated to speeds higher than the free flow gas velocity.
Variations and slip exist between the gas and liquid velocities occur primarily due to changes in
elevation, such as inclines in a pipeline or when a horizontal pipe cross section is uniformly almost
full. Certain of these cases can be extreme and should be avoided.

In general, slugs formed by low point collection in typical plant environments escape the vertical
confine and begin accelerating from an entry velocity on the order of one meter per second. These
slugs accelerate until friction overcomes the increase in potential rise due to the upstream change in
elevation. The slug calculator predicts the velocity that exists when the slug is fully in the horizontal
position, and when the slug has reached its maximum velocity. (These locations and velocities are
shown in the figure below.)

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.5
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Separators often serve as a terminator in a two phase flow pipeline system. In this capacity they
separate the liquid and gas, and trap any slugs (slug catcher).

Once a liquid becomes a traveling fluid column it can be either stable or unstable. If stable then it will
continue throughout the system until encountering a significant obstruction, manifold, volume change,
or vessel. (These are the slugs that piping engineers are most concerned about.) The stability of a slug
in a two-phase flow can be predicted, although the predictions are notoriously poor. The long-term
stability of a low-point collection slug cannot effectively be predicted, and so whenever they can exist,
if the system is critical, some evaluation of the resulting loads is often performed.

The maximum possible slug velocity in a horizontal run (not accelerated by an upstream vertical drop),
can be found from: [(QL)(ρL/ρG) + QG)] / (Ai). The assumption should only be used if the actual free
flow velocity of the gas cannot be obtained from the process engineer or measured. (QL) is the
volumetric flowrate of the liquid and (QG) is the volumetric flowrate of the gas, ρG is the density gas
and ρL liquid density, evaluated at flow conditions.

Slug stability can be predicted for isolated vertical and horizontal runs using the BOS Slug Calculator.
If a slug is stable, then it’s effect on the piping system can be calculated in the BOS Fluids LARGE
CAVITY-SLUG model. The LARGE-CAVITY simulation will show the effects of changes in
elevation, elbows, variations in diameter and intersections. The LARGE-CAVITY-SLUG analysis
demonstrates where and how the slug moves through the system, and calculates the unbalanced loads
caused by the slug momentum flux at elbow-elbow pairs and other “paired” changes in direction.
SPLASH predicts the waveform of the slug loading as it traverses elbows or changes in direction, for
example: bends, tees, reducers, impingement plates, or vortex breakers.

The low frequency content of the slug impact on the mechanical system is controlled by the length of
the slug. The slug length governs the total duration of the unbalanced load when the slug length is
short compared to the distance between changes in direction. The high frequency content of the slug
impact on the mechanical system is controlled by the shape of the impact waveform which can be
evaluated by (SPLASH). The time history of the slug load, whose primary variables are length and
waveform, can be entered into most pipe stress programs to compute stresses, displacements and
support loads. The BOS Fluids LARGE CAVITY slug simulation incorporates reasonable
assumptions for both the waveform and the length into the analysis and predicts the unbalanced load
time history that should be input directly into any time history analysis of a piping system.

The three BOS Fluids analysis tools described above are:

1) Slug Calculator
2) SPLASH
3) LARGE CAVITY-SLUG Simulation

For standard radius bends the approach outlined in CAESAR and used elsewhere for the duration of
the slug impact rise-time has been shown by SPLASH to be very reasonable. The CAESAR slug
impact model uses a linear ramp-up of the impact load, where the total duration of the ramp is equal to
(π/2) (R+r)/V, with R = bend radius, r = pipe radius, V = slug velocity. The maximum amplitude of
the impact is assumed to be caused by momentum change and is ρAV2. If the downstream pressure

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.6
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

drops due to the presence of the slug, as may occur in the presence of long, stable slugs that start with a
zero velocity, or where multiple long slugs travel in unison, then the slug load can be higher by an
amount on the order of PA for the outgoing component of the slug although this event is not thought
common. Other phenomena can exist in very high viscous flow situations, or where the pipe diameter
is very small, and these are not addressed.

The LARGE-CAVITY-SLUG impact facility in BOS Fluids provides coefficients for both the
momentum and pressure components of the slug loading so that the user can investigate separately or
in unison the contribution of each to the unbalanced load.

Whereas the variety of scenarios is considerable, the approach used to reasonably and conservatively
analyze a mechanical piping system for slug events is straight-forward and described via several
examples.

The exact approach to be employed depends on what the user already knows about the flow state:

1) If it is known that stable slugs will exist in the system, but the exact two-phase flow makeup of the
slug is not known, the user can run the Slug Calculator iteratively to obtain a range of flowrates and
velocities to be used in the slug design. Slugging conditions will typically only exist when the liquid
volumetric flowrate is greater than the gas volumetric flowrate. The controlling parameter for design is
the maximum horizontal length of straight pipe in the system and the largest vertical change in
direction between two consecutive straight sections, as these quantities have a significant impact on the
maximum possible slug velocity.

2) If the flow makeup is controlled precisely by the process conditions, then the exact stability of the
flowing system can be evaluated. Again, usually a variety of runs are made and both stability and
maximum slug velocities are estimated for each pair of runs. Hypothetically developed slug scenarios
and flow directions should be postulated, and the design adjusted accordingly based on maximum
forces calculated.

3) If the user wants to know what level of liquid in a line can generate stable traveling slugs, the Slug
Calculator can be run iteratively with the largest vertical and horizontal runs entered. The amount of
liquid in the line must be increased until a stable slug can be maintained. At this point the maximum
slug velocity can be determined from the Slug Calculator, and a design force equal to ρAV2 estimated.
If the slug can build up a substantial back pressure that also might be damaging, (like e.g. for large
diameter pipe, high viscosity flows, etc.), then the impact of the back pressure must be taken into
account in the definition of the “slug” load and the load magnitude is defined by ρAV2 + λPA, where λ
is a constant defining the back-pressure as a fraction of the line pressure Typically this constant is
between 0.0 and 0.3 and has to be estimated by the user.

4) If the user is not sure of the source of a slug, but because of process conditions, valving,
condensation, etc., it is desired to evaluate the results of a slug flow, then the user can run a variety of
horizontal and vertical pipe dimensions and volumetric ratios in the Slug Calculator, recording the
maximum possible slug flow. This can be compared to the free flow velocity of the gas and the
smaller value used. Additionally, the NOFLOW-Slug evaluation will produce a slug velocity that can
be checked against the Slug Calculator. The NOFLOW simulation will also evaluate the impact at
intersections and changes in diameter.
The three slug calculation tools in BOS Fluids are illustrated in the figure below:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.7
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.8
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

BOS NOFLOW-SLUG Calculation shows loads due to slug movement throughout the 3d piping
system.
SPLASH – Calculates loads and pressures due to slug impacts on various geometries. (Bend shown.)
Slug Calculator – Determines maximum slug velocity and stability based on flow properties.

BOS Slug Calculator – The slug calculator performs three major functions:

1) Reports the tendency to form a slug in a vertical section of pipe.


2) Reports the tendency to form a slug in a horizontal section of pipe.
3) Gives the maximum and minimum possible velocities.

The calculator is used most often iteratively to determine which system parameters have the greatest
effect on the tendency for a slug to form and on the maximum possible velocity. The user accesses the
slug calculator via the button sequence below:

Slug Calculator Input Text Cells

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.9
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

A parametric study should be run for the ranges of configurations experienced in the system. The user
should assume that slugs will initiate from any point where liquid can collect. The diameter and wall
thickness effect should be evaluated if the system has potential slug flow through a variety of pipe
sizes. The vertical pipe length input should be varied so that the user can see how the geometry affects
the slug velocity distance requirements. (Long downstream distances are often required to develop the
maximum slug velocity from slugs starting at a zero velocity.)

Three types of slug calculations are performed:

1) Slug stability:
The stability of a slug is based on the pipe orientation, the liquid and gas properties, and the
respective liquid and gas flowrates. These calculations are based on a “uniform”two phase flow,
with an essentially no-slip boundary between the liquid and the gas. Slug stability is not dependant
on the vertical elevation change entered in the calculator.
2) Minimum and maximum velocities.
The minimum and maximum slug velocities that can be produced from a slug starting at a zero
velocity is a function of the vertical elevation change that must be overcome by the slug before it
can begin accelerating in a downstream horizontal pipe section. A certain length of horizontal pipe
section must also be present for the slug to accelerate to its maximum velocity. The slug calculator
also predicts these values.
3) Free flow velocity of the slug.
The free flow velocity of the gas is a value often used as a crude estimate of the slug velocity. The
free flow velocity is estimated from the properties entered, and may not be representative of the
actual free flow gas potential. If possible this value should be obtained from a process engineer
more familiar with the flow.

The geometry input cells are described below:

Units: The user can check English or Metric units. Metric units are shown.
Pipe OD: Enter the outside diameter of the pipe to be studied. The slug calculator assumes only a
single, uniform cross section.
Pipe Thk: Enter the wall thickness of the pipe to be studied.
Roughness: Enter the pipe roughness used to calculate friction drop – 0.05mm (0.002 in.) is a typical
value used for clean steel pipe.
Vertical Pipe Length: Enter the largest vertical pipe length in the system. The user should investigate
areas of the system where slugging is likely to be initiated. These are low points where fluid or
condensate can collect. The longer the vertical run just downstream of the collection point, the higher
the velocity of the slug that is propelled from the top of the vertical collection leg when the cross
section is closed.

A typical input and output is shown below. All input to the program is provided in the text boxes on
the top, and the output is produced in the heavy, dark text in the “output” frames in the bottom of the
window. The program is designed so that the user can change a value then press “calculate,” to see
how the changed values affect the results. As the user steps through a parametric range of inputs, any
runs of interest can be printed.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.10
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The pipe roughness is a function of the pipe material, age and condition. Roughness often will change
as the pipe ages. This must be evaluated on a case-by-case basis, but 0.05 mm (0.002 in.) is a typical
value used for clean steel pipe.

The density of the gas is at the nominal line pressure, and the volumetric flowrate is the amount of gas
flowing in a two phase section. The total flowrate is the sum of the gas and liquid volume rates. The
liquid flowrate may be left blank. The calculator will find a low-end value of liquid flowrate that will
produce slugging in a horizontal section of the pipe. Liquid flowrates significantly below the
minimum liquid flowrate calculated will not produce stable slugs in two-phase horizontal lines.
Slugs can still be accelerated down these lines if collected in low points however. Just because a
slug is not stable, does not mean that one cannot be introduced into a gas stream
because of condensation, operation error or improper draining.
The surface tension of water is 0.072 N/m. (0.0004 lb/in.) Values for other fluids are readily available
in the literature.

If the user is unsure of the liquid volumetric flowrate, there are several options:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.11
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

1) If the flow is two-phase, and if the gas and liquid flow-rates are known and also an assumption for
the viscous effects on liquid hold-up is made the percent of the pipe that is full of liquid can be
calculated. (In the above example the gas-to-liquid volumetric flow ratio is 0.71 gas / 0.142 liquid = 5.
If it is assumed that viscous effect are small compared to inertia effects i.e. the Reynolds Number is
large the liquid hold-up is almost equal to the Liquid Volume fraction. In that case the pipe section is
0.71 / (0.71+0.142) = 0.833, or 83% gas filled and the corresponding liquid hold-up is 17%).
Obviously the liquid hold-up (the fraction of the flow area occupied by the liquid) increases with an
increment of the viscous effects. Typically in real life the liquid hold-up is a multiple of the liquid
volume fraction.

2) If the flow is two-phase, the liquid flowrate text cell can be left blank and the calculator will
calculate a lower-bound value for the liquid flowrate that will produce slugging in a horizontal line.
(This calculation is an estimate only and should be verified.)

3) The gas flow velocity can be entered, (which must be known), and the program can be used to
calculate the liquid flowrate. From these values the user can reduce the gas flowrate and increase the
liquid flowrate to produce a range of values where slugging is possible. From this range of values,
maximum velocities can be predicted.

The calculated free flow velocity of the gas is an estimate only, and should be verified by the process
engineer or by measurement. The value for maximum free flow slug velocity is intended to provide a
crude estimate of the free flow velocity of the gas given the flowrate of gas and liquid.

