IOP Conference Series: Materials Science and Engineering

PAPER • OPEN ACCESS You may also like

- A Novel Double Chamber Rotary Sleeve
Compressor map prediction tool Air Compressor- Part III: Leakage Losses
Model
Y A Alduqri, H M Kamar, M N Musa et al.
To cite this article: Arjun Ravi et al 2015 IOP Conf. Ser.: Mater. Sci. Eng. 90 012042
- Modeling Solar Wind Variations over an 11
Year Cycle with Alfvén Wave Dissipation:
A Parameter Study
Soumitra Hazra, Victor Réville, Barbara
Perri et al.
View the article online for updates and enhancements.
- SUPERPOSITION OF POLYTROPES IN
THE INNER HELIOSHEATH
G. Livadiotis

This content was downloaded from IP address 188.212.124.194 on 23/07/2022 at 10:23

9th International Conference on Compressors and their Systems IOP Publishing
IOP Conf. Series: Materials Science and Engineering 90 (2015) 012042 doi:10.1088/1757-899X/90/1/012042

Compressor map prediction tool

Arjun Ravi, Lukasz Sznajder and Ian Bennett.

Shell Projects and Technology
Lange Kleiweg 40
2288GC Rijswijk, The Netherlands

Abstract. Shell Global Solutions uses an in-house developed system for remote condition
monitoring of centrifugal compressors. It requires field process data collected during operation
to calculate and assess the machine’s performance. Performance is assessed by comparing live
results of polytropic head and efficiency versus design compressor curves provided by the
Manufacturer. Typically, these design curves are given for specific suction conditions. The
further these conditions on site deviate from those prescribed at design, the less accurate the
health assessment of the compressor becomes. To address this specified problem, a compressor
map prediction tool is proposed. The original performance curves of polytropic head against
volumetric flow for varying rotational speeds are used as an input to define a range of Mach
numbers within which the non-dimensional invariant performance curve of head and volume
flow coefficient is generated. The new performance curves of polytropic head vs. flow for
desired set of inlet conditions are then back calculated using the invariant non-dimensional
curve. Within the range of Mach numbers calculated from design data, the proposed
methodology can predict polytropic head curves at a new set of inlet conditions within an
estimated 3% accuracy. The presented methodology does not require knowledge of detailed
impeller geometry such as throat areas, blade number, blade angles, thicknesses nor other
aspects of the aerodynamic design - diffusion levels, flow angles, etc. The only required
mechanical design feature is the first impeller tip diameter. Described method makes
centrifugal compressor surveillance activities more accurate, enabling precise problem
isolation affecting machine’s performance.

1. Introduction
In the oil and gas industry, the remote performance monitoring of critical machinery is essential to
reduce downtime, maintenance costs and improve production rates of entire installations. A common
machine that is central to most production and processing assets is the centrifugal compressor.
Continuous performance assessment plays a vital part in evaluating the health of these machines. The
performance of a centrifugal compressor is usually described by how the polytropic head varies with
suction volume flow over a wide range of rotational speeds. Original equipment manufacturers (OEM)
typically provide the design curves of polytropic head and efficiency, which serve as a benchmark in
assessing the machine’s performance - but for only a specific set of inlet conditions.
Ideally, a compressor would operate close to the OEM design conditions. In reality, the process
conditions such as temperature, pressure, molecular weight and rotational speed deviate, sometimes
substantially. This deviation impacts the accuracy of the performance assessment. There comes a point
when the performance map provided by the manufacturer is no longer valid as a reference at new
suction conditions. Accurate performance assessment of compressors requires a map that is valid for
varying sets of suction conditions and rotational speeds. Unfortunately, obtaining new performance

Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution
of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
Published under licence by IOP Publishing Ltd 1
9th International Conference on Compressors and their Systems IOP Publishing
IOP Conf. Series: Materials Science and Engineering 90 (2015) 012042 doi:10.1088/1757-899X/90/1/012042

curves for an existing compressor from the vendor can be time consuming and resource intensive.
Therefore, the methodology of predicting the compressor map at changing process conditions is
proposed.
There are many approaches that can be adopted to obtain a compressor map. One of these is to
create a numerical model of the centrifugal compressor stage and conduct CFD simulations. An
alternative approach is to use empirical correlations for predicting performance maps. A detailed
information of the stage geometry is needed; at least on a one dimensional (1D) basis. The detailed
geometry is often unknown as this is propriety information of the vendor. Examples of 1D methods
are given by [1], [2], [5].This paper presents a methodology that requires neither aerodynamic
information (i.e. diffusion ratios, flow angles) nor the impeller geometry such as the throat areas, blade
thickness and blade angles. It only requires the design performance curve from the OEM and the
impeller tip diameter which is typically provided in the datasheet of the machine.
The following paper is structured as follows. Section 2 describes the methodology regarding how
to correct the original performance map for a different set of suction conditions, limitations, validation
and results. Conclusions are presented in the last section.

