Department of Chemical Engineering

TKP4171 P ROJECT DESIGN

Report
"How many stages should be used in a distillation column"

Written by:
Hilde E KRHEIM
ekrheim@stud.ntnu.no

Mahnaze E SMAEILPOUR A BARDEH

mahnaze@stud.ntnu.no

Tor Anders M ARVIK

torandma@stud.ntnu.no

Jonas K AASA
jonasre@stud.ntnu.no

Supervisors:
Sigurd S KOGESTAD
skoge@nt.ntnu.no

Maryam G HADRDAN
maryam.ghadrdan@chemeng.ntnu.no
MTKJ
April 23, 2012

Abstract
This report checks what ratio, between the real number of stages and the theoretical minimum number
of stages, that gives the lowest cost for a distillation column. Two different cost methods were used to
cost estimate the column, Sinnot and Turton. Sinnot gave a ratio of 1.9 for capital costs and 2.9 when
energy was included. For Turton the numbers were 2.3 and 3.1, respectively. The average ratios were
2.1 with no energy cost included and 3.0 with energy cost. Norwegian electricity was used to calculate
the energy cost. A more expensive energy source would give a higher optimum Nratio .
Data for the cost estimation were found by simulating separation of binary mixtures in a distillation
column in UniSim. 20 different binary mixtures were separated to generalize the prediction for the
optimum Nratio . In the simulation of the mixtures, feed fraction, distillate fraction, quality of the feed
b
and the numbers of stages were varied. An equation, VVmin = N/Nmin
1 + 1, was developed to curve fit the
results. It showed to be an inadequate fit.
Due to the sources of error in the simulations and the cost analyzes, the suggested Nratio is inaccurate,
but it indicates that the Nratio should be higher than the suggested value of two from literature.

Preface
This report was written as a part of the course TKP4170, "Process Design", during the spring semester
2012. The project is a study of the relation between the actual numbers of stages in a distillation column
and the minimum numbers of stages. The report requires basic knowledge in chemical engineering,
Matlab and UniSim.
The group would like to thank Ph.D candidate Maryam Ghadrdan for helpful guidance throughout the
process, and Professor Sigurd Skogestad for setting the project on the right track.

ii

Contents
Abstract

Preface

ii

Introduction

Theory
2.1 Outline of a continuous distillation column .
2.2 Simulating a continuous column . . . . . .
2.3 Shortcut metods in UniSim . . . . . . . . .
2.4 Gilliand . . . . . . . . . . . . . . . . . . .

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

Simulation
3.1 Components and property packages . . . . . . . .
3.2 Automation of the simulation . . . . . . . . . . . .
3.3 Extracted data . . . . . . . . . . . . . . . . . . . .
3.4 Curve fitting and estimating Nmin and Vmin . . . .
3.5 Analysis of the simulated data . . . . . . . . . . .
3.5.1 Trends between the different cases . . . . .
3.5.2 Average for all the cases . . . . . . . . . .
3.6 Gilliland analysis . . . . . . . . . . . . . . . . . .
3.6.1 Curve fitting to the Gilliland plot . . . . . .
3.6.2 Improving Nmin and Vmin by using Gilliland

.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.

2
2
2
3
3

.
.
.
.
.
.
.
.
.
.

4
5
6
7
8
10
11
11
13
13
14

Cost estimation
4.1 Size calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.1.1 Calculation of Column Diameter . . . . . . . . . . . . . . . .
4.1.2 Calculation of column size parameters . . . . . . . . . . . . .
4.1.3 Calculation of heat transfer area in the reboiler and condenser
4.1.4 Temperature of the streams in and out of the condenser . . . .
4.1.5 Temperature of the streams in and out of the reboiler . . . . .
4.2 Cost calculations assumptions . . . . . . . . . . . . . . . . . . . . .
4.3 Turton cost method . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.3.1 Purchased Equipment Cost . . . . . . . . . . . . . . . . . . .
4.3.2 Pressure factor . . . . . . . . . . . . . . . . . . . . . . . . .
4.3.3 Bare Module Cost . . . . . . . . . . . . . . . . . . . . . . .
4.3.4 Total Module and Grass Roots Costs . . . . . . . . . . . . .
4.4 Sinnott cost method . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.5 Accounting for Inflation . . . . . . . . . . . . . . . . . . . . . . . .
4.6 Effect of location factor . . . . . . . . . . . . . . . . . . . . . . . . .
4.7 Operating cost . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

15
15
15
15
16
17
17
17
18
18
19
19
20
20
21
22
22

Result from cost estimation

5.1 Distribution of cost . . . . . . . . . . . . . . . . . . . . . . . . .
5.2 Study parameters effect on the capital cost and the optimal Nratio .
5.2.1 The effect of changes in the parameters on the capital costs
5.2.2 Optimal N/Nmin for capital and overall costs . . . . . . .
5.3 N/Nmin when capital cost is considered . . . . . . . . . . . . . .

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

24
26
27
27
29
30

iii

.
.
.
.
.

.
.
.
.
.

5.4

N/Nmin when overall cost is considered . . . . . . . . . . . . . . . . . . . . . . . . . . 31

Discussion

33

Conclusion

34

List of symbols

35

References

37

Appendices

38

A Results

39

B Matlab scripts
B.1 main . . . . . . . . . . . . .
B.2 Stage change . . . . . . . .
B.3 curve fitting developed . . .
B.4 curve fitting Gilliland . . . .
B.4.1 Main gilliland script
B.4.2 Gilliland function . .
B.4.3 Gilliland plot . . . .
B.5 Analysis . . . . . . . . . . .
B.5.1 Analysis . . . . . .
B.5.2 Plot curve . . . . . .
B.5.3 Last square function

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

47
47
51
52
52
53
54
55
56
56
57
58

C Average N/Nmin when fixed capital cost is considered

60

D Average N/Nmin after cost calculation

65

E Calculation example
E.1 Calculation of column diameter . . . . . . . . . . .
E.2 Calculation of shell weight . . . . . . . . . . . . .
E.3 Calculation of area . . . . . . . . . . . . . . . . .
E.4 Cost calculation of capital cost when Turton is used
E.4.1 Calculation of purchased equipment cost .
E.5 Cost calculation when Sinnott is used . . . . . . .

iv

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

70
70
70
71
71
71
72

Introduction

Choosing the correct number of stages in an industrial distillation column is essential for a good design.
Several shortcut methods togive an estimate the numbers of stages needed in a distillation column have
been published. These methods calculate and approximate numbers of stages in the column as a function
of the reflux ratio and the minimum number of stages. This report will try to validate the correlation
found by Gilliland and find an improved correlation. This was done by running many different rigorous
simulations in UniSim and changing feed fraction, distillate fraction, feed quality, numbers of stages
and the components separated.
A cost estimation based in the simulated data is also carried out. Previously published reports has
suggested that two times the theoretical minimum number of stages is a good estimate for the number
of stages a column. In this report this rule of thumbs is discussed, and an improved value is proposed.
This is done by looking into both the construction cost and the operating cost. The energy consumption
increase when the numbers of stages in the column decreases. In this report the cost estimation methods
developed by Sinnott and Turton are used.

Theory

Distillation is a common separation technique in chemical industry. The method exploits that a boiling
mixture consisting of components with different boiling points, the composition in the liquid and the gas
will be different. The mixture can be vaporised and condensed several times to reach a desired purity.

2.1

Outline of a continuous distillation column

A continuous distillation column consist of a vertical shell house together with a condenser and a reboiler, see figure 2.1.

Figure 2.1: Standard setup of an industrial distillation column, together with reboiler and condenser [2]

The column is either filled with a packing material or consisting of stages, both to maximize the vaporliquid contact. Improving contact between the vapor and liquid phase makes each stage closer to the
equilibrium, and enhances the performance of the column. Thus reducing the numbers of stages needed
to reach the specified purity and reducing the energy consumption in the column.

2.2

Simulating a continuous column

Simulation of a column is dependent on how the phase equilibrium between vapor and liquid is calculated for each step. For ideal mixtures the vapor-liquid equilibrium can be derived from Raoults law and
mass balances . The equilibrium line for an ideal binary mixture is given in equation 2.1, [7].

yA =

xA
1 + ( 1)xA
2

(2.1)

 is the relative volatility, xA is fraction of component A in the liquid and yA is the fraction of component
A in the vapor. The simulation program UniSim uses property packages that include equations that describe vapor-liquid equilibrium for real systems. A property package is a set of equations that describes
the behavior of different chemicals. The property packages are designed to handle different types of
chemicals, Peng-Robinson is for an example used for non-polar hydrocarbons.

2.3

Shortcut metods in UniSim

The shortcut method in UniSim can be used to give an indication of the minimum number of steps
in a distillation column. The minimum numbers of trays occurs when the reflux ratio goes to infinity.
UniSim uses Fenskes formula to estimate the minimum number of stages. This method assumes that
the relative volatility is constant. Fenskes formula for binary mixtures is given in equation 2.2 [7].

Nmin =

xD 1xB
ln( 1x
)
D xB

(2.2)

Where xB and xD are the desired bottom and distillate liquid fractions respectively.

2.4

Gilliand

Gilliands correlation is a graphical method for estimating the real number of stages, N, required for a
separation at a given reflux ratio, R. In the correlation are the minimum numbers of stages, Nmin , and
the minimum reflux ratio, Rmin , needed. The graphical method plots
can also be written as

V V min
V +1 ,

NNmin
N+1

versus

RRmin
R+1 .

RRmin
R+1

where V is the vapor in the boil-up stream from the reboiler and Vmin

is the minimum size of the vapor stream, this can be done since

L
D

is equal to R and V is L + D. L is

the liquid stream that returns to the column after the condenser and D is the distillate stream from the
condenser [8].

Simulation

Simulation of a distillation column was performed in UniSim to find the vapor flow in the column for
20 binary mixtures at different feed fractions, z f , distillate fraction, zd, quality of feed, q, and number
of stages, N. The vapor flow is proportional to the energy consumption of the column. Quality of feed is
a measurement of the state of the stream. When q is equal to zero the feed is saturated vapor, and when
q is one the feed is saturated liquid. All combination of the cases presented in Table 3.1 were simulated.
A case is defined as unique set of variables from Table 3.1 and change of components in the column.
Table 3.1: The fixed variables for each simulation

z f [heavy, light]

zd[heavy, light]

zb[heavy, light]

[0.7, 0.3]
[0.5, 0.5]
[0.3, 0.7]

[0.01, 0.99]
[0.02, 0.98]
[0.04, 0.96]

0
0.5
1

[0.96, 0.04]
[0.96, 0.04]
[0.96, 0.04]

Figure 3.1 shows the process flow diagram for the simulation. The logical operation (VS box) applies
the same conditions to the feed to the rigorous simulation (F1) and the feed to the short cut column (F2).
The shortcut column gives an estimate for Nmin by Fenske equation 2.2.

Figure 3.1: Setup of simulation in UniSim. The rigorous columns is T-100 and the shortcut column T-101.