Help with Units and Conversions

BOS Fluids has a flowrate calculator available from most of the 3d piping input screens. It is accessed
by hitting the F10 function-key. The flow calculator panel is shown below. The inside diameter and
pipe length are entered. The calculator computes the flowrate, pressure drop and total volume. The
user can change any of the inputs to see its effect on the other items. For example if the user wanted to
see how many cu.m/sec came from 17988 GPM, they could read the value directly from the table
below. To see how much came from 2000 GPM, the user should just change the value in the GPM
field and observe how all other items in the table change based on the user input.

BOS Fluids Units Converter

The user can also print the results from the slug calculator using the “Print” button. An example output
is shown below.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.12
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

BOS Fluids Slug Flow Evaluator V4.110

Input:
Outside Diameter = 250. mm
Wall Thickness = 10. mm
Roughness = .05 mm
Vertical Length = 4. m

Gas Density = 20. kg/cu.m


Gas Flowrate = .4 cu.m/sec

Liquid Density = 900. kg/cu.m


Liquid Viscosity = 1. cp
Liquid Surface Tension = .07 N/m

VERTICAL SLUG STABILITY


Have bubbly flow regime (No slug)
HORIZONTAL SLUG STABILITY
Slug Flow Possible!!

Minimum Slug Velocity = 6.1008 m/sec


Maximum Slug Velocity = 18.00 m/sec
Free Flow Slug Velocity = 20.5175 m/sec
Maximum Slug Velocity Pipe Length Required = 267.18 m

The vertical and horizontal slug stability predictions, and the free flow velocity calculation are based
on the gas and liquid properties and flowrates only. The input length of vertical pipe will not affect
these values. Horizontal slug predictions can vary if the pipe is sloped, and once the line slope
approaches 20-to-30 degrees, the flow tends toward the vertical flow regime.

The maximum and minimum slug velocities, and the evaluated slug development are shown in the
figure below:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.13
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

When slugs are geometry dependant, the longest slug that will form is equivalent to the maximum
elevation difference. Slugs formed from an elevation potential are started from a low velocity, assumed
in the BOS slug calculator to be 1 meter per second. The distance the slug must travel to attain a
maximum velocity is calculated and printed. Prior to this length the velocity is less, and the velocity
reduces after the maximum velocity is attained because of friction. This model does not include any
potential build-up of gas pressure beyond that required to elevate the fluid column. When the
horizontal length of pipe downstream of the elevation change is shorter than the pipe length printed,
the maximum velocity of the slug will not be attained. The minimum slug velocity printed is the
velocity at the point when the slug is fully horizontally and traveling in the downstream leg. The slug
will accelerate from the minimum value, to the maximum value over the identified distance. (This
process is described in the figure above.)

The only parameters used to calculate the minimum and maximum velocities and distance to the
maximum velocity described above are:

1) Vertical elevation change


2) Pipe diameter and wall thickness
3) Friction factor

Summarizing:

1) Horizontal and vertical slug stability predictions are based on gas and liquid properties and
flowrates only. A modified Mundane flow regime map is used for horizontal two-phase flow slug
stability, and an Aziz Up-Flow regime map is used to predict vertical slug flow stability. Slug stability

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.14
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

predictions are based on uniform, non-slip flows of a two phase liquid and gas. (When slip is
significant, the volume flowrate differences are typically too great to produce slugging or other
geometric considerations such as pipeline slope interact significantly with the potential for slugs to
form and accelerate. These are considered special conditions however, and their evaluation should be
based on previous experience, especially where pressure buildup behind a closed pipe section can
occur.)
2) The predicted free flow slug velocity is a crude estimate based on the collective liquid and gas
flows. The actual value should be obtained from the process engineer!
3) The maximum slug velocity is based on a slug length equal to the vertical rise, introduced into the
horizontal run at a velocity of 1 meter per second. This slug is accelerated down the pipeline until
friction forces overcome the acceleration of the slug. The maximum velocity that is printed is the
maximum velocity attainable, and only occurs for an instant at the distance specified. This is intended
to be an upper limit for a slug accelerated by the specified vertical elevation change.
4) The minimum slug velocity is based on a slug whose length is equal to the vertical elevation change
just after it has been fully introduced into the horizontal pipe run.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.15
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

BOS Fluids Splash Slug Evaluation


The BOS Fluids Splash documentation can be found in Chapter 3 Section 1 of this User’s Guide. The
standard slug model for a slug impacting an elbow is shown below:

This model is constructed from the main menu of the Splash input screen via automatic input. An
elbow-elbow section can also be generated using an automodel of an elbow-elbow geometry. This
construction and a typical result is shown below.

An intersection, elbow and reducer geometry can be defined in the user-geometry section. The results
from this model are shown below.

With some modeling creativity many two- and three-dimensional intersection/elbow/diameter change
geometries can be investigated and the unbalanced load calculated.

The SPLASH unbalanced load is always calculated in the horizontal direction, and slugs should be
introduced accordingly.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.16
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

BOS LARGE CAVITY-SLUG Evaluation


The BOS Fluids LARGE CAVITY-SLUG model simulates essentially full section pipe slugs. These
are slugs whose “front” can be described by a relatively short axial “density” variation. These
liquid/gas boundary sections can be of five types:

SLUG – Impact Unstable Slugs


GLUG – Non-Impact Stable Slugs
PLUG – Intermittent Unstable Slug Flow.
FOAM – Uniform Low Density Flow
FLASH – Unstable Variable Flow Expansion

The SLUG, GLUG and FOAM slug types develop, and are driven by a pressure and continue with a
constant or reducing flow depending on the boundary condition. The PLUG and FLASH slug models
provide for an expansion of a liquid or gas traveling in varying densities.

A SLUGHEAD boundary condition is used when the slug flowrate should be constant. (This is the
most conservative assumption to make.) A TIMEHEAD boundary condition should be used when the
slug flowrate should reduce either as a function of time or distance.

Five worked examples illustrate the use of this method:

1) Typical slug flow


2) Steam flooding
3) Slug Catchers
4) Firewater pump start
5) Relief Valve Downstream Flashing

In all cases the pipe is voided by the application of a startup pressure boundary condition below
the vapor pressure. The vapor pressure specification in these instances can be the actual liquid
vapor pressure or an artificial vapor pressure. The examples below demonstrate how the vapor
pressure is used to void the pipe.

An artificial vapor pressure head must be used if any part of the system drops below the highest
specified boundary condition pressure head, or if the actual vapor pressure is less than one atmosphere
absolute. (The pressure boundary conditions must produce a NOFLOW pressure throughout the system
that is less than the vapor pressure so that the desired part of the system will be voided when the
transient solver starts. The examples below illustrate.)

The artificial vapor pressure head is often the elevation difference between the highest point in the
section of the line that is to be exposed to the slug (emptied/voided) and the lowest. The artificial vapor
pressure can be calculated by multiplying the maximum elevation difference by the density of the slug
liquid and by gravity. The actual liquid pressure driving the slug should be added to the artificial vapor
pressure and used as the SLUGHEAD maximum pressure so that the pressure difference is correct.

English Units Example:

Maximum Elevation Difference = 14 ft.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.17
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Liquid Density = 62.4 lb./cu.ft.


Artificial Vapor Pressure Used = (14)ft.(62.4)lb./cu.ft. / (144) = 6.0666 psig. = 20.76 psia.

Metric Units Example:

Maximum Elevation Difference = 5 meters


Liquid Density = 998 kg./cu.m.
Artificial Vapor Pressure Used = (5)m.(998)kg./cu.m.(9.81)m/s2 /(100,000) = 0.489 barg.

It should be verified that proper portions of the pipeline are voided by printing a cavity report before
the slug is introduced into the line. The cavity report should show the entire section of the pipe
downstream of the slug voided, i.e. 100% vapor.

This approach is demonstrated schematically below. The slug is introduced at node 5 and travels
through the system to node 10.

The 5 meter head of 0.489 bars is added because otherwise there could be a pressure at node 5 that was
below the vapor pressure of the fluid because of the 0.0 psig boundary at node 10 on the right and the 5
meter rise from node 5 to node 10. (The 5 meter rise will be supported by an increase in pressure at
node 5. This pressure will not exist in a system that is fully above the fluid vapor pressure and so to be
sure that the entire system is voided the vapor pressure will be set to the effective value of 0.489 barg.)

The method may not be clear until applied a few times. As a guideline to follow the user should:

1) Specify initial boundary conditions of 0.0 psig (or 0.0 barg) for each point in the system that is
attached to a part of the pipe that is voided on startup.

2) Find the largest elevation change between the highest boundary with a 0.0 pressure boundary and
the lowest point that should be voided.

3) Find the pressure that corresponds to this elevation change as shown in the example above for
English or Metric units.
© Copyright 2003, Paulin Research Group & Dynaflow Engineering
5.1.18
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

4) Set the vapor pressure (in absolute pressure units) equal to the pressure due to the elevation change.

5) Make sure that the first printout of a cavity report is before any transient is started so that the user
can confirm that the desired sections of pipe have been voided and that there is no flow in the empty
pipe sections, regardless of elevation change.

6) Add the calculated vapor pressure to the driving pressure, most typically the throat pressure at the
orifice of the relief valve.

The slug solution starts at a time of zero with a NOFLOW steady state solution. The NOFLOW
solution means that NO “liquid” is flowing anywhere in the piping system when the transient solution
starts. Inertia and friction effects at all points are zero.

Pipe slugs most likely won’t exist just downstream of large manifold inlets. Slugs are broken up at
sizeable increases in volume such as slug catchers, manifolds, and condensers. Slugs build up in low
areas where liquid can be trapped and raised in an adjoining vertical section of pipe. The vertical
section adjacent to the low horizontal section collects the fluid and determines the net head available to
accelerate the slug. The process engineer or plant personnel responsible for operation of the system
should be consulted so that proper operating conditions and valve positions are considered. (Slugs are
often introduced because improper valving results in a process condition that was not anticipated.) The
analyst must know what parts of the gas system are susceptible to liquid collection and what parts are
not. After a basic understanding of the system is developed, the BOS Fluids slug model can be
adjusted to fit the assumed parameters. In the steps below the basic slug model is developed and the
user taken step-by-step through the slug evaluation procedure. Inputs and outputs are displayed so the
user can reproduce both the thinking and the results.

ALWAYS BE SURE TO USE THE LARGE CAVITY, CAP COLUMN SEPARATION MODEL
FOR BOS LARGE CAVITY-SLUG simulations. Usually the GLUG option should be used, but
the user is encouraged to experiment with other options to see which best suits the anticipated
flow conditions and to make sure that the solution is not sensitive to the calculation procedure
employed.

The first step is to evaluate the system and decide where a slug will likely be introduced into the pipe.
Once these points are identified, a BOS Fluids model of the piping system downstream of the slug
introduction point must be constructed. The example will employ a basic slug geometry as shown in
the figure below:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.19
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The pipe to the right of the slug initiation point will be voided on transient solution startup by
introducing a pressure at the initiation point and the outlet to assure that the entire section of pipe
over which the slug will travel is below the vapor pressure at time zero. A “SLUGHEAD” boundary
condition is specified at the initiation point to define a time waveform that will simulate the
introduction of the slug. (The rise time used in the SLUGHEAD boundary condition is generally not
“known”, and the user should verify that the solution is not sensitive to the “rise-time” of the slug
introduction, although inertial effects generally assure that this is true. A first estimate of the rise-time
is made from (π/2) (R+r)/V, where “R” is the bend radius, “r” is the mean radius of the pipe and “V”
is the maximum slug velocity.)

A SLUGHEAD boundary condition is shown diagrammatically below.

As the boundary condition pressure rises above the vapor pressure, liquid is injected into the line and
simulates the flowing slug. The plateau pressure is the driving pressure behind the slug. After some
time: (tc), long enough to develop the design slug length, the flowrate is held constant and the pressure
increased to simulate a fixed length flowing slug. Tc is entered by the user and can be estimated from:

( Tc – Tb ) = L / V

where: L = the maximum length of the slug, and


V = the velocity of the slug.
Tb = time during the slug pressure rise.

Tb should be closer to the start of the rise time to produce a greater maximum slug velocity and closer
to the end of the pressure rise to reduce the slug velocity. If the maximum length of the slug is not
known, then Tc should be selected after the desired slug flowrate is established. If a slug flowrate is
not known, then the maximum length of the slug can be taken as 20 diameters, and Tc found from the
equation above. The user should be sure that the solution is not significantly affected by the selected
value for Tc, and usually, it is not.