1.1. Nomenclature
Dimp - Impeller tip diameter (m)
k - Specific heat ratio (-)
Mu - Tip speed Mach number (-)
MW - Molecular weight (kg/kmol)
Q - Volume flow rate (m3/s)
t - Temperature (°C)
U - Impeller blade tip speed (m/s)
R - Universal gas constant [J/ (mol K)]
Z - Compressibility (-)
Hp - Polytropic Head (kJ/ kg)
N - Rotational speed (rpm)
S - Cubic spline polynomial
a, b, c, d - Coefficients of spline polynomial (-)
σ - Second derivative of cubic spline polynomial
h - Equidistant spacing
m - Number of intervals in the piecewise cubic spline polynomial
i - Variable for iteration
Greek Symbols
Φ - Volume flow coefficient (-)
Ψ - Head coefficient (-)
Subscripts
OEM - Original Equipment Manufacturer
design - design inlet conditions
new - new inlet conditions
suc - suction conditions
p - polytropic

2. General methodology and verification.

The process inlet conditions such as temperature, compressibility, specific heat ratio, molecular weight
and circumferential speed all determine the compressor map. One non-dimensional parameter that
takes all these factors into account is the tip speed Mach number. The key principle is that if the Mach
numbers are the same for two different sets of suction conditions, then their corresponding non-
dimensional curve of head coefficient vs. volume flow coefficient are equivalent. This approach

2
9th International Conference on Compressors and their Systems IOP Publishing
IOP Conf. Series: Materials Science and Engineering 90 (2015) 012042 doi:10.1088/1757-899X/90/1/012042

follows similitude with equivalent Mach numbers, flow and head coefficient for the design and the
new set of inlet conditions. An equivalent non-dimensional curve also ensures aerodynamic similarity
which is important to simulate the equivalent flow of gas through the impellers. An equivalent flow
coefficient for the design and the new set of suction conditions ensures the velocity triangles at the
inlet of the stage to be the same. To visualize this concept a sample calculation has been done for two
different set of suction conditions, for which the proximity of Mach numbers i.e. 0.517 to 0.5 was
obtained. All the values are depicted in Table 1. Original polytropic head curves for these two set of
inlet conditions provided by the OEM are shown in Figure 1. Corresponding non-dimensional curves
obtained after conversion are displayed in Figure 2. Good agreement between these two curves
validates the mentioned hypothesis. In both the graphs the red curve represents the design suction
conditions and the black curve represents the new suction conditions.
Table 1 Set of inlet conditions and Mach numbers for corresponding performance curves depicted in
Figure 1 and Figure 2.
Temperature Pressure Molecular Rotational Mach
[°C] [bara] weight speed number[-]
[kg/kmol] [rpm]
Design inlet
29.9 30 22.59 6778 0.517
conditions
New inlet
40.9 30 17.24 7746 0.5
conditions

90 3.90

80 3.40

70
Head coefficient (Ψ ) [-]

2.90
Polytropic head [kJ/kg]

60
2.40
50
1.90
40
1.40
30

20 0.90

10 0.40
5000 15000 25000 35000 0.02 0.07 0.12 0.17
Volume flow [m3/h] Volume flow coefficient (Ф) [-]

Figure 1 Polytropic head curves for different Figure 2 Equivalent non-dimensional curves for
sets of inlet conditions. different sets of inlet conditions with the similar
Mach number.
To present an example of the complete curve prediction methodology, data from a randomly
chosen centrifugal compressor was used. The inputs are the design performance map of a variable
speed compressor and its corresponding suction conditions i.e. pressure, temperature and gas

3
9th International Conference on Compressors and their Systems IOP Publishing
IOP Conf. Series: Materials Science and Engineering 90 (2015) 012042 doi:10.1088/1757-899X/90/1/012042

composition for which the head curves were obtained. With the use of the presented methodology the
new compressor map i.e. polytropic head (Hp) vs. volume flow rate (Q) relationships for a new set of
suction conditions will be derived.

The coordinates of Hp and Q of original compressor map provided by the OEM at design suction
conditions and varied N are shown in Table 2.

Table 2 Polytropic head (Hp) and volumetric flow (Q) for design suction conditions and varying
rotational speeds (N) provided by the OEM. The rotational speeds are in % of the rated speed.