The simulation was done with a basis of 100 kmol/h feed and a column pressure at 1 bar. The temperature in the feed flow is a function of the feed quality and the composition. The reboiler pressure was set

to 105 kPa and the pressure drop in the condenser to 5 kPa. Distillate composition and bottom fraction
of the light component were used as the two degrees of freedom. It was assumed 100 % efficiency at
each stage. Modified Inside-out was used as solver for the column for increased robustness. Table 3.2
shows the specifications of the column.
Table 3.2: Simulation setup for rigorous column

Degree of freedom 1
Degree of freedom 2
Column pressure
Reboiler pressure
Pressure drop condenser
Solver

3.1

Fraction of heavy component in distillate

Fraction of light component in bottom
100 kPa
105 kPa
5 kPa
Modified Inside-Out

Components and property packages

Different types of components were separated to generalize the conclusion for the optimum ratio, Nratio ,
between the real number of stages and the theoretical number of stages. Alkanes, alkenes, alcohols,
phenols, polyols (glycerol), ether, chlorinated hydrocarbons and aromatic hydrocarbons were simulated.
These groups of components require different property packages for simulation. Peng-Robinson is the
most common used property package for hydrocarbons, but it does not work well with polar components.
For the alkanes, alkenes, ether and aromatic hydrocarbons Peng-Robinson was used as property package.
NRTL (Non-random two-liquid) was used for the simulations with alcohols, phenols and chlorinated
hydrocarbons, while for the polyols the glycerol package was used [5]. A complete overview of the
components used in the simulation, the component types and the property packages used can be found
in table 3.3.

Table 3.3: Setup simulation of rigorous column

3.2

Components

Type

Property package

1-butanol - 2-propanol

Alcohol

NRTL

Ethanol - methanol

Alcohol

NRTL

2,6 xylenol - phenol

Phenol

NRTL

2-chlorobutane - 1-chloropropane

Chlorinated hydrocarbon

NRTL

Ethylene dichloride - vinyl chloride

Chlorinated hydrocarbon

NRTL

Glycol - ethylene glycol

Polyol

Glycol package

1-butene - propene

Alkene

Peng-Robinson

m-cresol - phenol

Phenol

NRTL

Metyl-tert-butyl-ether - n-butane

Ether - alkane

Peng-Robinson

n-butane - propene

Alkane - alkene

Peng-Robinson

n-butane - n-propane

Alkane

Peng-Robinson

n-decane - n-nonane

Alkane

Peng-Robinson

n-nonane - n-octane

Alkane

Peng-Robinson

n-heptane - n-hexane

Alkane

Peng-Robinson

n-octane - n-heptane

Alkane

Peng-Robinson

n-pentane - n-butane

Alkane

Peng-Robinson

Propane - ethane

Alkane

Peng-Robinson

p-xylene - benzene

Aromatic hydrocarbon

Peng-Robinson

p-xylene - toluene

Aromatic hydrocarbon

Peng-Robinson

Toluene - benzene

Aromatic hydrocarbon

Peng-Robinson

Automation of the simulation

Changing z f , zd, q and N systematically and extracting V at the different conditions was done by connecting UniSim and MATLAB. MATLAB and UniSim were connected with object linking and embedding through ActiveX controls. The connection makes it possible to perform changes in and extracting
data from UniSim through Matlab. Appendix B shows how the connection was established and contains
the MATLAB scripts that performs the changes.
The sets of variables that were used in the simulation are presented in table 3.1. For each set of binary
mixture all the different combination of the variables in table 3.1 were simulated. This gives a total of
27 simulations with the same variable set. For each fixed set of variables the number of stages were
changed by multiplying Nratio with an estimate of Nmin , found by the shortcut column, and rounding
up the number. This was done for comparison of the different simulations, because all the simulations
would then have data for the same Nratio . The Nratio is defined by equation 3.1

Nratio =
6

N
Nmin

(3.1)

The Nratio values used for each simulation is presented in equation 3.2. As the critical area for cost
estimation is expected to have a Nratio around 2, the Nratio interval is short in this area.

Nratio = [ 1 1.1 1.2 1.3 1.4 1.5 1.6 1.8 1.9 2 2.1 . . .

(3.2)

2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 4 8 12 ]

Feed stage for each simulation was set by using equation 3.3. Fb is the molar flowrate of bottom and Fd
is the flowrate of the distillate in kilomoles per hour. Equation 3.3 does give an optimal feed location,
causing a higher vapor flow in the column than necessary [10].
Nrecti f ying
=
Nstripping

z flight
z fheavy



zbheavy
zdlight

2 

Fb
Fd

0.206
(3.3)

Nrecti f ying /Nstripping is the ratio between the number of steps in the rectifying and the stripping section.
This ratio was used to calculate the location of the feed stage as shown in equation 3.4. The equation
sets the feed location to the last stage in the stripping section. The plus one in the nominator corrects for
the stage added by the reboiler and the feed stage is rounded up to the closest whole number, as the feed
stage is an integer.

Feed tray = round

1+

N +1


Nrecti f ying
Nstripping

1

(3.4)

As previously mentioned, q, is a measurement of state of the feed flow, and it is calculated as shown in
equation 3.5. Hvap,sat is the enthalpy of the saturated vapor, H f is the enthalpy of the feed and Hliq,sat is
entalpy of the saturated liquid [9].

q=

3.3

Hvap,sat H f
Hvap,sat Hliq,sat

(3.5)

Extracted data

Parameters needed for later analysis were extracted from UniSim. Those parameters are found in Table
3.4. Notice that some of the variables set by the MATLAB script are again extracted. This is to assure
that the variables are inserted correctly and are not changed unintentionally during the simulation.

Table 3.4: List of variables extracted from the simulation

Symbol
q
Ff
xf
Fd
xd
V
N
Nmincurv
V mincurv
L
V
QR
QC
T1 R
T2 R
T1C
T2C

3.4

Unit
kmol/h
kmol/h

kmol/h

kg/h
kg/m3
kg/m3
kJ/h
kJ/h
oC
oC
oC
oC

Description
Quality of the feed
Feed mass stream
Feed composition in mol fraction
Distillate mass stream
Distillate composition in mol fraction
Feed tray
Vapor flow from the reboiler
Number of steps
Nmin calcualted by curve fitting and asymptote analysis
Vmin calcualted by curve fitting and asymptote analysis
Mass flow from the reboiler
Mass density of liquid in boil-up stream
Mass density of vapour in boil-up stream
Duty reboiler
Duty condenser
Temperature in to reboiler
Temperature out of reboiler
Temperature in to condenser
Temperature out of condenser

Curve fitting and estimating Nmin and Vmin

Once the feed fraction, z f , distillate fraction, zd and the quality of feed, q, are fixed, the column has a
Nmin and a Vmin independent of the column stages. Vmin is found when the numbers of trays go to infinity
and Nmin is found when V goes to infinity. To find Vmin and Nmin , V is plotted against N. An example of
this is shown in figure 3.2. The plot for the rest of the simulated cases is shown in appendix A.

(a) Normalized

(b) Simulation data

Figure 3.2: Simulated data of n-butane and propene. The left figure shows V /Vmin plotted against N/Nmin . The
right figure show N potted against V.

For each of the series in figure 3.2, Nmin and Vmin are the horizontal and vertical asymptotes respectively.
Approximate values for Nmin and Vmin were found by curve fitting equation 3.6 to the series. The curve
8

fitting was done by least square method. The Matlab script used for the curve fitting is in shown in
appendix B.

V=

b
+Vend
N Nmin

(3.6)

Vend is the vapor flow at the largest N in each simulation series. Vend was used since the number of stages
is large therefor close to the true Vmin . For the simulation of the alcohols the simulation of the largest
numbers of stages failed.
Equation 3.6 was the equation found that matched the path of the data points best. Three other equations
3.7-3.9 were also tested, but they did not give an adequate fit to the simulation data.

b1
+V min
a1 N Nmin
= a1 eb1 N

= a1 eb1 N + a2 eb2 N

(3.7)
(3.8)
(3.9)

3.5

Analysis of the simulated data

In order to validate the simulated data, the data were plotted varying only one of the three properties z f ,
zd and q. The two fixed variables were set to the middle of Table 3.1. Figure 3.3 shows number of stages
plotted against the vapor flow for the n-Butane and Propene.

Figure 3.3: Numbers of stages plotte agains the vapor flow, with the basis in the raw data. In the plot, two of the
3 degrees of freedom z f , zd and q is fixed to the middel simulated value.

When, q, is the only changed parameter, the plot indicates that Nmin is independent of q, since it converges to the same vertical asymptote. This corresponds to classical distillation theory. This plot also
shows that the vapor flow, which is proportional to the energy consumption, is higher when the feed
contains more liquid.
Once z f is the only adjusted variable the plot should converge to the same Nmin , according to distillation
theory. Unfortunately the simulation data those not give a smooth line and therefore do not give a clear
result. This is probably due to error in choosing the correct feed stage. When it comes to the vapor flow,
the plot shows that the vapor flow is lower once the feed contains more heavy component. This is due
10

to the specified composition in the bottom stream.

Once zd is the only adjusted variable it is hard to see if it converges to the same Vmin or not. Theoretically
the vertical and horizontal asymptotes for all the variation of zd should be different.

3.5.1

Trends between the different cases

A curve fit of the relation between N/Nmin and V /Vmin was made with the basis in all the collected
data from the simulation for each binary mixture. The trend for six of the different binary mixtures are
presented in figure 3.4. The equation used in the curve fitting is presented in equation 3.10, and is chosen
since it match the plot of V versus N well, see section 3.4.

V /Vmin =

b
N/Nmin 1

+1

(3.10)

Figure 3.4: Curve fit of the relation between N/Nmin and V /Vmin for the simulated data for six different binary
mixtures

3.5.2

Average for all the cases

An average value of V /V min as a function of N/Nmin for all cases was fund by taking the average of
the curve fit for the different binary mixtures. The function found is presented in equation 3.11.

11

V /Vmin =

0.22
+1
N/Nmin 1

(3.11)

This equation can be used as a shortcut method for finding the number of stages in the column for a
spesific V /Vmin , when Nmin is known. However it is not a good approximation as it is strongly dependent
on feed compositions, feed quality, type of components and distillate composition. This can be seen
from the large scattering of the data in figure 3.2 where one of the mixtures is plotted. The difference
between the different cases can be seen in figure 3.11.

12

3.6

Gilliland analysis

The data found by curve fitting was plotted as done by Gilliland [8], (N Nmin)/(N + 1) against (V
V min)/V . See also chapter 2.4. This plotting relieved a large scattering of the data in contrast to the
results found by Gilliland. Many of the series in the data set also had trajectories that seemed to not go
between the points [0,1] and [1,0] as seen in figure 3.5.
By definition of Nmin and Vmin, all the series should go between the points [0,1] and [1,0]. When V is
equal to Vmin N is infinite, and vice versa. This causes (N Nmin)/(N + 1) to approach 1 when V goes
to infinity (V V min)/V is then zero. The same logic can be provided to the other corner. An example
of the Gilliland plot with Nmin and Vmin found by curve fitting is shown in figure 3.5 to the right. The
black line is the Molokanov equation 3.12 which is a curve fit of the data found by Gilliland [6].