An initial estimate of the slug velocity can be found using the BOS Fluids slug
calculator. The velocity can be verified after the 3d simulation is performed and should
be between the minimum slug velocity and the maximum slug velocity predicted by the
calculator.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.20
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The user should verify that the segment of the voided line contains 100% cavities. BOS Fluids
generates a “Cavity Size” report that can be inspected by the user at particular times to see where the
pipe is filled and where it is voided (empty). The “Cavity Size” report can also be used to follow the
slug front as it moves through the system. The times are specified in the input. Up to 20 time inputs
may be entered on the Exclusion/Force/Cavity form. The user should experiment with these as
significant volumes of output can be produced.

The line shown above is an 80 deg.F air line operating at 150 psig, but has collected fluid in the past,
and has experienced slugging. The collected fluid is water, whose properties are entered in BOS Fluids
as shown below. (Note that for metric units, the WEIGHT density in NEWTONS per cubic meter
must be entered.)

Water Weight Density = 10,000 N/cu.m


Viscosity = 1.0 cp
Bulk Modulus = 2000 MPa
Vapor Pressure = 2.5 bar absolute

It is suspected that the water forms from condensation somewhere else in the process, but collects in
the low spot upstream of point 10 and then forms a slug after clearing the elbow at node 10.

The air flow is 23,000 SCF/min. A standard cubic foot of air weights approximately 0.075 lb./cu.ft.
The mass flowrate of air is:

23,000 (SCF)/min * 0.075 lb/SCF = 1725 lb./min.

At 80 deg.F and 150 psig, the density of air is:

ρ = p/RT = (150+14.7) (144) / [ (53.342)(540)] = 0.823 lbm/cu.ft

53.342 = Gas Constant in ft.lbf/lbm/degR.


540 = Temperature in deg.R.

The volumetric flowrate of air without the slug will be:

Q = 1725 lb/min / (0.823 lb/cu.ft) = 2096 cu.ft/min.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.21
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The pipe has a cross sectional area of 105.2 sq.in. The free flow velocity of the air is:

V = Q/A = (144 in2/ft2) (2096 cu.ft/min) / (105.2 sq.in) = 2869 ft/min (47.8 fps)

There is a five meter elevation change in the system, which can result in a pressure increase of:

ρh = (62.4)lb/cu.ft (5)m.(3.28 ft/m) = 1023 psf. (7.1 psi)


or in SI units:
ρhg = (998)kg/cu.m (5)m.(9.81 m/s2) = 48952 N/m2. (0.489 Bar)

The total driving pressure for a slug can be estimated from:

δP = 0.5*ρGV2 + ρLh

The first term corresponds to the dynamic pressure in the gas-flow and is responsible for the
propagation in the horizontal part in case the velocity difference between gas flow and fluid flow is
large. The second term is representing the pressure difference across the slug when the slug is in the
vertical pipe section and where “h” is the total vertical rise the slug must experience.

δP = (0.5)(0.823)[lb/cu.ft] (28692)[ft/min]2 (1/602)[min/sec]2 /32.2 [ft/s2]+ 1023 psf


= 29.2 + 1023 (psf)
= 1052 psf
= 7.31 psi

This is the pressure rise over the vapor pressure that should be used to provide the driving force for the
slug. (Note that the line pressure of 150 psig is not included in the driving pressure. Typically there is
not too much of a line pressure differential over the slug. Where the user believes that the full line δP
can form over the slug, it can be entered and the change on the loads observed.)

The vertical rise is 5m. (16.4 ft.) The head due to the elevation change is 1023 psf (7.1 psi.) The
elevation change is often the largest component of the slug driving pressure force. The vapor pressure
for the above configuration will be set at ( 7.1 + 14.7 ) = 21.8 psia. The driving boundary condition
pressure is the sum of the driving pressure difference and the vapor pressure:

Driving Pressure = (21.8 + 7.31 ) psia


= 29.11 psia
= 14.41 psig.

BOS Fluids uses gage pressure when describing boundary condition pressures, and so 14.41 psig will
be the driving slug pressure. This pressure differential is shown in the figure below.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.22
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The initial insertion of the slug into the pipeline can occur at any finite time after a start of the solution,
but one second is often used. The rise time can be estimated as described above using:

Trise = (π/2)(R+r)/V = (1.57)[ (1 ft/12 in) (18 + 6) in.] / [(2869) ft/min.] = 0.000697 min.
= 0.0418 seconds.

There is very little information available about the slug, but from the guidelines above, Tc can be
found from:
( Tc – Tb ) = L/V

Tb = 1.042 seconds. (Arbitrarily selected to be between Ta and the pressure plateau.)


L = 20D = (20)(12.75) = 255 in.
V = 2869 ft./min

Tc = L/V + Tb = (255)/[(2869)(12)/(60)] + 1.042 = 1.486 seconds.

These parameters used with a SLUGHEAD boundary condition are shown on the figure below:

Since slugs are not reflected, the total simulation time is taken as the time that it takes for the slug to
traverse the pipe section of interest. For the example considered, the total length of pipe for the slug to
travel is:

L = ( 120 + 90 + 5 ) (3.28) = 705 ft.

The approximate time that it will take for the slug to travel this distance is:

L/V = 705 / 2869 ft/min = 15 seconds.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.23
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The slug will probably travel slower than the free flow velocity of the gas, so 20 seconds will be
selected as a first pass total solution time. The total solution time can be always be adjusted if needed.
Where large numbers of intersections or changes in diameter are involved, more time will be most
likely be needed. When the model ends in a long, straight pipe run, less time will be needed.

The pressure selection procedure can be described as follows:

1) Each point in the system that should start empty must have a NOFLOW pressure less than the vapor
pressure. The vapor pressure can be set artificially high so that this is achieved.

2) The driving pressure for the slug can be given by the process or estimated from: P = 0.5ρV2 + ρLh.

3) Pressures at boundary conditions that are attached to empty pipes must be initialized at a pressure
below the vapor pressure. The lowest boundary condition pressure that can be specified is zero.

Slug modeling should be approached as a trial and error procedure whereby the user determines what
the critical system parameters are and how they affect important design loads. The slug calculator
should be run to verify that certain design rates are within reasonable ranges.

The values of interest are:

Free Flow Velocity of the Gas (No Liquid) = 2869 ft/min (47.8 fps)
Density of the Gas at Line T&P = 0.823 lb/cu.ft.
Gas Flowrate at Line T&P = 2096 cu.ft/min. (34.9 cu.ft/sec.)
Maximum Vertical Change: 5 m. (16.4 ft.)

The input cells for the slug calculator are shown below. Note that the liquid volumetric flowrate is left
blank. The program will calculate this value to produce a lower bound on the required two-phase liquid
flow needed to produce slugging in a horizontal pipe if the gas flowrate is sufficient. It is possible that
the gas flowrate is too small for the diameter of pipe to support slugging.

The “Calculate” button is pressed to produce the following output:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.24
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The slug calculator indicates that a minimum-to-maximum slug velocity will be in the 22
–to-68 ft/sec range. It is expected that the program calculated slug velocity will be in
this range. (Although the slug calculator shows that a 1163.98 ft. length is required to
accelerate a 16.4 ft. slug to a velocity of 68.05 ft./sec. Only 120 ft. is available in the
model, so it is reasonable that the maximum slug speed observed will be less than 68.05
ft./sec. The free flow velocity of the gas is given by the process as 47.8 ft./sec and
crudely estimated by the program as 58.05 ft/sec.
A “range” of anticipated slug speeds has just been calculated. The higher values are expected within
1200 ft. of the slug inlet and at the lower elevations. Lower values are expected at the inlet, after 1200
ft., at higher elevations, and after diameter changes or intersections.

The SLUGAA job cavity report for a time before the slug is introduced into the system is shown
below:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.25
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Note that the cavity % is 100.00 for each reach and internal node in the piping system. The line is
completed vacated on startup. There is no flow at 0.89 seconds because the SLUGHEAD boundary
condition does not start raising the boundary head above vapor pressure until after 1.0 second. Each
internal reach is shown in the cavity distribution list. A “reach” is an internal length of pipe needed to
properly solve the transient set of equations. The list starts at a distance of zero feet into the model and
runs the entire model length, i.e. to 703.7 ft.

The head is a collective value reflecting elevation and pressure. The head variation in a horizontal pipe
shows how the pressure distribution exists over the flowing slug length.

Since it is desireable to follow the cavity through the system a number of “time-slices” of the solution
will be printed. These are accessed from the option on the
main menu. The following input is used for the example job:

The times are chosen arbitrarily at first, somewhat evenly spaced throughout the total solution time so
that the behavior of the traveling slug can be watched during the solution.

Only a few changes are required on the output form to be sure that the large cavity capability is
activated. The applicable portion of this input is shown below:

1-The maximum simulation time is set to 20 seconds.


2-The large cavity model is set to GLUG
3-Cavities are plotted instead of the velocity (user preference)
4-Maximum pressure trapping is deactivated on output. (user preference)

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.26
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The (user preference) items selected on the optional screen shown above, make viewing and
understanding the large cavity output easier.

A user defined fluid is specified since it is the liquid properties that govern the magnitude of the
impacts. The input for water is shown below with the artificially calculated vapor pressure:

The geometry input can be read in from many typical pipe stress neutral files, or entered in BOS
Fluids. The nonzero boundary condition input required to describe the SLUGHEAD is shown below:

The boundary condition at the last system node, where the slug is to exit (node 25) is set to a fixed
head at 0.0 psig. This provides an essentially non-reflective boundary for the slug.

The ACTION form must be set to TRANSIENT with an initial steady state solution of NOFLOW: The
germain portion of this screen is shown below:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.27
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The model described is included in the BOSMODELS directory as SLUGAA.IFU and is ready to run.
A plot of the piping geometry is shown below:

The Cavity size output provides the most information for a slug flow. A truncated version of the
cavity report for the above model is shown below:
Cavity Size Distribution

Cavity Percentages at Time(sec) = 0.490 00 .. before slug introduced

ELEMENT 10 TO 25
"x" Cavity % Head Elevation Flow
0.000 100.00000 7.10378 0.00000 0.00000
43.982 100.00000 7.10555 0.00000 0.00000
307.875 100.00000 7.10555 0.00000 0.00000
351.857 100.00000 7.10370 1.50000 0.00000 …. Cavity = 100%, NO FLOW

Cavity Percentages at Time(sec) = 1.99000

ELEMENT 10 TO 25
"x" Cavity % Head Elevation Flow
0.000 0.02696 13.95359 0.00000 28.49443 … Slug beginning to move
43.982 16.51073 7.11488 0.00000 0.23674

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.28
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

87.964 99.83081 7.10555 0.00000 0.00008


131.946 99.99996 7.10555 0.00000 0.00000
571.768 100.00000 7.10370 16.40000 0.00000
615.750 100.00000 7.10370 16.40000 0.00000

Cavity Percentages at Time(sec) = 3.00000

ELEMENT 10 TO 25
"x" Cavity % Head Elevation Flow
0.000 0.01412 20.18008 0.00000 28.47187
43.982 0.02521 14.42877 0.00000 28.43780
87.964 28.21839 7.11024 0.00000 0.13647 … Slug Front
131.946 99.90517 7.10555 0.00000 0.00004
175.928 99.99998 7.10555 0.00000 0.00000
219.911 100.00000 7.10555 0.00000 0.00000

Cavity Percentages at Time(sec) = 6.00000

ELEMENT 10 TO 25
"x" Cavity % Head Elevation Flow
0.000 0.00353 21.07824 0.00000 28.49338
43.982 0.00489 21.96041 0.00000 28.12138
87.964 0.00752 21.90332 0.00000 27.92018
131.946 0.01410 20.20343 0.00000 27.86417
175.928 0.01429 7.33528 0.00000 27.78255
219.911 67.25726 7.10645 0.00000 0.01825 … Slug Front Advancing
263.893 99.99260 7.10555 0.00000 0.00000
307.875 100.00000 7.10555 0.00000 0.00000

Cavity Percentages at Time(sec) = 10.00000

ELEMENT 10 TO 25
"x" Cavity % Head Elevation Flow
0.000 0.00221 21.07824 0.00000 28.49347
43.982 0.00266 21.96041 0.00000 27.73622
87.964 0.00326 21.90332 0.00000 27.14905
131.946 0.00408 21.09364 0.00000 26.70190
175.928 0.00564 21.74401 0.00000 26.40232
219.911 0.00903 22.41485 0.00000 26.25791
263.893 0.01523 19.23173 0.00000 26.22868
307.875 0.02877 13.52295 0.00000 26.21837
351.857 33.22995 7.10926 1.50000 0.79628 …. Slug at 351 ft. down pipe
395.839 100.00000 7.10370 16.40000 0.00000
439.821 100.00000 7.10370 16.40000 0.00000
659.732 100.00000 7.10370 16.40000 0.00000
703.714 100.00000 7.10378 16.40000 0.00000

Cavity Percentages at Time(sec) = 17.99000

ELEMENT 10 TO 25
"x" Cavity % Head Elevation Flow
0.000 0.00156 21.07824 0.00000 28.49347
43.982 0.00178 21.96041 0.00000 27.31664
87.964 0.00201 21.90332 0.00000 26.30441
131.946 0.00227 21.09364 0.00000 25.42391
175.928 0.00263 21.74401 0.00000 24.67461
219.911 0.00309 22.41485 0.00000 24.06524
263.893 0.00358 20.99614 0.00000 23.56549
307.875 0.00449 22.49404 0.00000 23.17827
351.857 0.00401 7.53417 1.50000 22.74470
395.839 0.01471 19.65625 16.40000 22.46601
439.821 0.02201 15.49243 16.40000 22.46770
483.803 0.04350 11.34906 16.40000 22.46837
527.785 53.61598 7.10714 16.40000 8.65850 … Slug position @ 18 sec.
571.768 100.00000 7.10370 16.40000 0.00000
615.750 100.00000 7.10370 16.40000 0.00000
659.732 100.00000 7.10370 16.40000 0.00000
703.714 100.00000 7.10378 16.40000 0.00000

Since the slug introduction does not start until 1.0 seconds, the first time slice at 0.49 seconds shows
no flow and the entire system with 100% cavities. At the next time slice of 2.0 seconds, the slug front

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.29
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

has moved to a location that is 43.9 feet down the the pipe. After six seconds the slug appears at 219.9
feet down the pipe. The slug is moving approximately (219.9 – 43.9) ft./(6.0 – 2.0 ) sec. = 44 ft./sec,
and this is reasonable given the calculations performed above. The instantaneous flowrate at the end
of the slug is shown to be 22.46 cu.ft./sec. The volumetric flowrate can be divided by the inside area of
the pipe to get the local velocity:

[ (22.46) cu.ft/sec (144) in2/ft2 ] / [ (105.2) in2 ] = 30 ft/sec. locally.

The momentum flux force due to a slug traveling at a 30 ft./sec velocity impacting an elbow can be
found from:
ρ A V2 = [(62.4)(105)(30)2] / [(32.2)(144)] = 1271.7 lbf.

A 40 ft/sec. slug produces a load equal to:

ρ A V2 = [(62.4)(105)(40)2] / [(32.2)(144)] = 2260.8 lbf.

There are three separate sections of the line analyzed:


1) The first horizontal run before the first elbow (120m.)
2) The vertical 16.8 ft. run
3) The second horizontal run after the vertical rise (90m.)

The unbalanced force report for these three sections is shown below:

The unbalanced loads included by default include the momentum flux term (calculated above), and the
pressure term (PA). The user can turn off either the momentum flux term or the pressure term in the
optional screen. For slug loadings usually the pressure term is turned off because the pressure buildup
is not thought to contribute significantly to the unbalanced load. The option form with the PA term
deactivated is shown below along with the reduced loads that are produced:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.30
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

(P)(A) Term Deactivated

Loads with (P)(A) Term Deactivated

The slug loads are certainly in the expected range and vary as expected. The above loads can
certainly be used for static design by multiplying by a DLF of 2.0. Static designs
may not be as accurate as equivalvent dynamic designs, but a static design
consideration is certainly better than no consideration at all!
For users wishing to run transient time history analyses the unbalanced load in a section can be plotted
and sent to an ascii text file. The plot below is for the straight section that contains the node 13. Once
this plot is generated a PC13 and MC13 file will exist in a CAESAR ascii format for positive and
negative direction applications of the load. Additionally PP and MP files will exist in a PipePlus ascii
format. Other pipe stress programs can usually read one of these formatted ascii files. The user is
recommended to apply the time history load to the piping system and design accordingly. For the node
13 loading, the load will act in the axial direction along the pipe that contains the node 13. Typically
the loads are placed on the straight section just before either of the bends that attach to the system. For
slugging designs, the load can usually be placed at either elbow. Where extra caution should be
excercised, the load can be placed first at one elbow, (or change in direction), and then the other.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.31
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

SLUGAA run without pressure term in unbalanced force calculation

SLUGAA run with pressure term in unbalanced force calculation

Steam Flooding
In steam-flooding applications hot water is the prime product for pumping into wells and steam flows
over the hot water to maintain the water temperature. Since the objective is to transport water, it is
desireable to keep as much liquid in the pipe as possible. In these situations slugging is common and
the BOS Fluids slug calculations can be used to evaluate the system.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.32
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The problem:

Hot water is 75% of the pipe area. Saturated steam over top of the water is at 90 psig and 331 deg. F.

Hot Water specific volume = 0.01778 cu.ft/lb. Density = 56.24 lb/ft3


Steam specific volume = 4.2309 cu.ft/lb Density = 0.236 lb/ft3

The pipe is 10” std wall. (Ai = 78.854 in.2)

The water flow is 7.0 ft./sec

The water flow area = (78.854)(75%) = 59.14 in2.

The flowrate of water is: (7) ft/sec. (59.14)in2 / 144 = 2.875 cu.ft/sec.

Assume that the steam moves at the same velocity as the water.

Steam area = (78.854)(0.25) = 19.71 in2.

The flowrate of steam is: (7)(19.71)/(144) = 0.958 cu.ft/sec.


The slug calculator results are shown below.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.33
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The slug velocity range is quite large – from 8.62 fps to 62.65 fps. Since the line is 75% full of water,
the free flow slug velocity might be a reasonable estimate for slug velocity also. The maximum
elevation in the line is 2.0 ft and the loop elbows. The vapor pressure must be:

(2.0)ft.(62.4)/(144) = 0.867 psig. = 15.6 psia

The back pressure differential will be:

0.5ρGV2 + ρh
0.5ρGV2 = (0.5)(0.236)(34.2)2/[(32.2)(144)] = 0.03 psi
ρh = 0.867 psi

The pressure plateau used in the SLUGHEAD boundary condition will be 0.5ρV2 + ρh over the vapor
pressure = (Vapor Pressure ) + (0.5ρV2 ) + ( ρh ) = 0.867 + 0.03 + 0.867 = 1.764 psig.

The rise time for the slug introduction at the boundary will be:

(2/π) ((R+r) / V = (2/π)(15+5.1925) / [(34.22)(12)] = 0.03 sec.


The time for Tc when the SLUGHEAD boundary condition maintains the flow can be estimated from:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.34
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Tb + Trise + L/V

where Tb will be the start of opening time.


L = 10D = (10)(10.75) = 107.5 in.
V = 34.22 fps

Tc = Tb + Trise + L/V = 1.0 + 0.03 + (107.5)/(12)/(34.22)


= 1.3 sec.

The boundary condition for the SLUGHEAD entry at 10 is shown below:

The slug should travel a total of 200 + (8)(3) + (28)(2) + 80 + 100 = 460 ft.

Traveling at 34.22 fps it will traverse 460 ft. in 460/34.22 = 14.44 sec. (Use 20 sec since there is an
intersection to pass.) Use a GLUG large cavity model. The steady state will be a NOFLOW condition,
and the slug will be trapped at 0.1, 5, 10, 15 and 19 seconds. The 0.1 sec snapshot of the slug will be
taken to be sure that the lines starts fully voided. The remaining times are placed uniformly through
the solution to monitor the progression of the slug. A “USER” fluid is input so that the vapor pressure
of 15.6 psia can be entered. The BOS Fluids model is shown below:

The truncated slug cavity listing for the first run of the job STMFLOOD is shown below:
Cavity Size Distribution

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.35
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Cavity Percentages at Time(sec) = 0.10000

ELEMENT 10 TO 105
"x" Cavity % Head Elevation Flow
0.000 100.00000 0.90388 0.00000 0.00000
44.222 100.00000 0.90565 0.00000 0.00000
88.445 100.00000 0.90565 0.00000 0.00000
309.557 100.00000 0.90565 0.00000 0.00000
353.780 100.00000 0.90388 0.00000 0.00000

ELEMENT 105 TO 110


"x" Cavity % Head Elevation Flow
0.000 100.00000 0.90388 0.00000 0.00000
200.000 100.00000 0.90388 0.00000 0.00000

ELEMENT 105 TO 115


"x" Cavity % Head Elevation Flow
0.000 100.00000 0.90388 0.00000 0.00000
200.000 100.00000 0.90388 0.00000 0.00000

Cavity Percentages at Time(sec) = 5.00000

ELEMENT 10 TO 105
"x" Cavity % Head Elevation Flow
0.000 0.00724 14.40377 0.00000 17.05622
44.222 0.00974 7.84376 0.00000 16.94622
88.445 0.02851 7.38051 0.00000 16.88993
132.667 31.39098 0.90969 0.00000 0.00019
176.890 100.00000 0.90380 2.50000 0.00000
221.112 100.00000 0.90380 2.50000 0.00000
265.335 100.00000 0.90565 0.00000 0.00000

Cavity Percentages at Time(sec) = 10.00000

ELEMENT 10 TO 105
"x" Cavity % Head Elevation Flow
0.000 0.00316 14.40377 0.00000 17.05638
176.890 0.01689 11.83708 2.50000 15.95280
221.112 0.00177 6.53515 2.50000 14.30453
265.335 0.00067 1.19133 0.00000 13.32817
309.557 99.40369 0.90566 0.00000 0.00008
353.780 99.99975 0.90565 0.00000 0.00008

ELEMENT 105 TO 110


"x" Cavity % Head Elevation Flow
0.000 99.99975 0.90565 0.00000 0.00000
40.000 100.00000 0.90565 0.00000 0.00000
80.000 100.00000 0.90565 0.00000 0.00000

ELEMENT 105 TO 115


"x" Cavity % Head Elevation Flow
0.000 99.99975 0.90565 0.00000 0.00000
40.000 100.00000 0.90565 0.00000 0.00000
80.000 100.00000 0.90565 0.00000 0.00000

Cavity Percentages at Time(sec) = 15.00000

ELEMENT 10 TO 105
"x" Cavity % Head Elevation Flow
0.000 0.00242 14.40377 0.00000 17.05642
44.222 0.00263 7.84376 0.00000 16.47604
353.780 0.29119 1.53798 0.00000 15.10138

ELEMENT 105 TO 110


"x" Cavity % Head Elevation Flow
0.000 0.29119 1.53798 0.00000 7.39572
40.000 60.41387 0.90686 0.00000 0.00295
80.000 99.99520 0.90565 0.00000 0.00000
120.000 100.00000 0.90565 0.00000 0.00000

ELEMENT 105 TO 115


"x" Cavity % Head Elevation Flow
0.000 0.29119 1.53798 0.00000 7.39572

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.36
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

40.000 60.41387 0.90686 0.00000 0.00295


80.000 99.99520 0.90565 0.00000 0.00000
120.000 100.00000 0.90565 0.00000 0.00000

Cavity Percentages at Time(sec) = 18.99000

ELEMENT 10 TO 105

Cavity Size Distribution

"x" Cavity % Head Elevation Flow


0.000 0.00237 14.40377 0.00000 17.05640
44.222 0.00258 7.84376 0.00000 16.46296
221.112 0.00806 10.91969 2.50000 15.06745
265.335 0.01483 13.35549 0.00000 15.00930
309.557 0.02514 8.24924 0.00000 15.01752
353.780 0.08289 3.13163 0.00000 15.01630

ELEMENT 105 TO 110


"x" Cavity % Head Elevation Flow
0.000 0.08289 3.13163 0.00000 7.54773
40.000 0.16262 2.03936 0.00000 7.59542
80.000 23.91095 0.91153 0.00000 0.01842
120.000 99.95208 0.90565 0.00000 0.00000
160.000 100.00000 0.90565 0.00000 0.00000
200.000 100.00000 0.90388 0.00000 0.00000

ELEMENT 105 TO 115


"x" Cavity % Head Elevation Flow
0.000 0.08289 3.13163 0.00000 7.54773
40.000 0.16262 2.03936 0.00000 7.59542
80.000 23.91095 0.91153 0.00000 0.01842
120.000 99.95208 0.90565 0.00000 0.00000
160.000 100.00000 0.90565 0.00000 0.00000
200.000 100.00000 0.90388 0.00000 0.00000

The following can be observed by checking the cavity listing:

1) The vapor pressure settings are correct because the first time step at 0.1 seconds shows the system
completely voided, i.e. 100% cavities throughout.