1 + 54.4x
y = 1 exp
11 + 117.2x

x1
p
(x)

!
(3.12)

Where y is (N Nmin)/(N + 1) and x is (V V min)/V

Figure 3.5: Simulated data of n-butanepropene plotted as done by Gilliland. The black line is the Molokanov
equation which is a curve fit of the results by Gilliland

3.6.1

Curve fitting to the Gilliland plot

It is assumed that the scattering in the right plot of figure 3.5 is due to errors in the curve fitting method
used in section 3.4. In addition notice that all the series go towards the point [0, 1]. This is a consequence
of the assumption that V is equal to Vmin on the largest simulated N in each serie. The true Vmin will
be smaller than this approximated Vmin.
In this particular simulated case most of the series lays below the Molokanov line. To make the series
match the Gilliland plot Nmin has to be larger if Vmin is fixed. It is also possible to only adjust Vmin
keeping Nmin constant, then Vmin has to become larger. But this can not be correct as the true Vmin is
below the already chosen Vmin.

13

3.6.2

Improving Nmin and Vmin by using Gilliland

As the cost analysis is dependent on a correct Nmin to be reliable, it is essential to get a good value for
this parameter. Gilliland was assumed to give a good description of the relationship between N, Nmin,
V and Vmin, and used to find an improved Nmin and Vmin. This was done by minimizing the square
error between the Molokanov equation and the simulated data. This operation is mathematically correct,
but it might not find the physically correct values, even if the Molokanov equation is correct. This is
caused by having two adjustable variables. If Nmin is fixed Vmin can be justed so it fits the Molokanov
line. If Vmin is fixed Nmin can be adjusted so it fits the Molokanov equation. The minimizing has still
been performed as it is assumed that it will mostly vary on Nmin.
The curve fitting was proved to be very sensitive to the guess of Nmin and Vmin. In some cases it was
not possible to find guesses that resulted in convergence for all series. The left figure in figure 3.5 shows
an example of the data with the improved Nmin and Vmin. It can be seen that one of the series has not
been converged.

14

Cost estimation

A cost estimation with the basis in the simulated data has been performed to find the Nratio which gives
the lowest total cost of the column. To do this it is essential to have a good method to determine the cost.
In this project there have been used two different cost methods to calculate the fixed capital cost, those
are the cost metods deveoped by Turton and Sinnott. Both methods are strongly dependent on a good
size estimation.

4.1

Size calculations

The size calculations are based on data found from the simulation.

4.1.1

Calculation of Column Diameter

To estimate the column diameter equations from Sinnot [12] were used. The maximum allowable vapor
velocity, uw , is used to find the diameter. Equation 4.1 and 4.2 shows how the diameter. This equation is
designed to optimize the column productivity, but still avoids flooding.
r
L V
uw = (0.171lt 2 + 0.27lt 0.047)
V

(4.1)

In the equation lt the plate spacing between the trays and is chosen to be 0.9 m for every case. This value
is chosen so it is large enough for easy maintenance of the column. L and V is the density of the liquid
and vapor in the boil-up stream respectively. The diameter is calculated using the equation 4.2 below.
In this equation the maximum allowable vapor velocity is used to calculate the diameter of the column,
and it also is assumed that the vapor volume is much greater then the liquid volume.
s
D=

4Vg
V uw

(4.2)

Vg kg/s is the vapor rate, and is found in the boil-up stream from the reboiler. The density that is used in
the equation is also found in the boilup stream and is given in kg/m3

4.1.2

Calculation of column size parameters

Then the height of the distillation column can be found be using equation (4.3)
L = N lt

15

(4.3)

Where N is the number of stage in the column. The volume in the column is found by multiply the
length of the column with the cross-sectional area in the column.
To estimate the column wall thickness, t equation 4.4 found in Sinnott [12] is used.
t=

Pi Di
(2SE 1.2Pi )

(4.4)

Pi is the maximum allowed pressure in the column, and is assumed to be 10 bar in this project. In this
equation, SE is the maximum allowable stress for type 304 stainless steel at 300 o F, which is assumed
to be 15 Ksi which is equal to 103.42 N/mm2 , found in Table 13.2 in Sinnott [12].
The shell mass, M, is found by using the following equation 4.5.

M = tw Dc Lc

(4.5)

Where tw is the thickness of the wall, Dc is column diameter and Lc is the column height.

4.1.3

Calculation of heat transfer area in the reboiler and condenser

In the project it is assumed that the reboiler and condenser are counterflow shell and tube heat exchangers. Steam is used to heat the cold stream in the reboiler and water, from a source near the plant, is used
to cool down the hot stream in the condenser. The area for the reboiler and condenser is found by using
equation 4.6 given in [12].

Q = UATlm

(4.6)

The duty, Q, for the condenser or the reboiler is extracted from UniSim. A is the area in m2 , Tlm is
the logarithmic mean temperature. U W /m2 C is the overall heat-transfer coefficient given in W /m2 C.
The value for the heat-transfer coefficient is given in Table 4.1 and is found in [3].
Table 4.1: Overall heat-transfer coefficient

U-tube

250 750

W /m2 C

Kettle reboiler

500 1000

W /m2 C

In this project the value 750 W /m2 C and 1000 W /m2 C is used for the U-tube and the kettle reboiler.
The logarithmic mean temperature difference, Tlm , can be found be using equation 4.7 found in [12]
Tlm =

T1 T2
1
ln( T
T2 )

16

(4.7)

T1 and T2 is the difference in the temperature between the two streams on both sides of the heat
exchanger.

4.1.4

Temperature of the streams in and out of the condenser

The temperature of the hot streams in and out of the condenser is found in UniSim. The hot inlet stream
in the condenser is found in the vapor stream in the top of the column. The cooled outlet stream in the
condenser is found in the distillate stream.
In this project it is assumed that cooling water in the process is pumped up from a water source near
the plant, and that the heated cooling water is returned to the water source. It is assumed that the
temperature for the cooling water in the condenser goes from Tin =10 C to Tout =15 C, due to the
minimum temperature difference given in Sinnott [12]. The difference in the temperature, Tmin , in the
to outlet streams for a condenser will usually be between 5 C and 7 C when cooling water is used as a
cooling agent. For every case in this project it is assumed that the minimum temperature difference is
5 C or more. If this lower limit is not satisfied, the temperature in the cooling stream has to be changed
to a lower value.

4.1.5

Temperature of the streams in and out of the reboiler

The temperatures of the in and out streams in the reboiler is also found in UniSim. The cold inlet stream
in the reboiler is found in the liquid stream in the bottom of the column, and the heated outlet stream in
the reboiler is found in the boil-up stream.
In the reboiler steam is used to heat the liquid stream. It is assumed that low presseure steam is used.
Low pressure steam has a pressure at 4 bar and the temperature is 160 C. According to this assumption
the inlet temperature for the steam is set to 160 C. The minimum temperature difference, Tmin , for a
reboiler has to be between 10 C and 30 C. In this project it is assumed that Tmin is 20 C. According
to this relationship the temperature for the steam out of the reboiler is set to be 20 C higher than the
heated outlet stream in the reboiler. If the outlet temperature for the steam is higher than the temperature
in low pressure steam, medium pressure steam or high pressure steam have to be used instead. Values
for medium (212 C and 20 bar) and high pressure (250 C and 40 bar) steam is found in [12].

4.2

Cost calculations assumptions

N/Nmin giving the lowest cost are found in two cases, with and without energy costs.
For both of these cases the cost evaluation is based on the following assumptions:
The distillation column is a vertical process vessel, and the maximum operating pressure is 10 bar
Sieve trays are used in the column

17

 The condenser is a U-tube heat exchanger and water is used as cooling agent
The reboiler is a kettle reboiler and steam is used as heating agent
Norwegian electricity is used to generate steam
Equipment for the support systems to the column is not included
The construction material is stainless steel
The pay back time is 5 years
The column is operating 8000 hours per year
Increase in the costs due to loans are not included
In one of the cost method equations in Turton [13] is used to calculate the cost. The factors and equations
that are used gives the cost in 2001 US gulf coast basis. In the other cost method equation in Sinnott [12]
is used, which give the cost in 2001 US gulf coast basis. In both the methods the inflation is taken to
considering to get the value to the present time by using equation 4.17 given in the chapter below.

4.3

Turton cost method

Equations given in Turton [13] is used to calculate the capital costs for the distillation column, including
condenser and reboiler.

4.3.1

Purchased Equipment Cost

The value of purchased cost, C0p , for the plant is calculated at ambient operating pressure and carbon
steel is used as construction material. This material cost is later converted to the cost for stainless steel
in chapter 4.3.3. Purchased cost can be calculated by using equation 4.8 given is [13]

log(C0p ) = K1 + K2 log(A) + K3 log(A)2

(4.8)

In this equation A is the size parameter. Kn is purchased cost factors for different equipment, and is
given in Table 4.2

18

Table 4.2: Equipment Cost Data, given in [13]

4.3.2

Equipment

K1

K2

K3

Size parameter

Process vessels

3.4974

0.4485

0.1074

volume [m3 ]

Sieve trays

2.9949

0.4465

0.3962

area [m2 ]

U-tube

4.1884

0.2503

0.1974

area [m2 ]

Kettle reboiler

4.4646

0.5277

0.3955

area [m2 ]

Pressure factor

The cost for the process vessel is depending on the operating pressure in the column. The pressure
factor, Fp is calculated by equation 4.9, which is a function of the column pressure and diameter.

Fp =

(p+1)D
2(849,61.2(p+1))

+ 0.00315

0.0063

(4.9)

where p is the operating pressure barg and D is the diameter m in the column. It has been assumed that
the operating pressure was 10 bar in the column. The cost for the other equipment is independent of the
pressure, when a low pressure is assumed.

4.3.3

Bare Module Cost

Bare module cost is the sum of direct and indirect cost for each equipment. The bare module cost for
the heat exchanger and the process vessel can be calculated by using the equation 4.10
CBM = C0p FBM

(4.10)

FBM = B1 + B2 FM Fp

(4.11)

where FBM is given in equation 4.11

Bn is the bare module factor and FM is the material factor. In this project there were assumed that all
the equipments material are stainless steel. Bn and FM for stainless steel are given in Table 4.3. Fp is
the pressure factor given above. The cost of the heat exchanger is independent of the pressure, when the
pressure is low, and Fp is then equal to one.

19

Table 4.3: Bare module factor and material factor for stainless steel and carbon steel

Equipment

B1

B2

FM Stainless steel

FM Carbon steel

Process vessels

2.25

1.85

3.33

Heat exchanger

1.63

1.66

2.7

1.33

The bare module cost for the sieve trays is calculated with equation 4.12.