2) At 5 seconds the slug has moved approximately 88 ft., and is traveling with a flowrate of 16.89
cuft/sec. Since the inside area of the pipe is 78.854 in2 = 0.546 ft2, the velocity at the slug front can be
estimated from 16.89/0.546 = 30.84 fps.

3) After 10 seconds, the slug is still short of the intersection because the 105 to 110 and 105 to 115
segments have not begun to show significant flow increases.

4) After 15 seconds, the slug has definitely split at the intersection. Note how the flow down each
branch is halved. This liquid flowrate can be checked in the slug calculator to determine if the slug will
be stable in the branches.

The section of pipe between the anchor and the intersection is shown above. The calculated load is
1655 lb. A static design load for the anchor due to the slug impact would be (1655)(2) = 3,310 lb.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.37
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Much smaller design loads are required in the loop piping because the slug front passes the
downstream elbow before the full momentum change occurs in the first. This is a very big benefit of
the BOS Fluids slug analysis. The BOS user does not need to assume that the slug impacts each elbow
with the same force.

The short duration unbalanced force in the loop is shown below:

The time history of the slug load on the anchor is shown below:

Slug Catchers
© Copyright 2003, Paulin Research Group & Dynaflow Engineering
5.1.38
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

There are a variety of slug catcher designs. Two primary designs involve using either one or more
large drums or vessels, or a series of parallel large diameter pipes. Either is intended to separate the
anticipated liquid column of oil that separates and/or is included in an entrained, but primarily gas
flow. The slug velocity and total anticipated volume entering the slug catcher is usually specified in
these cases and the procedure outlined above can be used to analyze the inlet slug loads, and the
distribution of the slug through the multiple outlets of the header. Nodes can be placed appropriately
in the model so that needed support loads and time histories can be extracted.

The loads in the axial direction are needed for support design, and the designer wants to be sure that
the branch sizes will adequately handle the slug from the header:

The inlet line transfers 6.5 barrels per second, and the oil specific gravity is 0.61. The inside area for
the 20” inlet line is 1.968 ft2.

(6.5) barrels/sec. (42) gallons/barrel (0.134) ft3/gallon= 36.6 ft3/sec.

The slug velocity will be: 36.6 / 1.968 = 18.6 fps.

From the buried inlet to the manifold the slug will rise 14 ft.

The pressure rise due to the 14 ft. elevation change is: (14)(62.4)(0.61) / (144) = 3.7 psi

No data on the gas flow are given. From experience however we know that the pressure term ρGV2 is a
small number for gas densities related to pressures below 1000 psi.
Therefore the term related to ρGV2 is neglected in this example.

The vapor pressure to make sure that the entire slug catcher model is drained will be: 3.7 psig or
14.7+3.7 = 18.4 psia.

The driving pressure rise is:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.39
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

(Vapor Pressure)gage + ρV2 + ρh = 3.7 + 0.0 + 3.7 = 7.4 psig.

The rise time for the pressure will be: (π/2)(R+r)/V = (π/2)(30+10) / [(12)(18.6)] = 0.28 sec.

The Tc value for the SLUGHEAD boundary condition will be found from:

Tb + Trise + L/V

where Tb will be the start of opening time. (1.0 sec.)


L = 10D = (10)(20) = 200 in.
V = 18.6 fps

Tc = Tb + Trise + L/V = 1.0 + 0.28 + (200)/(12)/(18.6)


= 2.18 sec.

The total length the slug should travel to evaluate just the inlet (since that is all that is modeled) is: 200
+ 14(1.4) + 30 + (8)(3) + 15 + 2 + 12 + 9 + 100 = 411.6 ft.

For a slug traveling 18.6 fps, this would take 411.6/18.6 = 22.1 seconds. The slug traveling through
the changes in direction will travel considerably slower in the 42” barrels and so 30 seconds will be
used as a first estimate of the total run time.

FIXHEAD boundary conditions are placed at the end of the barrels = 0.0 psig. These will serve as exit
points for the collected slug. (This boundary condition is placed here to simplify the outlet piping
configuration. Seldom is it necessary to analyze the exit piping for slug loading, but the user certainly
could if it was determined that there was a concern, possibly because of improper valving, etc.)

The BOS Fluids model is shown below and included in the \BOSMODELS folder. The jobname is
CATCHER.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.40
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Once the model is developed it is a simple matter to close one of the manifold outlet valves for
servicing the barrel, and see what affect this has on flows and loads in the system.

A condensed cavity report is shown below:


Cavity Size Distribution

Cavity Percentages at Time(sec) = 0.49975 … before slug – all empty, 100% cavity

ELEMENT 5 TO 20
"x" Cavity % Head Elevation Flow
0.000 100.00000 3.70384 0.00000 0.00000
2.035 100.00000 3.74821 0.00000 0.00000
4.070 100.00000 3.74821 0.00000 0.00000
118.022 100.00000 3.74821 0.00000 0.00000
120.057 100.00000 3.74821 0.00000 0.00000
122.092 100.00000 3.74821 0.00000 0.00000
250.289 100.00000 3.70376 14.00000 0.00000
252.323 100.00000 3.70376 14.00000 0.00000

ELEMENT 20 TO 25

Cavity Size Distribution


"x" Cavity % Head Elevation Flow
0.000 100.00000 3.70376 14.00000 0.00000
2.000 100.00000 3.70376 14.00000 0.00000
8.000 100.00000 3.70376 14.00000 0.00000

ELEMENT 20 TO 120
"x" Cavity % Head Elevation Flow
0.000 100.00000 3.70376 14.00000 0.00000
2.114 100.00000 3.70376 14.00000 0.00000
38.053 100.00000 3.70376 5.00000 0.00000
40.167 100.00000 3.70376 5.00000 0.00000

ELEMENT 120 TO 125


"x" Cavity % Head Elevation Flow
0.000 100.00000 3.70376 5.00000 0.00000
8.000 100.00000 3.70376 5.00000 0.00000
10.000 100.00000 3.70376 5.00000 0.00000

ELEMENT 120 TO 130


"x" Cavity % Head Elevation Flow
0.000 100.00000 3.70376 5.00000 0.00000
1.935 100.00000 3.70376 5.00000 0.00000
114.194 100.00000 3.70376 5.00000 0.00000
116.129 100.00000 3.70376 5.00000 0.00000
118.065 100.00000 3.70376 5.00000 0.00000
120.000 100.00000 3.70384 5.00000 0.00000

ELEMENT 100 TO 135


"x" Cavity % Head Elevation Flow
0.000 100.00000 3.70376 5.00000 0.00000
2.000 100.00000 3.70376 5.00000 0.00000
8.000 100.00000 3.70376 5.00000 0.00000
10.000 100.00000 3.70376 5.00000 0.00000

Cavity Percentages at Time(sec) = 24.99974 … slug location at 25 seconds

ELEMENT 5 TO 20
"x" Cavity % Head Elevation Flow
0.000 0.09795 49.08898 0.00000 103.89542
2.035 0.09844 48.86162 0.00000 103.89542
238.079 0.29484 18.78081 14.00000 103.89614
240.114 0.29936 18.55339 14.00000 103.89630
242.149 0.30401 18.32597 14.00000 103.89617
244.184 0.30882 18.09855 14.00000 103.89634
246.219 0.31378 17.87113 14.00000 103.89620

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.41
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

248.254 0.31889 17.64371 14.00000 103.89637


250.289 0.32418 17.41628 14.00000 103.89623
252.323 0.32965 17.18886 14.00000 103.89639

ELEMENT 20 TO 25
"x" Cavity % Head Elevation Flow
0.000 0.32965 17.18886 14.00000 77.57259
2.000 0.33222 17.08471 14.00000 77.57246
8.000 0.34016 16.77226 14.00000 77.57264

ELEMENT 20 TO 120
"x" Cavity % Head Elevation Flow
0.000 0.32965 17.18886 14.00000 26.32381
6.342 0.40438 14.69678 14.00000 26.32383
8.456 0.43744 13.86608 14.00000 26.32384
35.939 4.19197 4.76420 7.58333 26.32595
38.053 4.86854 4.61683 5.00000 26.32657
40.167 54.02693 3.78604 5.00000 26.32657

ELEMENT 25 TO 30
"x" Cavity % Head Elevation Flow
0.000 0.34016 16.77226 14.00000 51.59788
6.000 0.34380 16.63398 14.00000 51.59780
8.000 0.34503 16.58788 14.00000 51.59792

ELEMENT 25 TO 100
"x" Cavity % Head Elevation Flow
0.000 0.34016 16.77226 14.00000 25.97477
2.114 0.36260 15.96356 14.00000 25.97477
38.053 4.98028 4.59635 5.00000 25.97183
40.167 52.92916 3.78774 5.00000 25.97183

ELEMENT 30 TO 35
"x" Cavity % Head Elevation Flow
0.000 0.34503 16.58788 14.00000 25.78005
6.000 0.34595 16.55333 14.00000 25.78002
8.000 0.34626 16.54181 14.00000 25.78008

ELEMENT 30 TO 80
"x" Cavity % Head Elevation Flow
0.000 0.34503 16.58788 14.00000 25.81788
2.114 0.36784 15.78887 14.00000 25.81788
10.570 0.50010 12.59278 14.00000 25.81791
12.684 0.54949 11.79375 14.00000 25.81792
35.939 4.45611 4.70135 7.58333 25.82013
38.053 5.04177 4.58546 5.00000 25.82071
40.167 53.92323 3.78619 5.00000 25.82071

ELEMENT 35 TO 40
"x" Cavity % Head Elevation Flow
0.000 0.34626 16.54181 14.00000 -0.00001
2.000 0.34626 16.54181 14.00000 -0.00001
4.000 0.34626 16.54182 14.00000 0.00000

ELEMENT 35 TO 60
"x" Cavity % Head Elevation Flow
0.000 0.34626 16.54181 14.00000 25.78010
2.114 0.36918 15.74511 14.00000 25.78009
38.053 5.06271 4.58182 5.00000 25.78064
40.167 54.63489 3.78512 5.00000 25.78064

ELEMENT 120 TO 125


"x" Cavity % Head Elevation Flow
0.000 54.02693 3.78604 5.00000 0.68520
2.000 52.53218 3.78838 5.00000 -0.52601
8.000 62.02397 3.77543 5.00000 2.59337
10.000 44.33787 3.80402 5.00000 2.59264

ELEMENT 120 TO 130


"x" Cavity % Head Elevation Flow
0.000 54.02693 3.78604 5.00000 26.85442
1.935 53.52930 3.78680 5.00000 25.21070
120.000 100.00000 3.70384 5.00000 0.00000

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.42
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

ELEMENT 100 TO 135


"x" Cavity % Head Elevation Flow
0.000 52.92916 3.78774 5.00000 -0.86724
2.000 52.87217 3.78783 5.00000 1.31013
8.000 54.62691 3.78513 5.00000 -2.38436
10.000 51.44937 3.79016 5.00000 -2.38509

ELEMENT 100 TO 140


"x" Cavity % Head Elevation Flow
0.000 52.92916 3.78774 5.00000 25.11876
1.935 53.74685 3.78646 5.00000 26.49618
3.871 52.56614 3.78832 5.00000 25.44730
118.065 100.00000 3.70376 5.00000 0.00000
120.000 100.00000 3.70384 5.00000 0.00000

ELEMENT 80 TO 145
"x" Cavity % Head Elevation Flow
0.000 53.92323 3.78619 5.00000 0.56178
2.000 53.17217 3.78736 5.00000 -1.30926
4.000 55.08062 3.78446 5.00000 2.34802
6.000 51.72380 3.78970 5.00000 -3.97878
8.000 57.01231 3.78173 5.00000 5.31390
10.000 50.62852 3.79156 5.00000 5.31462