CBM = C0p NFBM Fq

(4.12)

FBM is the bare module factor, which is dependent of the material for the equipment and are given in
stainless steel for the trays. For sieve trays FBM = 1.33. N is the number of actual trays in the column.
Fq is a quantity factor for trays, and can be found by using equation 4.13.
log(Fq ) = 0.4771 + 0.08516 log(N) 0.3473(log(N))2

(4.13)

for N < 20
Fq = 1 for N > 20

4.3.4

Total Module and Grass Roots Costs

Total module costs refers to the cost of making small-to-moderate expansion to an existing plant. Grass
roots cost, also known as the fixed capital cost, refers to the cost to a completely new facility.
Total module cost can be found by using equation 4.14.
n

CT M = CT M,i = 1.18 CBM,i

i=1

(4.14)

i=1

The grass root cost, CGR , can be estimated by using equation 4.15.
n
0
CGR = CT M + 0.50 CBM,i

(4.15)

i=1

0 in equation4.15, can be found by calculating the

In both equations n is the number of equipment. CBM

bare module cost for the equipment with ambient pressure and carbon steel as material.

4.4

Sinnott cost method

In the second cost method the cost equations presented in Sinnott [12] are used to calculate the cost for
the distillation column. The purchased cost, Ce , for the whole column, can be calculated by using eq.
20

(4.16)

Ce = a + bSn

(4.16)

Where S is a size parameter, a and b are cost constants and n is the exponent for the type of equipment; all
given in Table 4.4 . The calculation is done at ambient operating pressure and the construction material
is Stainless steel 304. The factors and equations that are used give the cost in 2006 and US gulf coast
basis.
Table 4.4: Equipment Cost Data

Equipment

Size parameter

Process vessels

15000

68

0.85

Shell mass [kg]

Sieve trays

110

380

1.8

Diameter [m]

U-Tube Condenser

24000

46

1.2

Area [m2 ]

Kettle reboiler

25000

340

0.9

Area [m2 ]

In the case of sieve trays, condenser and reboiler, the data of Table 4.4 is given for carbon steel and the
equation needs a factor of 1.3 to be applicable for stainless steel. The cost estimated by equation 4.16
needs to multiply by a factor of 6 in order to get the total fixed capital cost.

4.5

Accounting for Inflation

Since the fixed capital cost calculated in the two methods is not in the value for the present time. The
fixed cost have to be converted into value for the present time before the cost analyse can be done. This
can be done be using the equation below 4.17, given in [13].

C2 = C1

I2
I1

(4.17)

in this equation, C1 is the cost for the base time when the cost is known, C2 is the cost value for the
present time, I1 is the cost index for the base time and I2 is the cost index for the present time. The
CEPCI-index gives the values for I1 and I2
The value for I1 when the cost method in Turton is used is 397 [13]. The CEPCI-index when the cost
method in Sinnott is used is 509.7 [12]. For I2 the CEPCI-index for February 2012 is used. The value is
590.8 [4].

21

4.6

Effect of location factor

The equations and factors above give the cost in US Gulf Coast (USGC) basis. The total cost for a plant
is dependent on the location for the plant. By using equation 4.18 [12] the cost for the plant built in an
other country can be calculated [12].
Cost o f plant in location A = cost o f plant in USGC LFA

(4.18)

LFA is the location factor for location A relative to USGC basis.

In the case when the energy source in the reboiler is electricity, the cost for Norwegian electricity is used
in the cost evaluation. The column is therefore located in Norway. It has been assumed that the location
factor for Norway is 1.3 relative to USGC basis [11].

4.7

Operating cost

In this project, in the case of the operating costs, only the energy required in the reboiler that has been
considered. It is assumed that the cooling water that is used in the process, has been pumped up from
a water source near the plant. The cost of the electricity needed to pump up the water and the possible
cost for returning the heated cooling water to the water source is not included in the calculation of the
utility.
It has been assumed that steam is being used in the reboiler, and that electricity is used to heat water to
steam. The amount of electricity needed in this process is found by using the duty needed in the reboiler
and multiply it by 1.1 to account for a 10 percent heat loss. The cost for electrity was calculated by
approximating the average value for the last four years in figure 4.1 for the blue line.

Figure 4.1: Price for the electric power in Norway. The green line is households, the orange line is service industry
and the blue line is industry without high power industry and pulp and paper industry

22

The blue line in the figure show the price for the electric power in the Norwegian industry. In this
project it has been assumed that the price for the electricity is 0.35 NOK/kW h, which is equal to 0.061
USD/kW h when the currency is converted to USD [1].

23

Result from cost estimation

The cost estimation has been done for all the simulated mixtures in this project, except for the mixture
with a boiling point below 0 oC for atmospheric pressure, since that gives negative temperature values
for the vapor entering the condensor. Another cooling agent than water has to be used in these situation.
The cases filling these circumstances were discarded to from the cost estimation. That concerns the
cases: n-Butane and Propane, Ethylene dicloride and Vinyl chloride, i-Butene and Propylene, Methyl
tert-butyl ether and Butane,n-Butane and propylene, n-Pentane and Butane, Propane and Ethane.
Figure 5.1-5.3 shown below are the results from the simulation of the separation of Toluene and Benzene.
The usage of colors and symbols are the same in all these three figures. The blue, red and green lines
present the mole fraction for the binary mixture in the distillation stream in the top of the column. The
red lines represent when zd is 0.01, the green lines when zd is 0.02 and the blue lines when zd is 0.04.
The different symbols represent the changes in z f . The filled circle is when z f is 0.7, the unfilled circle
when z f is 0.5 and the cross when z f is 0.3.
Figure 5.1 shows the N/Nmin at the x-axis and the total capital cost at the y-axis. The optimal N/Nmin is
when the capital cost has a minimum.The minimum in fixed capital cost for toluene and benzene with
different properties can be observed in the figure 5.1 shown below.

(a) Cost method from Turton

(b) Cost method from Sinnott

Figure 5.1: Fixed capital cost plotted against N/Nmin

In figure 5.1 can it be seen that for the Turton cost method is the optimal N/Nmin slighty higher than two,
and for the Sinnott method is it slightly lower than two.
Figure 5.2 shows N/Nmin plotted against the price for the energy needed in the reboiler in one year.

24

Figure 5.2: Energy cost plotted against N/Nmin

From the theory it is know that the energy in a distillation column decrease when the height of the
column increases. This correlation can be observed in Figure 5.2.
Figure 5.3 shows the N/Nmin on the x-axis and the sum of the annualized capital cost and the energy
cost on the y-axis.

(a) Cost method from Turton

(b) Cost method from Sinnott

Figure 5.3: Capital cost and energy cost plotted against N/Nmin

In figure 5.3 it can be observed that the minimum for the overall cost has a higher value for the Nratio
compared to the case when only the capital cost was plotted. against Nratio . The optimal value for Nratio
is now closer to three than to two. This shows that increasing the energy costs increases the optimal
Nratio for both cost methods.

25

5.1

Distribution of cost

Figure 5.4 and 5.5 shows the cost distribution for column when the minimum total cost occurs for
separation of Toluene and Benzene. This occurs when the mole fraction for the heavy component in
the feed stream is 0.7, the feed stream is vapor and the mole fraction for the heavy component in the
distillate is 0.04. In figure 5.4 the distribution for the capital cost is considered.

(a) Cost method from Turton

(b) Cost method from Sinnott

Figure 5.4: Distribution of capital cost

In the pie diagrams it can be observed that the cost for the column is lower when the Turton method
is used compared to the cost method in Sinnott. The reason for this can be that when the Turton cost
equations are used for the column, the volume of the column is the cost parameter, while the Sinnott
method it is the weight of the column shell that is the cost parameter. It can also be observed in the
diagrams that the cost for the reboiler is higher when the cost method from Turton is used compare to
the cost method in Sinnott. In the both the cost equations for the reboiler, it is the area of the reboiler
that is the cost parameter. These difference in distribution shows how the cost parameters are differently
weighted in the two methods.
In figure 5.4 the cost for the energy is also included in the cost distribution.

26

(a) Cost method from Turton

(b) Cost method from Sinnott

Figure 5.5: Distribution of the overall cost

The pie diagram shows that the cost for energy is the largest expense for the overall costs, accounting for
over 50 % in both the methods. This is even though a cheap source of energy were used for generating
steam to the reboiler.

5.2

Study parameters effect on the capital cost and the optimal Nratio

Studies were done to check the effect of z f , zd and q on the capital costs and the optimal Nratio for the
capital costs and the overall costs. This was done by holding two of the parameters constant and varying
the third.

5.2.1

The effect of changes in the parameters on the capital costs

Figure 5.6 is the result when z f is equal to 0.5, q is equal to 0.5 and zd is varied. The serie with circles
is when zd is 0.01,the squares when zd is 0.02 and the triangle when zd is 0.04.

(a) Cost method from Turton

(b) Cost method from Turton

Figure 5.6: Fixed z f and q while zd varies. The sum of capital plotted against N/Nmin

27

The result shows that the capital cost for the column, in both cost methods, increases when the product in
the distillation stream becomes purer. This result is reasonable since a purer product needs a distillation
column with a higher number of trays.
Figure 5.7 is the result when z f is equal to 0.5 and zd is 0.02. The serie with circles is when the feed
stream is saturated vapor, the serie with squares is when the amount of vapor and liquid in the feed
stream is equal, and the serie with triangles is when the feed stream is saturated liquid.

(a) Cost method from Turton

(b) Cost method from Sinnott

Figure 5.7: Fixed z f and zd while q varies. The sum of capital plotted against N/Nmin

The results shows that the capital cost for the column increase when the feed quality goes from saturated
vapor to saturated liquid. The trend can be seen for both cost methods. This result is reasonable since
the energy requirement in the reboiler is higher when the inlet stream is liquid, and then the reboiler area
then has to be larger, causing an increase in the cost for the reboiler.
In figure 5.8 is zd and q kept constant at 0.02 and 0.5, respectively. The circles represent when z f is 0.7,
the squares when z f is 0.5 and the triangles when z f is 0.3.

(a) Cost method from Turton

(b) Cost method from Turton

Figure 5.8: Fixed zd and q with varying z f . The sum of capital cost is plotted against N/Nmin

In figure 5.8 it can be observed that the cost increase when the amount of heavy component in decreases.
28

This is due to the setup of the simulation in UniSim. The size of the bottom stream and the purity is fixed,
so in order to achieve the given specifications is it necessary to increase the fluid volume in the reboiler.
Then more heat is required in the reboiler to vaporize the liquid and area of the reboiler increases.