ELEMENT 80 TO 150
"x" Cavity % Head Elevation Flow
0.000 53.92323 3.78619 5.00000 26.31987
1.935 54.11786 3.78590 5.00000 25.53253
3.871 53.68328 3.78656 5.00000 26.41565
102.581 59.90157 3.77797 5.00000 23.48782
104.516 72.82969 3.76479 5.00000 8.83304
106.452 93.62556 3.75124 5.00000 2.71812
108.387 99.02775 3.74865 5.00000 1.69743
110.323 99.86756 3.74827 5.00000 1.55110
112.258 99.98582 3.74822 5.00000 1.53193
114.194 99.99953 3.74821 5.00000 1.53068
118.065 100.00000 3.70376 5.00000 0.00000
120.000 100.00000 3.70384 5.00000 0.00000

ELEMENT 60 TO 155
"x" Cavity % Head Elevation Flow
0.000 54.63489 3.78512 5.00000 -0.14727
2.000 53.18576 3.78734 5.00000 -0.13327
10.000 43.93643 3.80493 5.00000 1.78166

ELEMENT 60 TO 160
"x" Cavity % Head Elevation Flow
0.000 54.63489 3.78512 5.00000 25.84709
1.935 54.87726 3.78476 5.00000 25.79014
27.097 58.24404 3.78008 5.00000 27.12186
29.032 51.78264 3.78960 5.00000 25.90043
46.452 58.14099 3.78021 5.00000 24.83654
48.387 58.08346 3.78029 5.00000 24.56650
92.903 65.65377 3.77146 5.00000 24.92766
94.839 51.55785 3.78998 5.00000 33.17286
96.774 51.58434 3.78993 5.00000 22.23539
104.516 80.96780 3.75866 5.00000 5.79087
106.452 96.44064 3.74985 5.00000 2.16482
114.194 99.99994 3.74821 5.00000 1.53068
116.129 100.00000 3.70376 5.00000 0.00000
118.065 100.00000 3.70376 5.00000 0.00000
120.000 100.00000 3.70384 5.00000 0.00000

At 0.5 sec., before the slug is introduced the cavity report shows that all reaches in the system are
empty, i.e. 100% cavity.

At 25 seconds, after the slug is setup. (The model in the \BOSMODELS\ folder, named: CATCHER
prints slug locations at intermediate times to that the user can investigate the slug distributions.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.43
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

At 25 seconds, the flow established is 103.89 ft3/sec. The design flowrate is 36.6 ft3/sec. The driving
pressure is producing too large slug velocity.

A look at the maximum pressure and flowrate report shown below also indicates that there was a
stability problem during the movement of the slug front. The maximum pressure is well in excess of
the boundary value prescribed and the maximum flowrate in the inlet 20” run is shown to be 120.8
cuft./sec. The design flowrate to produce the slug rated barrels per second is only 36.6 cuft/sec?

Since the pressure is proportional to the square of the flowrate, an adjusted pressure rise can be used to
produce the desired slug flow rate:

Pi = (Qi+1 / Qi)2 (Pi) = [ (36.6)/(103.89) ]2 (10.45-3.7) = 0.8377 psi.

This is the pressure rise over the vapor pressure, and so the next pressure to try is

3.7+0.8377 = 4.54 psig

In addition, the value K1 on the optional data form can also be decreased to 0.01 instead of 0.1. This
will tend to prevent the “reverberations” observed in the time history pressure and flow plots. Also the
PA term should be set to zero since there is no anticipation that a pressure buildup will occur. It is
when the user needs to start adjusting the stability criteria that the output should be studied carefully.
The new jobname is CATCHER2.

The Integrated Nodal Flowrate report (shown below) can be used to determine the collective
distribution of flow through the manifold take-offs. As can be seen in the printout below, the first
intersection at 20 receives 2503 ft3 after 30 seconds, while the next intersection 25 only receives 1867
ft3. The flow is split based on pressure and respective flow area. SPLASH should be used to
determine how the momentum and directionality of the flow condition effects the distribution of the
slug stream.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.44
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Every effort has been made to establish reasonable limits for stability criteria that will work in the
majority of situations. This is certainly not possible for all geometries and flowrates however, and so
the user must be sure that the solution calculated seems reasonable upon inspection.

The slug calculator should be consulted for this problem:

The elevation potential is for the 14ft. elevation present in the inlet pipe. The minimum slug velocity
reported is 21.10 fps, greater than the design slug in the problem. It might be that once the slug gets
onshore, that it is no longer provides a full section flow. Further inspection is certainly warranted as
the given parameters produce a result that is unstable in a solver tuned for typical slug flow properties.
The SPLASH analysis of this problem suggests that the flow distribution in the header will be altered
by the directional tendency of the flow to bypass a side opening in favor of continuing straight on.
Time sequences from the splash model “SlugCatcher” included in the \BOSMODELS folder are
shown below:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.45
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Pump Startup into Empty Systems


This simulation is typically used for firewater systems where the pump is turned on into an open or
nitrogen filled system and the intent is to get water to the fire as soon as possible. Piping may be
fiberglass reinforced and not suffer pressure spikes well or the loads due to the sudden influx of water
may need to be analyzed. In either case starting a pump into an open system produces primarily
momentum change loads similar to the slug loadings encountered in the first three examples. These
loads can be analyzed using BOS Fluids.

The BOS Fluids model to be analyzed is named FIREWTR and is included in the \BOSMODELS
folder under the installation directory. A 30 psi pressure head is to be added to a suction side pressure
and discharged into the vertical riser, up 60 feet, through a control valve and over to a fire ring.
Nozzles on the ring can be modeled using a variety of techniques. Small valve models are effective
when simulating nozzles because the user can adjust the opening of the valve to control the water
delivery rate and to ultimately determine the type of nozzle to be employed. PSV’s could be used to
employ the opening and closing characteristic of that valve, but in the example FIREWTR job, small,
LUMPED elements were used to simulate the open nozzle and simplify the model.

There are several keys to pump startup into an empty system:


1) The pump startup time is primarily pump controlled, but may also be materially affected by adverse
suction and discharge pressures. The user is responsible for determining the correct startup time and
entering it into the BOS Fluids transient pump data form.
2) Most of the downstream section of pipe must be voided, but the pump discharge nozzle and first
reach must be below water so that the pump is effectively “primed.”

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.46
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

3) The suction side of the system must be above the system vapor pressure. This might require a shift
in the effective problem pressures, but only requires adding the same pressure shift to each node so that
the vapor pressure and head differences can be maintained.

An annotated BOS Fluids plot of the “firewtr” system is shown below:

Calculations applicable to the FIREWTR problem:

The maximum elevation from the inlet pipe to the pump to the top of the riser is 3+20+20+20 = 63 ft.
This represents a pressure head of:

(63) ft. (62.4) lb/cu.ft (1/144) = 27.3 psi.

Vapor pressure in psi (absolute) = 27.3 + 14.7 = 42 psia. = 27.3 psig (gage)

This will be the system artificial vapor pressure and will essentially be the system “zero” boundary
condition. (Or any boundary condition pressure below 27.3 psig will be an essential zero boundary
condition.

The pump takes seawater at three feet of suction head and adds 30 psi to it, to overcome the elevation
change (26 psi worth) and to supply firewater (4 psi worth). The nodes from 5 to 10 describe the
suction inlet pipe 1-0 long in the vertical direction. From the pump suction to the discharge is 2-0 in

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.47
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

the vertical direction. The pump will be submerged to satisfy the NPSH of 3-0 at the suction flange
(node 10). The following diagram illustrates:

The boundary condition at 5 will be a gage pressure equal to a 4 ft. head of water column = (4)(62.4) /
(144) = 1.73 psi. This will be added to the vapor pressure of 27.3 = 27.3 + 1.73 = 29.03 psig. The
LUMP element nozzles at the ring header appear as shown below:

These nozzles are the only other boundary conditions in the firewater system. The boundary condition
set at these nozzles is 0.0 psig. Since the artifical vapor pressure is 27.3 psig, the nozzles will have a
boundary condition pressure of 27.3 psig (or anything smaller). Relative to the pipe at its present
elevation, any positive pressure internal to the pipe greater than 27.3 will produce a flow out of the
nozzle. The LUMP elements are given a diameter and wall thickness so that the ID is 0.726 in., which
is the open flow area for the spring loaded nozzle. It is assumed that the most significant momentum
change at the nozzles will not come by the force to lift the light nozzle spring, but by the reduction in
flow area from the 4” ring header to the 0.726 in. nozzle ID.

Summarizing:

The artificial vapor pressure (gage) = 27.3 psig.


The artificial vapor pressure (absolute) = 42 psia

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.48
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The pump suction inlet pipe pressure boundary condition = 1.73 + 27.3 = 29.03 psig
The nozzle outlet pressure boundary condition = 0.0 + 27.3 = 27.3 psig.

The pressure head specified for the pump is the relative pressure head from the suction to the discharge
flange and so no adjustment needs to be made. These pressure boundary conditions are shown below.

The flow through one of the nozzles appears as shown below.

These inputs are included in the model FIREWTR included in the \BOSMODELS folder.

Flashing:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.49
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Fluid flashing downstream of a relief valve on a process vessel is a function of the contents and size of
the vessel, any chemical reaction, and both the upstream and downstream sizes, lengths, pressures and
temperatures.

In long or flexibly supported vent stack lines the change in pressure and mass flow that can occur in
bubbly and slug flow regimes when flashing occurs can have a significant effect on the impulse loads
developed on the piping elbows and supports. It is not uncommon for pipe elbows to be opened and
supports to be bent when liquids having boiling points at or below atmospheric conditions are vented
due either to accidental overpressure of the containment vessel or pipe. Such conditions most
commonly occur due to failed sensors, flow or pressure cutoff switches, or due to reactions because of
impurities or other unexpected product combinations in the flow.

When a liquid flashes, there is an accompanying pressure drop and volumetric flowrate expansion that
can produce sound waves in the downstream gas passages. These sound waves can resonate and
produce mechanical vibration in concert with the pressure and moment change forces that exist due to
the rapid increase in flow downstream of the flashing valve.

Common storage and transportation systems that may experience flashing problems include:

1) Chlorine
2) Propane, Butane, Methane
3) Ethylene Oxide
4) Amonia
5) LPG
6) Cyclohexane/hexane/Tolulene

In general, the process of flashing will be stochastic, but will occur in a recognizable vicinity, (i.e. just
downstream of the relief valve or volume change). The net volumetric flowrates to be used in
mechanical load calculations can be expressed by the following equation:

Qd = (Qu) (Cf)

Qd – discharge flowrate used in BOS Fluids (cu.ft./sec) or (cu.m/sec)


Qu – upstream flowrate (typically the discharge from the relief valve.)
Cf – Flashing or Reaction constant

These properties are presented in the figure below:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.50
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The constant Cf can be found for the flashing of a liquid into a gas from the following equation:

Heat content of the fluid per m 3


Cf =
Heat of vaporization per m 3

ρ fl • c s • ∆T R • TVLE
Cf = •
H vap p vapour
where:

ρfl = density of the fluid to flash [kg/m3]


cs = specific heat of the applicable fluid [kJ/kg°K]
TVLE = Vapor Liquid Equilibrium temperature at the specified pressure [°K]
(From Process Engineer or Handbook of Chemistry & Physics)
∆T = Difference between the fluid temperature and the TVLE [°K]
Hvap = Heat of vaporization of the applicable fluid [kJ/mol]
R = Universal Gas Constant = 8314.5 [J/mol°K]
pvapor = Pressure of the flashed vapor [N/m2]

The flashing constant Cf is equal to the ratio the specific volume (volume per kg or per mol) of the
flashed species before and after the flashing process. Cf is related to the specific heat and the enthalpy
of vaporization of the applicable fluid.

Where the mechanism that produces the increased downstream flowrate is a chemical reaction, the user
should note that the process engineer can typically supply the steady state values for the upstream and
downstream volumetric flowrates, i.e. Qd , and Qu. Where the downstream flow represents the reaction
products, and the upstream flow represents the products reacting. From these values the coefficient Cf
can be found from:
Cf = Qd / Qu

Cf is entered directly in BOS Fluids on the control valve data form. The BOS
Fluids flashing model flashes the fluid inside of the CONTROL VALVE model,

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.51
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

and the results are seen downstream of the valve. The orientation of the valve
determines the upstream and downstream direction for the fluid flashed.

Flashing is NOT a large cavity simulation, although some characteristics of large


cavity behavior can be observed.
The procedure demonstrated in the example below simulates hexane flashed from a temperature of 150
degC to a reservoir pressure of 2 bara:

By means of the data provided in the Handbook of Chemistry and Physics, Section 6, the Vapor Liquid
Equilibrium temperature for hexane at an absolute pressure of 2 Bara is found to be 364.3° K.

Hence ∆T, the difference between the fluid temperature and the TVLE: ∆T = 58.85 [°K]

The value for the specific heat of Hexane is found to be: cs = 2.260 [kJ/kg°K]
The Heat of vaporization of Hexane is: Hvap= 28850 [kJ/mol]
The density of Hexane as a fluid is: ρfl= 656.3 [kg/m3]

The result for the flashing constant Cf for Hexane when subject to the indicated conditions is:

Cf = 656.3* 2.260 * 58.85* 8314.5*364.3 /(28850*200000) = 45.8

The geometry of the system downstream of the relief valve is shown below:

If the data is not available from the process engineer, the mass flow from a pressurized vessel can be
conservatively estimated as:

w = (C) (A) [ (2gρ)(P-PA) ]1/2

w = mass flow in lb/sec.


C = discharge coefficient (0.6)
A = discharge area (sq.ft.)
g = gravitational constant 32.2 ft/sec2

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.52
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

ρ = source liquid density lb/cu.ft


P = absolute source pressure (psf)
PA = absolute ambient pressure (psf)

For the hexane relief example, the orifice at node 5 has an inside diameter of 2.02 in. with a
corresponding area of 3.205 sq.in. (0.0226 sq.ft.) The absolute source pressure is ( 290+14.7 ) (144) =
43,877 psf., and the ambient pressure is (14.7)(144) = 2116.8 psf.

The mass flowrate is found from:

w = (0.6)(0.0226)[ (2)(32.2)(40.95)(43,877 – 2116.8) ]1/2


= 142 lb/sec.

The volumetric flowrate is the mass flowrate divided by the density:

Q = w / ρ = 142 / 40.95 = 3.47 cu.ft./sec.

The effective downstream flowrate for mechanical load estimates is found from:

Qd = (Qu) (Cf)

The density of the downstream gas can be used to improve the characteristics of the flashed pressure
waves. The flashed gas density can be found from the equation:

ρg = mp/RT

“m” is the molecular weight of hexane – C6H14 = (12)(6) + 14 = 86 kg/mole.

ρg = mp/RT = [(86) kg/mole (200000) N/m2] / [ (8314.5) Nm/moledegK (364.3) degK ]


= 5.67 kg./m3

The ratio of specific heats (Cp/Cv) for the hexane and the viscosity is also used to describe the flash
gas properties. These will be taken as:

(Cp/Cv)Hexane = 1.4
ViscocityHexane = 0.01 cp

And so all quantities used in the effective downstream model are available:

Estimate the time that the flow will take to travel the length of the vent pipe. The 6” standard wall pipe
area is: π/4 (6.0652) = 28.89 sq.in. (0.2 sq.ft.)

The low and high side velocities will be:

VLO = (3.47) / (0.2) = 17.35 ft/sec.


VHI = (3.47)(45.8) / (0.2) = 794.6 ft./sec.

The total length of pipe from the exit of the vent stack to the entrance to the 14” collection header is:

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.53
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

210 + 95 + 80 + 170 = 555 in. (46.25 ft.)

The time it will take for the interface to travel from the exit of the vent stack to the collection header
will be:
TLO = (46.25) / (17.35) = 2.66 sec.
THI = (46.25) / (794.6) = 0.0582 sec.

Run the simulation for 4 seconds after the valve is completely opened and then inspect the solution to
see if the desired flow has been established. The total solution time can always be adjusted.

Notes for starting the BOS Fluids calculation:

1) The boundary condition on the header at nodes 35 and 40 will be at 14.5 psig (1 barg).
2) A TIMEFLOW boundary condition will be provided at the entrance to the control valve used to
simulate the orifice of the RV and to produce a “flashed” flowrate. The timeflow will start at 0.0
cu.ft./sec at time = 0.0 sec. and ramp up to 3.47 cu.ft./sec. (see calculation above), after 0.2 seconds, a
typical opening time for relief valves. (Typical RV opening times are between 0.1 and 2 seconds.)
3) The downstream piping will be defined as a FLASH type pipe. (See the pipe input form. This will
be for the pipe from nodes 5 to 40.) The large cavity model will be set to FLASH, and Cf = 45.8 will
be specified as the “flashing” constant on the control valve data sheet.
4) The steady state solution will be set as NOFLOW.
5) The orientation of the control valve should be such that the flashing occurs at
the downstream “TO” end of the element. This is true for both vortex shedding
and for flashing downstream of the control valve type.
The inlet piping configuration is shown below:

The BOS Fluids model is shown below: (The input file is flasht03.ifu in the <installation>\bosmodels
folder.)

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.54
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Germane inputs are given below:

Column Separation Model: CAP


Action: TRANSIENT
Initial Flow State: NOFLOW
Total Solution Time: 4 sec.
Deactivate Maximum Pressure Trapping: YES
Large Cavity Option: FLASH
Elements from 10-to-40 are called out as FLASH pipe element types.
On the Column Separation and Flashing form the following was entered:
Flashed Gas Weight Density = 3.0 lb./cu.ft
Flashed Gas Mean Pressure = 4.0 psig.
Flashed Gas Viscosity = 0.02 cp
Flashed Gas cp/cv = 1.4
Fixhead Boundary Condition @35 and 40 Steady State Pressure = 14.5 psig
TIMEFLOW Boundary Condition @100 Steady State Flowrate = 0.0 cu.ft/sec.
Time (sec) Flowrate
(cuft/sec)
0.0 0.0
1.0 0.0
1.2 3.47
10 3.47
Fluid: USER, LIQUID
Weight Density: 40.95 lb/cu.ft.
Viscosity: 0.36 cp
Bulk Modulus: 147,000 psi
Vapor Pressure (ABS) 3 psia.

Cavity Profile Times (sec): 0.5 1.4 1.8 2.2 2.6 3.0 3.4 3.8
The number of reaches printed per report is 10

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.55
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The Cavity profile report for a few of the time steps is shown below.
Cavity Percentages at Time(sec) = 1.09964

ELEMENT 100 TO 1
"x" Cavity % Head Elevation Flow
0.000 0.39602 70.72949 2.25000 1.71673
3.000 0.43574 63.21669 6.25000 1.68818
6.000 0.65361 55.27021 6.25000 1.68555

ELEMENT 5 TO 30
"x" Cavity % Head Elevation Flow
0.000 0.05628 4.01985 6.25000 1.68555
4.975 0.07045 19.46989 6.25000 27.27611
9.950 0.08917 12.34212 6.25000 17.22507
14.925 0.10625 8.03190 6.25000 10.91596
19.900 0.15530 0.92245 6.71154 0.00000
24.875 0.15625 0.82840 11.22561 0.00000
29.850 0.15688 0.76713 14.16667 0.00000
34.825 0.15683 0.77194 13.93590 0.00000
39.800 0.15597 0.85665 9.86973 0.00000
44.775 0.15493 0.95946 4.93487 0.00000
49.750 0.15390 1.06227 0.00000 0.00000

ELEMENT 30 TO 40
"x" Cavity % Head Elevation Flow
0.000 0.15390 1.06227 0.00000 0.00000
3.256 0.15390 1.06227 0.00000 0.00000
29.302 0.15390 1.06227 0.00000 0.00000
32.558 0.15390 1.06227 0.00000 0.00000

ELEMENT 30 TO 35
"x" Cavity % Head Elevation Flow
0.000 0.15390 1.06227 0.00000 0.00000
3.256 0.15390 1.06227 0.00000 0.00000
6.512 0.15390 1.06227 0.00000 0.00000
29.302 0.15390 1.06227 0.00000 0.00000
32.558 0.15390 1.06227 0.00000 0.00000

Cavity Percentages at Time(sec) = 1.29964

ELEMENT 100 TO 1
"x" Cavity % Head Elevation Flow
0.000 0.06150 519.06997 2.25000 3.46966
3.000 0.06174 516.98331 6.25000 3.46464
6.000 0.09262 516.02938 6.25000 3.46021

ELEMENT 5 TO 30
"x" Cavity % Head Elevation Flow
0.000 0.02350 37.68108 6.25000 3.46021
4.975 0.02665 75.25475 6.25000 74.02810
9.950 0.03039 64.22123 6.25000 67.83384
14.925 0.03454 54.76249 6.25000 61.88541
19.900 0.03912 46.65032 6.71154 56.20398
24.875 0.04600 37.50801 11.22561 47.28308
29.850 0.05345 30.26555 14.16667 39.67004
34.825 0.06150 24.41108 13.93590 32.99742
39.800 0.07009 19.64221 9.86973 27.09581
44.775 0.09863 9.77006 4.93487 28.71889
49.750 0.13741 2.92714 0.00000 26.49887

ELEMENT 30 TO 40
"x" Cavity % Head Elevation Flow
0.000 0.13741 2.92714 0.00000 13.24913
3.256 0.14011 2.59187 0.00000 10.87284
6.512 0.14639 1.85919 0.00000 5.66653
9.767 0.15390 1.06282 0.00000 0.00387
13.023 0.15390 1.06227 0.00000 0.00000
16.279 0.15390 1.06227 0.00000 0.00000
19.535 0.15390 1.06227 0.00000 0.00000
22.791 0.15390 1.06227 0.00000 0.00000
26.047 0.15390 1.06227 0.00000 0.00000

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.56
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

29.302 0.15390 1.06227 0.00000 0.00000


32.558 0.15390 1.06227 0.00000 0.00000

ELEMENT 30 TO 35
"x" Cavity % Head Elevation Flow
0.000 0.13741 2.92714 0.00000 13.24913
3.256 0.14011 2.59187 0.00000 10.87284
6.512 0.14639 1.85919 0.00000 5.66653
9.767 0.15390 1.06282 0.00000 0.00387
13.023 0.15390 1.06227 0.00000 0.00000
16.279 0.15390 1.06227 0.00000 0.00000
19.535 0.15390 1.06227 0.00000 0.00000
22.791 0.15390 1.06227 0.00000 0.00000
26.047 0.15390 1.06227 0.00000 0.00000
29.302 0.15390 1.06227 0.00000 0.00000
32.558 0.15390 1.06227 0.00000 0.00000

The maximum pressure/flow/load report is shown below:


Pressure in : psi (Max = 1212.89 Min = -2.09 )
Flowrate in : cu.ft./sec.
Distances in : ft
Cavity size in : Percent (Max = 2.0052% )
Forces in : lb

From 100 max pres.= 1212.892 @ 0.000 from node 100


To 105 min pres.= 12.723 @ 4.000 from node 100
max flowrate = 3.4783 @ 4.000 from node 100
min flowrate = 0.0000 @ 4.000 from node 100
max cavity = 1.3368 @ 4.000 from node 100
max force = 93743.

From 105 max pres.= 1210.797 @ 0.000 from node 105


To 1 min pres.= 12.723 @ 2.000 from node 105
max flowrate = 3.4788 @ 2.000 from node 105
min flowrate = 0.0000 @ 2.000 from node 105
max cavity = 2.0052 @ 2.000 from node 105
max force = 93743.

From 5 max pres.= 138.564 @ 0.167 from node 5


To 10 min pres.= 0.932 @ 0.000 from node 5
max flowrate = 80.7327 @ 0.667 from node 5
min flowrate = 0.0000 @ 0.000 from node 5
max cavity = 0.1557 @ 0.000 from node 5
max force = 2982.

From 10 max pres.= 133.827 @ 0.000 from node 10


To 15N min pres.= 0.932 @ 10.856 from node 10
max flowrate = 80.9343 @ 16.750 from node 10
min flowrate = 0.0000 @ 0.000 from node 10
max cavity = 0.1552 @ 0.931 from node 10
max force = 2982.

From 15N max pres.= 87.387 @ 0.000 from node 15N


To 15F min pres.= 0.916 @ 2.000 from node 15N
max flowrate = 80.9726 @ 2.000 from node 15N
min flowrate = 0.0000 @ 0.462 from node 15N
max cavity = 0.1554 @ 2.000 from node 15N
max force = 2982.