5.2.2

Optimal N/Nmin for capital and overall costs

Figure 5.9 shows the optimal Nratio found from capital cost estimation for the 27 different simulation
series in the separation of Toluene and Benzene. The cost estimation method on the left in figure 5.9 is
for Turton and the right is for Sinnott. For the first three circles are the fraction of heavy in the distillate
increased from 0.01 to 0.04, while q and z f are kept constant. For the next three circles are z f kept
constant, q is increased to 0.5, while the cycle for zd is repeated.In the next three circles are z f still kept
constant, q is increased to 1 and the cycle for zd is repeated. Next is z f decreased from 0.7 to 0.5 and the
loops for q and zd are repeated. The arrows below the graphs illustrates how the parameters are changed.
In the left graph it can be seen that a reduction in the purity of the product, the heavy fraction in the
feed and the vapor fraction gives a higher Nratio . Reducing the vapor fraction increases the area of the
reboiler, due to an increased heat usage, giving higher capital costs for a low vapor fraction. Reducing
the purity decreases Nmin , according to distillation theory. The reduced Nmin gives a higher optimal
Nratio . As previously mentioned does a decrease of the heavy fraction in the feed increase the capital
cost, contrary to the distillation theory. This is due the specifications for the column in the simulation.
The cost method for Sinnott does not show any clear trends, showing how the optimal Nratio , in addition
to be dependent upon the parameters, are dependent of the cost estimation method. Similiar graphs for
all the cost estimated cases can be found in appendix B.5.3.
Figure 5.9: N/Nmin vs. various cases of different compositions for the minimum amount of capital cost

(a) Cost method from Turton

(b) Cost method from Sinnott

The cost estimation series were not curve fitted, so the minimum costs were found for the Nmin provided
from the simulation data.

29

Figure 5.10 below shows the overall costs for the column, and it is built up in the same way as figure
5.9. Now is it the Sinnott cost estimation that shows the clear trends. The graph for the Sinnott method
follows the same pattern as the Turton method described for figure 5.9. In appendix C can plots for the
other cases be found.

(c) Cost method from Turton

(d) Cost method from Sinnott

Figure 5.10: N/Nmin vs. case, when capital cost and energy cost is considered

5.3 N/Nmin when capital cost is considered

The data plotted in 5.2.2 were used to calculate an average optimal N/Nmin for the capital cost for each
of the cost estimated cases. The average for each of the cases are shown in Table 5.1.

30

Table 5.1: Average optimal N/Nmin when different components are separated

N/N min
Components

Cost method 1.

Cost method 2

Toluene and Benzene

2.3413

1.8451

1-Butanol and 2-propanol

2.5188

1.9805

26-xylenol and phenol

2.2085

1.7353

C4-Chloride and C3-Chloride

2.3170

1.8289

Ethanol and Methanol

2.2463

1.7979

Glycerol and Ethylene Glycol

2.3919

2.3515

m-cresol and phenol

2.1982

1.7727

n-Decane and n-Nonane

2.1150

1.7856

n-Heptane and Hexane

2.2554

1.8236

n-Nonane and n-Octane

2.0809

1.7728

n-octane and Heptane

2.2772

1.7875

p-Xylene and Benzene

2.5430

2.0297

p-Xylene and Toluene

2.3705

1.8587

Average N/Nmin

2.3027

1.8746

The two different cost methods gives to different results for the average optimal N/Nmin . The reason for
this is that the cost for the column and reboiler are differently weighted in the two methods. In Turton the
column have a lower price while the reboiler is more expensive, compared to Sinnott. This indicate that
good cost estimation is necessary to find the number of stage in the column that gives the total minimum
cost. The total average optimal N/Nmin for the capital costs are 2.1.

5.4 N/Nmin when overall cost is considered

The average optimal N/Nmin corresponding the minimum overall cost is found for the cost estimation of
the different binary mixtures. Table 5.2 shows the optimal N/Nmin for each mixture, estimated by both
the costs methods.

31

Table 5.2: Optimal N/Nmin for the overall costs

N/Nmin
Components

Cost method from Turton

Cost method from Sinnott

Toluene and Benzene

3.1423

2.8577

1-Butanol and 2-propanol

3.7093

3.3567

26-xylenol and phenol

2.8562

2.7089

C4-Chloride and C3-Chloride

3.0063

2.7839

Ethanol and Methanol

3.0515

2.9505

Glycerol and Ethylene Glycol

4.0877

4.1773

m-cresol and phenol

2.9820

2.8413

m-Decane and n-Nonane

2.8175

2.6893

n-Heptane and Hexane

3.0329

2.8227

n-Nonane and n-Octane

2.8612

2.7301

n-octane and Heptane

2.9266

2.8073

p-Xylene and Benzene

3.6442

3.2687

p-Xylene and Toluene

3.0691

2.8065

Average N/Nmin

3.0916

2.8853

When the operating cost is included in the total cost, the average N/Nmin for the two cost methods are
more similar. The optimal N/Nmin increases compared to the N/Nmin for the capital costs. The average
N/Nmin for the two methods are 3. When capital costs are estimated, can the energy used be considered
as free. The increase in the optimal N/Nmin for the overall costs indicates that N/Nmin should be higher
than 2.

32

Discussion

All the calculations in this project is based on the simulation data from UniSim. Possible sources of
errors in the simulation will therefore be important for the result. The property package used for each
simulation, feed stage location and the tolerance in the equilibrium calculations in the column. Reducing
these sources of error would decrease the uncertainties for the results and perhaps decrease the fluctations. Optimizing the feed stage location would for an example decrease the energy usage in the column,
and therefor affect the optimal Nmin for the overall costs. z f , zd, q and N were varied to generalize the
results from the simulations.
1 bar was used as feed pressure for the column, causing many of the components to have a bubble point
below 0oC, giving negative temperatures in the condenser. Water was used as cooling agent in this
project, and because that it was not possible to cool down the vapor in the condensor. Seven simulations
were discarded from the cost analysis due to these circumstances.
The curve suggested by the group for curve fitting showed a decent fit to the simulation series. It did
however estimate a too low Nmin as well as a too high Vmin . This was seen when the results were plotted
with the same axes as Gilliland used. Using curve fitting to find Nmin and Vmin is a source of error for the
recommended Nratio , since the ratio is dependent upon Nmin . The Nmin found in this report is higher than
the true Nmin , making the recommended Nratio lower than it should be.
The results found in the report is dependent on the cost estimations method by Turton and Sinnott. The
energy source used to create steam for the reboiler is Norwegian electricity, which is a relative cheap
energy source. Using a more expensive source of energy, would increase the recommended Nratio . An
increase in the capital costs would give a lower optimal Nratio , since the capitol cost would be a larger
portion of the total costs. The number of years for the pay back time of the column is also important. A
longer pay back time would increase the recommended Nratio .
Due to the sources of error is there uncertainties concerning in the optimal Nratio that was found in the
report.
For further analysis it is suggested that more accurate values are found for the overall heat-transfer
coefficient, used in the cost estimation and for Nmin . An improved value for Nmin can be found by using
McCabe-Thieles method. Performing a similiar analysis for a multicomponent mixture would increase
the generalization of the result.

33

Conclusion

This report have checked if the ratio between the real number of stages and the theoretical number of
stages should be two, as mentioned in literature. A suggestion for a new ratio was found by simulating
different binary mixtures in UniSim. Two different types of cost estimation methods, Sinnot and Turton,
were used to check the ratio. With the Sinnot method it was found that the ratio giving the lowest
capital cost is 1.9, and the ratio giving the lowest overall cost is 2.9. For the Turton method are the
ratio for lowest capital cost and the lowest overall costs 2.3 and 3.1, respectively. The average ratio for
capital cost is 2.1 and the average for the overall costs is 3.0 for both methods. Comparison of the two
optimal ratios for capital costs and overall costs shows that the optimal Nratio increases as the energy
cost increases.
It was proposed an equation, V /Vmin = b/(N Nmin ), for curve fitting the simulation series and to be
used as a shortcut method for finding Nmin and Vmin . The equation was found to not give a good fit, the
Nmin found is too low and the Vmin is too high. To find a better estimate for Nmin and Vmin , the simulations
series were curve fitted to the Molokonov equation in a Gilliland plot.
There are some sources of error concerning the Nratio found in this rapport. The Nratio found depend
upon the results from the simulations, the sizing of the column and the thorougness of the cost analysis.
In the cost analyzes were many assumptions made for the column and the energy costs. Due to these
sources of error it is not certain that the optimum Nratio should be 3.0 for the overall costs, but the results
shows that the Nratio should be higher than two when the energy costs is included.

34

List of symbols
Table 8.1: List of symbols

Symbol

Unit

Description
Relative volatility

kg/m3

mass density of liquid in boil-up stream

kg/m3

mass density of vapor in boil-up stream

Tmin

oC

Minimum temperature difference

constant curve fitting

constant Sinnott cost estimation

a1

constant curve fitting

a2

constant curve fitting

Size parameter for cost estimation

constant curve fitting

constant Sinnott cost estimation

b1

constant curve fitting

b2

constant curve fitting

Cp

US$

Perched equipment cost

CBM

US$

Bare module cost

Ce

US$

Purchased cost, Sinnott cost estimation

CGM

US$

Fixed capital cost

CT M

US$

Total module cost

Dc

Diameter column

FBM

Bare module factor

Fp

Pressure factor used in cost estimation

Fq

Quantity factor for trays

Hf

kJ

Enthalpy feed

Hliq,sat

kJ

Enthalpy feed as saturated liquid

Hvap,sat

kJ

Enthalpy feed as saturated vapor

Hf

kJ

Enthalpy feed

K1

Constant for cost estimation

K2

Constant for cost estimation

K3

Constant for cost estimation

lt

Plate spacing in spacing

kmol/h

Liquid flow

LC

Hight of distillation column

35

Symbol

Unit

Description

constant Sinnott cost estimation

Number of trays

Nmin

Minimum number of steps

Nmincurv

Nmin calculated by curve fitting and asymptote analysis

V mincurv

Vmin calculated by curve fitting and asymptote analysis

kg

Mass of column

Nrecti f ying

Number of stages in the rectifying section

Nratio

Ratio between Number of steps and minimum number of steps

Nstrippinng

Number of stages in the stripping section

Vapor mol fraction of component A

yA
Pi

N/mm2

Maximum allowable column pressure

Duty of re-boiler

quality of the feed

Reflux ratio

Minimumn reflux ratio

Size parameter in Sinnott cost estimation

SE

Ksi

Maximum allowable stress

Wall thickness

T1 R

Temperature reboiler in

T2 R

Temperature reboiler out

T1C

Temperature condenser in

T2C

Temperature condenser out

Logarithmic middle temperature

Tlm
uW

m/s

Maximum allowable vapor velocity

kmol/h

Vapor flow

Vg

kg/s

Gas flow in boil-up for size calculation

Vmin

kmol/h

Minimum vapor flow

xA

Liquid mol fraction of component A

xD

Liquid mol fraction in distillate

xD

Liquid mol fraction in bottom

zf

Feed composition in mol fraction

z fheavy

Feed heavy component in mol fraction

z flight

Feed light component in mol fraction

zd

Distillate composition in mol fraction

zb

Bottoms composition in mol fraction

36

References
[1] http://www.dn.no/finans/valuta/, Sited: 31. march 2001.
[2] http://lorien.ncl.ac.uk/ming/distil/distileqp.html, Sited: 22. April 2012.
[3] http://www.engineeringpage.com/technology/thermal/transfer.html, Sited:

31.

march 2012.
[4] Cepci. Chemical Engineering, 119(2):5152, 2012. Copyright - Copyright Access Intelligence
LLC Feb 2012; Language of summary - English; Pages - 51-52; ProQuest ID - 921996196; Place of
publication - New York; Corporate institution author - Anonymous; DOI - 2588841711; 67448622;
29283; CEG; INODCEG0007694400.
[5] Dont gamble with physical properties for simulations. http://www.eq.uc.pt/~nuno/eps/
old/2010-11/aula09/Carlson%20(1996).pdf, Sited: 21. April 2012.
[6] R. S. Al-Ameeri and A. S. Said. Simple formula for the gilliland correlation in mulitcomponent
distillation. Separation Science and Technology, 20(7-8):565575, 1985.
[7] C.J. Geankoplis. The Transport processes and separation process principles. Pearson Education,
4th edition edition, 2003.
[8] E. R. Gilliland. Multicomponent rectification estimation of the number of theoretical plates as a
function of the reflux ratio. Industrial and Engineering Chemistry, 32(9):12201223, 1940.
[9] Ivar J. Halvorsen and Sigurd Skogestad. Distillation theory. http://www.nt.ntnu.no/users/
skoge/studenter_backup/halvorsen/public_html/publications/thesis/Chapter_
2.pdf, Sited: 31. march 2012.
[10] Ernst J. Henley and title= J. D. Seader.
[11] Mangne Hillestad. personal communication, 2012.
[12] R. Sinnot and G Towler. The Chemical Engineering Design. Elsevier, 5th edition edition, 2009.
[13] Bailie

R.

C.

Whiting

W.

B.

"Turton,

R.

and

J.

Analysis, Synthesis and Design of Chemical Processes. 2nd edition edition.

37

A"

Shaeiwitz.

Appendices

38

Results

The numbers of steps in the simulated column was potted towards the vapor flow. The graphs for the
simulated cases is found in the figures right figure below. Nmin and Vmin was found by curve fitting, the
procedure for this calculation is found in chapter 3. The normalized graphs with the Nmin and Vmin
found by curve fitting is presented to the left in the following figures.

Figure A.1: Simulated data of 1-butanol2-propanol

Figure A.2: Simulated data of 26-xylenolphenolfitted

Figure A.3: Simulated data of c4chloridefitted

39

Figure A.4: Simulated data of edcvcmfitted

Figure A.5: Simulated data of ethanolmethanol

Figure A.6: Simulated data of glycoleglycol

40

Figure A.7: Simulated data of i-butenepropene

Figure A.8: Simulated data of m-cresolphenol

Figure A.9: Simulated data of MTBEn-butane

41

Figure A.10: Simulated data of n-butanepropane

Figure A.11: Simulated data of n-butanepropene

Figure A.12: Simulated data of n-decanen-nonanefitted

42

Figure A.13: Simulated data of n-heptanen-hexane

43

Figure A.14: Simulated data of n-nonanen-octane

Figure A.15: Simulated data of n-octanen-heptane

Figure A.16: Simulated data of pentanebutane

44

Figure A.17: Simulated data of pentanebutane

Figure A.18: Simulated data of p-xylenebenzene

Figure A.19: Simulated data of p-xylenetoluenefitted

45

Figure A.20: Simulated data of toluenebenzene

Figure A.21: Simulated data of xylenolphenolfitted

46

Matlab scripts

Some of the matlab script that is used in the simulation is found in the following subchapters. The rest
of the matlab script that is used will be given on request.

B.1

main

Main.m changes zf, zd, q and N, either in the document or by using other subscripts. All the changes
and the corresponding V values are stored in the data matrix. After all the simulations have been run,
is the data matrix stored as a matlab datafile. Main.m also performs the curve fitting of the results to
equation ??.

%% Main.m
Summary:
UniSim is used to simulate a destillation column. Number of
steps, minimum number of steps, and V is calulated by UniSim. This script
systematically changes the following parameters in UniSim
q
quality of feed
zf
composition of feed
zd_heavy Heavy fraction of destillate
N
Number of trays
author:
Tor Anders Marvik and Jonas Kaasa
organization: Department of Chemical Engineering, NTNU, Norway
contact:
torandma@stud.ntnu.no
license:
requires:
matlab, UniSim and SCP scrips
since:
2012.02.24
version:
1.0

%
%
%
%
%
%
%
%
%
%
%
%
%
%
%
%
%
% started
% changed:
% note:
%
%Input:
%Output:
%
%
%
%
%
%
%
%
%
%

February 2012
Blabla.
Data from UniSim and scripts that conects with UniSim
alfa
Relative voletility
q
quality of feed
Ff
Molar flow feed [kmol/h]
zf
Molar fraction feed [Heavy Light]
Fd
Molar flow destillate [kmol/h]
zd
Molar fraction destillate [Heavy Light]
V
Vapor follow from reboiler [kmol/h]
Vmin
Minimum wapour flow [kmol/h]
N
Number of stages in rigorus column model
Nmin
Nmin from shortcut column (Fenske)

47

%
% Guide to use the script
% 1. Open the UniSim simulation
% 2. Copy the UPC scripts on to the script folder in UniSim
% 3. Specify the path to the script folder in UniSim, line 57 in main.m
% 4. Manually change the components and the fluid package in UniSim. The
%
choice of fluid package have to be discussed for each case.
% 5. Run the matlab script
% 6. UniSim stops if the parameters specified is infeasible. Do not
%
fulfill the mass balance
%%
clc
clear all
close all
global script_folder
%makes script_folder global
script_folder='D:\Skole\4aar\New folder\UPC scripts';
%the directory
%containing the UPC scripts
addpath([pwd,'\q_feed'])
%add the folder q_feed to search path
serv = actxserver('UniSimDesign.Application'); %connects to UniSim ActiveX
%server
hyCase = serv.ActiveDocument;
%UniSim model
col=hyCase.Flowsheet.Operations.Item('T100');
%rigorous column in UniSim
specs=col.ColumnFlowsheet.Specifications;
%specs in column
stream=hyCase.Flowsheet.MaterialStreams;
%stream path
shortcut=hyCase.Flowsheet.Operations.Item('T101');
%shortcut column path
EnergyStreams=hyCase.Flowsheet.EnergyStreams;
%energy stream path
data=[];
%Matrix to save created data
%data=[q,Ff,zf,Fd,zd,tray,V(i),N(i),1,1,massflow,rhoL,rhoV,Duty_reboiler,
%
Duty_condenser,T1R,T2R,T1C,T2C];
color=['y';'m';'c';'r';'g';'b';'w';'k'];
symbol=['.','o','x','+','*'];
count_symbol=0;
count=0;

%for color in plot

%for symbol in plot

%keeps track of the series in the data file

%Loop that changes zf

zf_matrix=[0.7:0.2:0.3;0.3:0.2:0.7]';
%[zf_heavy, zf_light]
for j=1:length(zf_matrix)
%systematic change of feed composition
zf_change(zf_matrix(j,:),script_folder,hyCase);
%change feed
%composition in UniSim [heavy light]
count_symbol=count_symbol+1;
%Loop that changes q
count_color=0;
for q=0:0.5:1

48

q_feed(q,100,hyCase);
count_color=count_color+1;

%iterate on temperatur to reach q, 100C=To

%Loop that changes zd

zd_heavy=[0.01,0.02,0.04];
%fraction of heavy component in distillate
for zd_heavy_teller=1:length(zd_heavy)
specs.Item('Comp Fraction').goalValue=zd_heavy(zd_heavy_teller); %set heavy
%fraction distillate
%Retrive molar flow and compositions from column streams
zf= stream.Item('F1').componentMolarFractionValue;
%mole fraction feed
%stream [heavy light]
Ff=stream.Item('F1').molarFlow.value*3600; %molar flow feed stream [kmol/h]
zd= stream.Item('D1').componentMolarFractionValue;
%mole fraction
%distillate[heavy light]
Fd=stream.Item('D1').molarFlow.value*3600;
%molar flow distillate[kmol/h]
zb= stream.Item('B1').componentMolarFractionValue;
%mole fraction bottom
%[heavy light]
Fb=stream.Item('B1').molarFlow.value*3600;
%molar flow bottom [kmol/h]
EnergyReboiler=EnergyStreams.Item('E_btm1').heatFlowValue*3600;
%energy
%reboiler [kJ/h]
%Data from rigourous UniSim simulation, not changing with stage
hyCase.Flowsheet.operations.Item('SPRDSHT1').Cell('B21').CellValue=zd(1);
%fraction of heavy compononent in distillate in shortcut column
hyCase.Flowsheet.operations.Item('SPRDSHT1').Cell('B20').CellValue=zb(2);
%fraction of light compononent in bottom in shortcut column
Nmin=hyCase.Flowsheet.operations.Item('SPRDSHT1').Cell('B6').CellValue;
%Nmin from shortcut model, via spreadsheet
%
Nratio=[1 1.1 1.2 1.3 1.4 1.5 1.6 1.8 1.9 2 2.1 2.2 2.4 2.6 2.8 3 3.2...
3.4 3.6 3.8 4 8 12];
%ratio between N and Nmin, 11 elements long
count=count +1;
for i=1:length(Nratio);
N(i)=ceil(Nratio(i)*Nmin);

%set number of stages. Does not include

%reboiler
Nr_Ns=((zf(1)/zf(2))*(zb(2)/zd(1))^2*(Fb/Fd))^(0.206);
%Ratio between
%rectifying and stripping stages in column
tray=round((N(i)+1)/(1+1/Nr_Ns));
%set the feed tray
stage_change(N(i),script_folder,hyCase);
%change N in rigorous model
feedstagechange(tray,script_folder,hyCase);
%change feed tray
V(i)=...
col.ColumnFlowsheet.MaterialStreams.Item.('boilup').molarFlowValue...
%from sub flowsheet, vapor to reboiler [kmol/h]
*3600;
%data importing for the cost estimation
massflow = ...