From 15F max pres.= 81.815 @ 0.000 from node 15F


To 20N min pres.= 0.783 @ 6.417 from node 15F
max flowrate = 81.0853 @ 6.417 from node 15F
min flowrate = 0.0000 @ 4.226 from node 15F
max cavity = 0.1567 @ 6.417 from node 15F
max force = 832.

From 20N max pres.= 64.204 @ 0.000 from node 20N


To 20F min pres.= 0.767 @ 2.000 from node 20N
max flowrate = 81.1216 @ 2.000 from node 20N
min flowrate = 0.0000 @ 1.077 from node 20N
max cavity = 0.1569 @ 2.000 from node 20N
max force = 492.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.57
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

From 20F max pres.= 58.661 @ 0.000 from node 20F


To 25N min pres.= 0.767 @ 5.167 from node 20F
max flowrate = 81.2027 @ 5.167 from node 20F
min flowrate = 0.0000 @ 4.227 from node 20F
max cavity = 0.1569 @ 1.253 from node 20F
max force = 492.

From 25N max pres.= 44.731 @ 0.000 from node 25N


To 25F min pres.= 0.767 @ 0.000 from node 25N
max flowrate = 81.2294 @ 2.000 from node 25N
min flowrate = 0.0000 @ 0.154 from node 25N
max cavity = 0.1569 @ 0.000 from node 25N
max force = 1200.

From 25F max pres.= 39.308 @ 0.000 from node 25F


To 30 min pres.= -2.087 @ 13.417 from node 25F
max flowrate = 81.6276 @ 13.417 from node 25F
min flowrate = 0.0000 @ 13.417 from node 25F
max cavity = 0.1930 @ 13.417 from node 25F
max force = 1200.

From 40 max pres.= 14.500 @ 0.000 from node 40


To 30 min pres.= 1.062 @ 0.000 from node 40
max flowrate = 0.0000 @ 0.000 from node 40
min flowrate = -63.9049 @ 0.000 from node 40
max cavity = 0.1930 @ 0.155 from node 40
max force = 0.

From 30 max pres.= 14.500 @ 0.000 from node 30


To 35 min pres.= -2.088 @ 0.155 from node 30
max flowrate = 63.9049 @ 33.333 from node 30
min flowrate = 0.0000 @ 33.333 from node 30
max cavity = 0.1930 @ 0.155 from node 30
max force = 0.

The integrated nodal flowrate report is shown below:


Integrated Nodal Flowrates

Node Total Volume

100 10.0603
1 10.0521
5 10.0521
30 207.2161
40 103.8029
35 103.8029

The integrated flowrate report gives the total flow passed by a given node during the integration
window specified. (Default – the entire solution time period.)

The maximum unbalanced load is found in the first horizontal section and is 492 lb. The load can be
estimated from:

Force = (w)(V) + (P)(A)

where: w = 142 lb./sec. (earlier calculation)


V = 17.3 ft/sec. (earlier calcualation)
P = 8 psi (from cavity profile)
A = 28.89 sq.in. (from earlier calculation)

Force = (142) lb/sec (17.3) fps / 32.2 + (8) psi (28.89) sq.in.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.58
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

= 76 lb. + 231 lb.


= 307 lb.

A look at a time history plot of the unbalanced load appears below:

There are a number of peaks in the unbalanced load plot probably because of pressure spikes. Once the
flashing calculation is performed we want to look at reducing the spikes so that a more uniform
unbalanced load curve can be generated. User’s should be careful extracting maximum load values
from the tabular results without inspecting the transient nature of the load to be sure that the proper
calculation is made.

To begin the flashing calculation the GLUG option is changed to FLASH on the optional data form.
This model change is reflected in the input file flash003.ifu.

As with most dynamic loadings, if the load application time is short, the net effect on the mechanical
system is small. If the load application time is long, then the effect on the mechanical system can be
greater. The BOS Fluids flashing model is intended to provide a first order estimate of the loads and
flows that can develop in a transient manner downstream of a relief system in a single or multiple relief
configuration where all valves are tied into a common header.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.1.59
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

Vortex Shedding at Tees and Control Valves


In steady state gas and liquid flows pulsations can originate from side branches that have little or no
flow and from control valves. BOS Fluids computes the flow excitations when these components exist
and will add them to an already running transient analysis if the user so directs. These flow
perturbations are illustrated in the figure below:

The pulsation magnitudes and frequency generation methods are taken predominantly from the
following two papers:

“Design Stage Acoustic Analysis of Natural Gas Piping Systems in Centrifugal Compressor Stations,”
by L.E. Rogers, and “A Computational Approach for flow-acoustic coupling in Closed Side
Branches,” by Paul Radavich and Ahmet Selamet.

Quoting from the Rogers’ paper: “Traditionally, the occurrence of flow generated pulsation problems
in centrifugal compressor piping was not excessive and could be addressed by post-design remedial
action. A trend toward increased flow rates in the late 80’s has lead to generally higher gas velocities
and more complex piping configurations. These conditions in turn have lead to flow generated
pulsation that is stronger and occurs over wider ranges of frequency. Such pulsation excites acoustic
resonances of the piping and results in excessive vibration and other pulsation related problems. The
key objective of design stage acoustic analysis is to eliminate predicted pulsation driver problems.”

This is exactly the objective of the BOS Fluids vortex shedding transient feature. It is designed so that
the user can start the transient and then switch on and off the pulsations due to vortex shedding at
stagnant branch pipes and/or the pulsations due to vortices caused by partially closed control valves.
The input for these features is found on the optional data form. The default parameters are shown
below.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.2.1
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The default values shown above leave the vortex shedding pulsations inactive. The user must make
changes to the values above to get the automatic vortex calculations activated. There are two types of
vortex shedding pulsations described by the fields above. The first two date input lines control the
vortex shedding at tee branches. The third and fourth input lines above control the vortex shedding at
CONTROL valves. Each will be discussed separately.

Start/Stop Time for Tee Vortices – Vortices from tee branches occur during the entire time the flow in
the header is passing an essentially stagnant branch. (The vortex shedding routine ignores the
calculation for any time step where the branch flowrate is greater than the header flowrate, for in these
cases the mechanism to produce vortex shedding is not present.) Large pulsations can result when the
quarter wave acoustic resonance of the branch is in tune with the Strouhal “shedding” or “separation”
frequencies associated with flow past the opening. Usually, the BOS Fluids user wants to observe the
effect of these pulsations on the steady state flow system, and if they are significant then attenuating
measures need to be explored. To see the effect of the vortices on the transient solution, it is
convenient then, to be able to turn on and off the effects through time to be able to observe the change
in the pulsations in the system. This is exactly what the start and stop time for the tee vortices do. For
a particular system, the vortex shedding is started at 3.0 seconds and stopped at 4.0 seconds. After 4.0
seconds, the resonant pulsations created at the open-closed quarter wave resonator, reflect for some
considerable time after the pulsations have stopped. This behavior is shown in the time history plot
below.

A frequency transform of this plot is shown below. Note how the major Strouhal frequencies are
shown, and that the first is the strongest, as it is coincident with the first resonant frequency of the
closed branch attachment.

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.2.2
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

The second line is also used with tee/branch vortex shedding and gives the radius factor and the
tolerance. The Strouhal frequencies for flow past branches is effected by the radius of curvature of the
fitting used to make the branch. Larger radii reduce the Strouhal frequency. The redius factor is used
with each tee found in the piping system to compute an equivalent radius. The factor entered is
multiplied by the inside diameter of the branch to get the branch radius used in the Strouhal frequency
calculation. A value of zero may be used. The tests run by Rogers and others at Nova varied from the
equations used in some instances. About plus or minus 5% was experienced. The user can enter 1.05 or
0.95 to test the full service range of the shedding equation to be sure that any possible side branch
resonance is avoided.

Typical values for the first two lines in the vortex shedding data input for tees are:

3.0 4.0
0.05 0.95

The first line would tell BOS Fluids to allow pulsation energy for flow by a stagnant branch to exist
between 3.0 and 4.0 seconds during the simulation. BOS Fluids finds the geometric branches and
headers and makes the pulsation calculations according to Rogers.

To have the vortex shedding at tees active for the entire simulation, then the second or “ending” time
should be set to a value in seconds that is greater than the total time for the simulation. The first value
should be left zero as this will get the tee vortex shedding calculation to start from time equal to zero.

The third and fourth lines in the vortex shedding data are for control valves. The third line contains
the starting and stopping time for the control valve vortex shedding. This is used just like the data for
tees. The user can turn on and off the vortex shedding at control valves so that the user can see its
effect on the transient model. Control valves must be entered into the BOS Fluids model before the
program can calculate the pulsation that is emanating away from the valve. The strength of the
pulsation is as calculated by Rogers, but is a function of the flow and pressure drop across the valve.
The basic equation used is:

Flow Disturbance Magnitude = (k1)(Q)(dP/P)k2

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.2.3
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

where Q = the flow through the valve,


k1 = a constant for the valve.
dP/P = the pressure drop across the valve divided by the mean pressure
k2 = a constant for the valve.

The user enters k1 and k2 on the fourth line of the vortex shedding input. The values found by Nova
from their tests are:

k1 k2
Ball, V-Ball Valve: 0.04000 0.9768
Globe, Cage Plug: 0.00245 0.4363

The CONTROL valve type was added to be used with this vortex shedding feature. The valve
otherwise performs just like a gate, globe or ball valve, and the user should enter the desired discharge
coefficient, percentage of opening and valve inside diameter if it is different than the attached pipe.
Valves other than control valves will not have vortex shedding pulsations generated. Pulsations can
also be generated downstream of orifice plates. The user worried about his condition can model the
orifice plate as a CONTROL valve that is partially open. The remainder of the analysis is the same.

The excitation is produced downstream of the valve. This is at the second node in the valve
description as it is described on the pipe element form. The flow should be positive going from the
first to the second node (i.e. the FROM to the TO node of the element) to get the best performance.
Note that the control valve element also has the ability to flash the liquid stream entering the valve.
The flashing and the vortex shedding feature are designed to be used together if necessary, but the user
is encouraged to carefully verify the results.

Excitation from the flow past tees will only manifest itself into large pressure disturbances when there
is a resonance between the quarter-wave branch stub and one of the Strouhal frequencies developed
due to flow separation past the branch.

Control valve pulsations however, tend to exist in a broad band fashion across frequencies ranging
from zero to two-hundred hertz, and it is indeed this broad-banded excitation that is replicated in BOS
Fluids. Since flow and pressure drop are the maximum contributors to control valve pulsations, any
problem usually occurs during the highest flow through the valve, and when the valve is throttled back
from a fully opened position.

Rogers makes a few other comments that should be of general interest to the BOS Fluids user:

“pulsation which is generated in the suction system can transmit through the [centrifugal] compressor
into the discharge system and vice-versa. … a centrifugal compressor can amplify or attenuate
pulsation transmitted through it depending on the operating conditions. (This can also be true for
centrifugal pumps, which can be at pressure or velocity nodes in the suction-discharge piping system.)

“The most effective approach to solving acoustically driven problems in centrifugal compressor piping
is to alter the source of pulsation directly. This is in contrast to reciprocating compressor piping, where
effective solutions involve decoupling the source from the piping because the source of pulsation is
largely unalterable by the analyst.”

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.2.4
BOS Fluids Version 4.111 September 25, 2003 www.paulin.com

“Tee Strouhal vortex shedding problems can often be solved by the insertion of a vortex spoiler into
the sidebranch. A suitable change in the sidebranch length eliminates coincidence of excitation and
response frequencies. “

“Reduced pulsation levels from control valves can be obtained directly in two ways. One way includes
various types of modifications, including silencers and noise reducing trims that stage the pressure
drop across the valve. Splitting the flow between two or more valves in parallel is another way to
reduce pulsation.”

“The sound pressure level (SPL) is checked for the worst combination of flow and pressure drop to
determine whether detailed acoustic modeling is required. Experience has shown that valves with
maximum SPL below 95 dBa will not result in excessive vibration for properly supported piping.”

“Acoustic analysis is restricted to frequencies below 200 Hz. Experience indicates beam (lateral)
modes of piping vibration are not significantly excited at freqencies above 200 Hz. Significant
response frequencies for equipment sensitive to pulsation (for example metering devices and relief
valves) have also been found to occur below 200 Hz.)

© Copyright 2003, Paulin Research Group & Dynaflow Engineering


5.2.5

You might also like