49

col.ColumnFlowsheet.MaterialStreams.Item('boilup').massFlowValue;
%massflow boilup[kg/s]
rhoL=...
hyCase.Flowsheet.operations.Item('SPRDSHT1').Cell('B33').CellValue;
%density liquid boilup stream [kg/m3]
rhoV=...
hyCase.Flowsheet.operations.Item('SPRDSHT1').Cell('B34').CellValue;
%density vapor boilup stream [kg/m3]
Duty_reboiler=...
hyCase.Flowsheet.operations.Item('SPRDSHT1').Cell('B36').CellValue;
%Energy reboiler [kJ/h]
Duty_condenser=...
hyCase.Flowsheet.operations.Item('SPRDSHT1').Cell('B37').CellValue;
%Energy condenser [kJ/h]
T1R=hyCase.Flowsheet.operations.Item('SPRDSHT1').Cell('B40').CellValue;
%temperature in to reboiler [C]
T2R=hyCase.Flowsheet.operations.Item('SPRDSHT1').Cell('B41').CellValue;
%temperature out of reboiler [C]
T1C=hyCase.Flowsheet.operations.Item('SPRDSHT1').Cell('B38').CellValue;
%temperature in to condenser [C]
T2C=hyCase.Flowsheet.operations.Item('SPRDSHT1').Cell('B39').CellValue;
%temperature out of condenser [C]
data=[data;q,Ff,zf,Fd,zd,tray,V(i),N(i),1,1,massflow,rhoL,rhoV,...
Duty_reboiler,Duty_condenser,T1R,T2R,T1C,T2C,count]; %values stored
%in data file
%Plots
figure(2)
plot(N(i),V(i),[color(count_color) symbol(count_symbol)])
%plot N vs V
xlabel('N')
ylabel('V')
title('Number of steps in a destillation column against the vapor stream')
hold on
legend(['q= ' num2str(q)],...
['zf= ', num2str(zf)],['zd= ', num2str(zd)],'Nratio')
end
%Calculation of Nmin and V min using curv fittion by least square
V_calc=data((endlength(N)+1):end,9);
N_calc=data((endlength(N)+1):end,10);

%store column 9 in data as V_calc

%store column 10 in data as N_calc

Nmin=0;
%Guess for Nmin, have to below the true Nmin
b=0.1;
%Guess for b
Vmin=V_calc(end);
%sets Vmin equal to last V_calc
[x,error,flag] = fminsearch('least_square',[Nmin,b],[],N_calc,V_calc,Vmin);
%minimizes the function least_square. Nmin and b are output. N_calc, V_calc
%and Vmin are input

50

Nmin=x(1);
b=x(2);
data(endlength(N)+1:end,11)=data(endlength(N)+1:end,9)./Vmin; %calculate
%ratio between V/Vmin
data(endlength(N)+1:end,12)=data(endlength(N)+1:end,10)./Nmin; %calculate
%ratio between N/Nmin
figure(3)
%plot the ratios N/Nmin vs V/Vmin
V_Vmin=data(endlength(N)+1:end,11);
N_Nmin=data(endlength(N)+1:end,12);
xlabel('N/Nmin')
ylabel('V/Vmin ')
title...
('Number of steps in a destillation column vs the vapor stream normalized')
plot(N_Nmin,V_Vmin,[color(count_color) symbol(count_symbol)])
hold on
figure(2)
N_space=linspace(min(N),max(N));
plot(N_space,b./(N_spaceNmin)+Vmin,[color(count_color),''])
end
end
end
save('data\toluenebenzeneData.mat','data');

'Success!'

B.2

%store the datamatrix as a

%.mat file
%end of simulation

Stage change

This script changes the number of stages in the column in UniSim. The m-script shows how the changes
in matlab were implemented in UniSim by UPC scripts.

function []=stage_change(N,script_folder,hyCase)
%
%
%
%
%
%
%

Summary:
author:
contact:
changed:
Input:
Output:

Changes the number of stages in rigorous column

Tor Anders Marvik
????
25 feb 2012
N, script folder to unisim with the
script\change_stage.SCP
None

51

% Assumes:

nothing

fid = fopen([script_folder,'\change_stage.SCP'], 'r+');

%opens the UPC
%script change_stage.SCP
%and enables writing
fseek(fid, 148, 'bof');
%find symbol 148
fwrite(fid, num2str(N,'%10.14e\n'));
%writes the number of stages before
%symbol 148
fclose(fid);
%closes change_stage.SCP
hyCase.Application.PlayScript([script_folder,'\change_stage.SCP']); %plays
%change_stage.SCP in UniSim
end

B.3

curve fitting developed

Curve fitting the developed curve to the simulation series using the least square method.

%
%
%
%
%
%
%

Summary:
author:
contact:
changed:
Input:
Output:
Assumes:

Least square method for curve fitting

Tor Anders Marvik
????
25 feb 2012
Nmin (x(1)), b (x(2)), N, V, Vend
None
nothing

function error=least_square(x,N,V,Vend)

error=0;
Nmin=x(1);
b=x(2);
error = sum(((b./(NNmin)+Vend)V).^2);

%Initial error
%Guess sent from main
%Guess sent from main
%equation that the curve is
%fitted to

end

B.4

curve fitting Gilliland

Curve fitting the simulation series to the Molokanov equation 3.12 in a Gilliland plot.

52

B.4.1

%
%
%
%
%
%
%
%

Main gilliland script

Summary:
author:
contact:
changed:
Input:
Output:
Assumes:

Curve fit simulation series to Molokonov equation in

Gilliland plot
Tor Anders Marvik
????
20 March feb 2012
data files from simulation
None
nothing

clear all
clc
hold on
close all
load([pwd,'\data\ferdigkjrt\nbutanepropeneData'])
%give the plot different colors and symbols
color=['y';'m';'c';'r';'g';'b';'w';'k'];
symbol=['.','o','x','+','*'];

%for color in plot

%for symbol in plot

count_color=1;
count_symbol=1;
k=1;
lengde(1)=1;
figure(4)
hold on

xlabel('(VVmin)/V')
ylabel('(NNmin)/(N+1)')

for i=1:size(data,1)
if data(i,22)~=k
k=k+1;
lengde(k)=i;
end
end
lengde(28)=size(data,1)+1;

%loop that finds the size of each series

i=1;
for count_zd=1:3
for count_color=1:3
for count_symbol=1:3

%changes zf
%changes q
%changes zd

53

figure(4)
N=data(lengde(i):lengde(i+1)1,10);
V=data(lengde(i):lengde(i+1)1,9);
Vmin=1;
Nmin=1;
[min,error,flag] = fminsearch('gilliland_Vmin_Nmin',[Nmin...,
Vmin] ,[],N,V);
%find minimum value of the function in
%gilliland_Vmin_Nmin
Nmin=min(1);
Vmin=min(2);
xaxis=1Vmin./V;
yaxis=((NNmin)./(N+1));
plot(xaxis,yaxis,[color(count_color)... %plot simulation series
symbol(count_symbol),''])
%in Gilliland plot
data(lengde(i):lengde(i+1)1,23)=Vmin;
%New Vmin in
data(lengde(i):lengde(i+1)1,24)=Nmin;
%New Nmin in
data(lengde(i):lengde(i+1)1,25)=V./Vmin; %New V/Vmin
data(lengde(i):lengde(i+1)1,26)=N./Nmin; %New N/Nmin
figure(2)
xlabel('(VVmin)/V')
ylabel('(NNmin)/(N+1)')

data(:,23)
data(:,23)
data(:,23)
data(:,23)

%gilliland with curve fitting

hold on
[x_curv,y_curv]=plot_gilliland(data(lengde(i)... %calculate axis
:lengde(i+1)1,:));
plot(x_curv,y_curv,[color(count_color)...
%plot curve fit
symbol(count_symbol),''])
xlim([0,1])
i=i+1;
end
end
end

figure(4)
%Plot of the the molokanov equation
x=linspace(0,1);
y=1exp(((1+54.4.*x)./(11+117.2.*x)).*((x1)./sqrt(x)));
plot(x,y,'k')
figure(2)
%gilliland with curve fitting
x=linspace(0,1);
y=1exp(((1+54.4.*x)./(11+117.2.*x)).*((x1)./sqrt(x)));
plot(x,y,'k')

B.4.2

Gilliland function

54

%
%
%
%
%
%
%

Summary:
author:
contact:
changed:
Input:
Output:
Assumes:

Curve fit to Molokonov equation using least square method

Tor Anders Marvik
????
25 feb 2012
min, N, V
None
nothing

function error=gilliland_Vmin_Nmin(min,N,V)
Nmin=min(1);
Vmin=min(2);

x=1Vmin./V;
y=((NNmin)./(N+1));

%gilliland x axis
%gilliland y axis

m=1exp(((1+54.4.*x)./(11+117.2.*x)).*...
((x1)./sqrt(x)));

error=sum((my).^2);

B.4.3

%
%
%
%
%
%
%

%Molokonov equation

%calculates the difference

Gilliland plot

Summary:
author:
contact:
changed:
Input:
Output:
Assumes:

Plot the Molokonov curve fit in a Gilliland plot

Tor Anders Marvik
????
10 March 2012
data
x_curv, y_curv
nothing

function [x_curv,y_curv]=plot_gilliland(data)
N_Nmin=data(:,12);
V_Vmin=data(:,11) ;
N=data(:,10);
V=data(:,9);
Nmin=N./N_Nmin;
Vmin=V./V_Vmin;

y_curv=(NNmin)./(N+1);
x_curv=(VVmin)./V;

55

%N/Nmin
%V/Vmin

%calculate Nmin
%calculate Vmin

B.5

Analysis

Plot the average

B.5.1

%
%
%
%
%
%
%

V
Vmin

values against

N
Nmin .

The plot is discretized to give smooth curves.

Analysis

Summary:
author:
contact:
changed:
Input:
Output:
Assumes:

Find the average curve fit for the simulation cases

Tor Anders Marvik
????
26 feb 2012
None
None
nothing

clear all
close all
clc
hold on
figure(4)
xlabel('N/Nmin')
ylabel('V/Vmin')
hold on
N_Nmin=linspace(1,5);
%discretize from 1 to 5
load([pwd,'\data\Ferdigkjrt\toluenebenzeneData.mat']);%loads the .mat file
[b]=plot_curvfit(data);
%curve fits the data
V_Vmin_average=b./(N_Nmin1)+1;
plot(N_Nmin,V_Vmin_average,'r')

%find the average V/Vmin value

%plot the average V/Vmin value

load([pwd,'\data\Ferdigkjrt\1butanol2propanolData.mat'])
[b]=plot_curvfit(data);

V_Vmin_average=b./(N_Nmin1)+1;
plot(N_Nmin,V_Vmin_average,'c')
%
load([pwd,'\data\Ferdigkjrt\c4chloridec3chlorideData.mat'])
[b]=plot_curvfit(data);

56

V_Vmin_average=b./(N_Nmin1)+1;
plot(N_Nmin,V_Vmin_average,'g')
%
load([pwd,'\data\Ferdigkjrt\glycoleglycolData.mat'])
[b]=plot_curvfit(data);

V_Vmin_average=b./(N_Nmin1)+1;
plot(N_Nmin,V_Vmin_average,'b')

load([pwd,'\data\Ferdigkjrt\nheptanenhexaneData.mat'])
[b]=plot_curvfit(data);

V_Vmin_average=b./(N_Nmin1)+1;
plot(N_Nmin,V_Vmin_average,'k')
%
load([pwd,'\data\Ferdigkjrt\mcresolphenolData.mat'])
[b]=plot_curvfit(data);

V_Vmin_average=b./(N_Nmin1)+1;
plot(N_Nmin,V_Vmin_average,'m')
%

axis([1,3,1,5]);

%sets the axis to 13 in xdirection and

%15 in ydirection

legend('Toluene and Benzene','1butanol and 2propanol',...

'c4chloride and c3chloride','Glycole and Glycol',...
'nheptanen and Hexane','mcresol and PhenolData')

B.5.2

Plot curve

clear
Nmin=7;
Vmin=200;

57

N= [8 9 11 13 15 16 19 22 26 29 58];
V=[138.0300 96.5994 72.1627 61.8815 56.9952 56.2987 53.2823 52.3745 51.8653 51.7354 51.5799];
b=100;
c=10;
[x,error,flag] = fminsearch('least_squeare',[Nmin,Vmin,b],[],N,V);

Nmin=x(1);
Vmin=x(2);
b=x(3);
%c=x(4);
plot(N,V,'o')
hold on
N_curv=[8:0.1:40];

%for i=1:length(N_curv)
%V_curv(i)=b*exp(c/(N_curv(i)Nmin))+Vmin;
%end
V_curv=(b./(N_curvNmin)+Vmin);
plot(N_curv,V_curv)
plot([Nmin,Nmin],[0,max(V_curv)],'r')
plot([0,max(N)],[Vmin,Vmin],'r')
Vmin
Nmin

B.5.3

Last square function

function error=least_square_data(x,N,V)
%
%
%
%
%
%
%

Summary:
author:
contact:
changed:
Input:
Output:
Assumes:

Find the average curve fit for the simulation cases

Tor Anders Marvik
????
25 feb 2012
b (x(1)), N, V
None
nothing

error=0;
b=x(1);
error = sum(((b./(N1)+1)V).^2);

58

end

59

Average N/Nmin when fixed capital cost is considered

The figure below shows the average N/Nmin for different components when the properties for the components is changing. It is just the fixed capital cost that is considered. The different properties for the
point in the figures below is shown in figure D.1

Figure C.1: Properties for the different points for the figures below

(a) Cost method from Turton

(b) Cost method from Sinnott

Figure C.2: 1-Butanol and 2-Propanol

(a) Cost method from Turton

(b) Cost method from Sinnott

Figure C.3: 26-xylenol and phenol

In the case for separate 26-Xylenol and Phenol high pressure steam used instead of low pressure steam.
Since the inlet temperature that is being heated in the reboiler has a higher temperature then the temperature in the low pressure steam.

60

(a) Cost method from Turton

(b) Cost method from Sinnott

Figure C.4: C4-Chloride and C3-Chloride

(a) Cost method from Turton

(b) Cost method from Sinnott

Figure C.5: Ethanol and Merhanol

(a) Cost method from Turton

(b) Cost method from Sinnott

Figure C.6: Glycerol and Ethyl Glycol

For the case when Glycerol and Ethyl Glycol is being separate the inlet temperature in the reboiler is
higher the the temperature for high pressure steam. Then it is been assumed that the inlet temperature
for the steam in the reboiler is 40 oC higher then the outlet temperature for the steam.

61

(a) Cost method from Turton

(b) Cost method from Sinnott

Figure C.7: n-Decanen and n-Nonane

In the case for separate n-Decane and n-Nonane medium pressure steam used instead of low pressure
steam since the inlet temperature that is being heated in the reboiler is higher then the temperature for
the low pressure steam.

(a) Cost method from Turton

(b) Cost method from Sinnott

Figure C.8: n-Heptane and n-Hexane

(a) Cost method from Turton

(b) Cost method from Sinnott

Figure C.9: n-Nonanen and n-Octane

In the case for separate n-Nonane and n-Octane medium pressure steam used instead of low pressure
steam since the inlet temperature that is being heated in the reboiler is higher then the temperature for
the low pressure steam.
62

(a) Cost method from Turton

(b) Cost method from Sinnott

Figure C.10: n-Octane and n-Heptane

(a) Cost method from Turton

(b) Cost method from Sinnott

Figure C.11: m-Cresol and Phenol

In the case for separate m-Cresol and Phenol high pressure steam used instead of low pressure steam.
Since the inlet temperature that is being heated in the reboiler has a higher temperature then the temperature in the low pressure steam.

(a) Cost method from Turton

(b) Cost method from Sinnott

Figure C.12: p-Xylene and Benzene

63

(a) Cost method from Turton

(b) Cost method from Sinnott

Figure C.13: p-Xylene and Toluene

(a) Cost method from Turton

(b) Cost method from Sinnott

Figure C.14: Toluene and Benzene

64

Average N/Nmin after cost calculation

The figure below shows the average Nratio for different components when the overall cost is considered.
In the different points in the figure below the properties for the components is changing and properties
for the different point
The different properties for the point in the figures below is shown in figure D.1

Figure D.1: Properties for the different points for the figures below

(a) Cost method from Turton

(b) Cost method from Sinnott

Figure D.2: 1-Butanol and 2-Propanol

(a) Cost method from Turton

(b) Cost method from Sinnott

Figure D.3: 26-xylenol and phenol

In the case for separate 26-Xylenol and Phenol high pressure steam used instead of low pressure steam.
Since the inlet temperature that is being heated in the reboiler has a higher temperature then the temperature in the low pressure steam.

65

(a) Cost method from Turton

(b) Cost method from Sinnott

Figure D.4: C4-Chloride and C3-Chloride

(a) Cost method from Turton

(b) Cost method from Sinnott

Figure D.5: Ethanol and Merhanol

(a) Cost method Turton

(b) Cost method from Sinnott

Figure D.6: Glycerol and Ethylene Glycol

For the case when Glycerol and Ethylene Glycol is being separate the inlet temperature in the reboiler
is higher the the temperature for high pressure steam. Then it is been assumed that the inlet temperature
for the steam in the reboiler is 40 oC higher then the outlet temperature for the steam.

66

(a) Cost method from Turton

(b) Cost method from Sinnott

Figure D.7: n-Decanen and n-Nonane

In the case for separate n-Decane and n-Nonane medium pressure steam used instead of low pressure
steam since the inlet temperature that is being heated in the reboiler is higher then the temperature for
the low pressure steam.

(a) Cost method from Turton

(b) Cost method from Sinnott

Figure D.8: n-Heptane and n-Hexane

(a) Cost method from Turton

(b) Cost method from Sinnott

Figure D.9: n-Nonanen and n-Octane

In the case for separate n-Nonane and n-Octane medium pressure steam used instead of low pressure
steam since the inlet temperature that is being heated in the reboiler is higher then the temperature for
the low pressure steam.
67

(a) Cost method from Turton

(b) Cost method from Sinnott

Figure D.10: n-Octane and n-Heptane

(a) Cost method from turton

(b) Cost method from Sinnott

Figure D.11: m-Cresol and Phenol

In the case for separate m-Cresol and Phenol high pressure steam used instead of low pressure steam.
Since the inlet temperature that is being heated in the reboiler has a higher temperature then the temperature in the low pressure steam.

(a) Cost method from Turton

(b) Cost method from Sinnott

Figure D.12: p-Xylene and Benzene

68

(a) Cost method from Turton

(b) Cost method from Sinnott

Figure D.13: p-Xylene and Toluene

(a) Cost method from Turton

(b) Cost method from Sinnott

Figure D.14: Toluene and Benzene

69

E
E.1

Calculation example
Calculation of column diameter

Constants from simulation and assumptions

lt = 0.9m
L = 874.7425kg/m3
V = 3.1040kg/m3
Vg = 1.5776kg/s
uw

r
L V
= (0.171lt + 0.27lt 0.047)
V
s
2

uw = (0.171 0.9m2 + 0.27 0.9m 0.047)

(E.1)
874.7425kg/m3 3.1040kg/m3
3.1040kg/m3

uw = 0.9634 m/s
s
4Vg
D =
V uw
s
4 1.5776kg/s
D =
3.1040kg/m3 2.5385m/s

(E.2)

D = 0.8196 m

E.2

Calculation of shell weight

The calculate the wall thickness equation in Sinnott [12] is used and given below
P = 10bar
P = 1N/mm2
N
SE = 103.42 mm
2

D is the diameter in column and calculated above. L is the tray spacing and assumed to be 0.9 m
= 8030kg/m2
Pi Di
(2SE 1.2Pi )
t = 3.9856mm
t =

M = ptw DL

(E.4)
3

M = 1.3349 10 kg

70

(E.3)

E.3

Calculation of area

In the example the reboiler area is calculated. Constants from simulation and assumptions
Tin,reboiler = 108.8316oC
Tut,reboiler = 109.7654oC
Tin,steam = 159oC
Q = 5.7587 105W
U = 1000W /m2
Tout,steam = Tut,reboiler + 20oC = 129.7654oC
Q = UATlm
Q
A =
U Tlm
T1 T2
Tlm =
1
ln( T
T2 )

(E.5)
(E.6)
(E.7)

T1 = Tout,steam Tin,reboiler

(E.8)

T1 = Tin,steam Tout,reboiler

(E.9)

Tlm = 33.0339oC
A = 17.4327m2

E.4
E.4.1

Cost calculation of capital cost when Turton is used

Calculation of purchased equipment cost

In the example the purchased equipment cost for the reboiler is shown. Parameters for calculated the
cost for the reboiler is given below and found in Turton [13]
K1 = 4.4646
K2 = 0.5277
K3 = 0.3955
A is the area calculated above and is equal to 17.4327
log(C0p ) = K1 + K2 log(A) + K3 log(A)2

(E.10)

C0p = 1.7896 104 US$

In the example the bare module cost for the reboiler is shown. Parameters for calculated the bare module
cost for the reboiler is given below and found in Turton [13]
B1 = 1.63
B2 = 1.66
71

FM = 2.7 which is the material factor an is given for stainless steal

Fp = 1
C0p is the purchased cost and given above

CBM = C0p FBM

(E.11)

FBM = B1 + B2 FM Fp

(E.12)

CBM = 1.6039 105 US$

To find the total capital cost the two equations below is used.
CT M = 1.18CBM,i

(E.13)

0
CGR = CT M + 0.50CBM,i

(E.14)

0
CBM,i
is found when the material is carbon steal and the column operate with ambient pressure.
0
CBM,i
= C0p (B1 + B2 FM Fp )

(E.15)

B1 , B2 and Fp are equal to the value given above and FM are equal to one.
0
CBM,i
= 8.6335 104
0
CGR = 1.18CBM,i + 0.50CBM,i

(E.16)

CGR = 2.3224 105

This gives the value in USGC basis in 2001.

E.5

Cost calculation when Sinnott is used

In the example the cost for the reboiler is shown. The cost equation used below is found in Sinnott
a=25000
b=340
n=0.9
S it the area in the reboiler and is calculated above and is equal to 17.4327
Ce = a + bSn
Ce = 2.7486 104 US$

72

(E.17)

Ce gives the cost in carbon steal, to get the cost in stainless steal, Ce is multiplied with 1.3
Ce = 3.5732 104 US$
To find the fixed capital cost the new Ce value is multiplied with 6.
Ce = 2.1439 105 US$ in USGC basis for 2006

73