N1 N2 N3 N4 N5
100% 90% 80% 70% 55%
Q1 Hp1 Q2 Hp 2 Q3 Hp 3 Q4 Hp 4 Q5 Hp 5
1 20171.2 137.7 15462.1 110.1 12505.4 89.4 10559.1 65.4 8152.6 41.3
2 22837.4 133.3 18499.0 110.2 15539.7 85.0 13594.6 63.2 10451.3 37.9
3 25593.6 125.4 21716.0 103.5 18113.8 80.5 16535.5 56.4 12289.8 34.6
4 27978.4 111.7 24562.2 92.2 20776.0 69.2 18923.0 47.4 14402.5 27.8
5 31183.5 84.4 28132.2 59.2 24535.3 45.3 21764.6 28.1 17246.1 11.9

2.1. Main calculation procedure

The design compressor map conversion to new suction conditions is done with the use of following
non-dimensional parameters. Head coefficient (Ψ), volume flow coefficient (Ф) and Mach number
(Mu) are calculated using outlined set of equations (2),(3) and (4) according to [3].
Mentioned equations require the conversion of N into impeller blade tip speed (U), which is done
as follows:
N ( 1)
U=Dimp
60

2H p
ψ (2)
U2

4Q (3)

πDimp
2
U

U
Mu 
 
(4)
k suc Z suc 273.15+t suc R MW (-1 )

The polytropic head and volume flow curves presented in Table 2 are converted to non-
dimensional curves in the form of Ψ and Ф for varied Mach numbers using the abovementioned
equations. Mach number is varying as the rotational speed is changing. Obtained coordinates of
converted curves, now in the form of non-dimensional curves are presented in Table 3 and visualized
in Figure 3. Tip speed Mach number Mu1 corresponds to the non-dimensional curve for the reference
rotational speed i.e. N1 as shown in Table 2. The same logic holds for Mu 2, Mu 3, Mu 4 and Mu 5.This
non-dimensional map for varying Mach number is used as a baseline for the calculation of the
compressor map for the new set of inlet conditions.

4
9th International Conference on Compressors and their Systems IOP Publishing
IOP Conf. Series: Materials Science and Engineering 90 (2015) 012042 doi:10.1088/1757-899X/90/1/012042

Table 3 The non-dimensional performance map values of head coefficient (Ψ) and volume flow
coefficient (Ф) for varying Mach number (Mu) at design suction conditions.

Mu1 Mu2 Mu3 Mu4 Mu5

0.74 0.67 0.59 0.52 0.407
Ф [-] Ψ [-] Ф [-] Ψ [-] Ф [-] Ψ [-] Ф [-] Ψ [-] Ф [-] Ψ [-]
1 0.087 3.595 0.074 3.554 0.067 3.662 0.065 3.479 0.064 3.567
2 0.097 3.516 0.087 3.524 0.082 3.533 0.080 3.424 0.080 3.369
3 0.108 3.338 0.100 3.363 0.096 3.276 0.096 3.201 0.096 3.007
4 0.119 3.016 0.114 2.995 0.111 2.853 0.111 2.744 0.111 2.420
5 0.134 2.168 0.134 1.892 0.132 1.813 0.134 1.478 0.135 0.996

4.30
Mach
3.80 Number
(Mu)

3.30
Head Ceofficient (Ψ ) [-]

0.74

2.80 0.67

2.30 0.59

0.52
1.80
0.407
1.30

0.80
0.05 0.07 0.09 0.11 0.13 0.15
Volume flow coefficient (Φ) [-]

Figure 3 Invariant non-dimensional curves of head coefficient (Ψ) and volume flow coefficient (Ф)
for different Mach numbers at design inlet conditions. The legend describes the impeller tip speed
Mach number (Mu) for each curve.
At this point the Mach numbers for new set of inlet condition and shaft speeds are calculated
(Munew). To draw the full non-dimensional curves for new conditions i.e. new head coefficient (Ψnew)
against new volume flow coefficient (Фnew) the curve approximation method is used.
The mathematical method employed by the Authors was cubic spline interpolation technique,
which is briefly presented in subchapter 2.2.1. It allowed obtaining coordinates of non-dimensional
Ψnew vs. Фnew curves for desired Munew based on existing Ψ vs. Ф curves for different Mu.
In Table 4 the new coordinates of head coefficient (Ψnew)and volume flow coefficient (Фnew)
obtained in given example are presented The same is visualised in Figure 4

5
9th International Conference on Compressors and their Systems IOP Publishing
IOP Conf. Series: Materials Science and Engineering 90 (2015) 012042 doi:10.1088/1757-899X/90/1/012042

Table 4 Volume flow coefficient (Фnew ) and Head coefficient (Ψnew) for new suction conditions.

Munew= 0.65 Munew= 0.62 Munew= 0.56 Munew= 0.50 Munew= 0.44
Фnew Ψnew Фnew Ψnew Фnew Ψnew Фnew Ψnew Фnew Ψnew
1 0.0717 3.5732 0.0689 3.6353 0.0662 3.6042 0.0646 3.4536 0.0640 3.5042
2 0.0854 3.5300 0.0831 3.5410 0.0810 3.4948 0.0802 3.4024 0.0798 3.3720
3 0.0992 3.3568 0.0972 3.3219 0.0958 3.2440 0.0957 3.1760 0.0956 3.0666
4 0.1130 2.9796 0.1114 2.9206 0.1105 2.8070 0.1112 2.7046 0.1114 2.5223
5 0.1336 1.8701 0.1326 1.8525 0.1326 1.7038 0.1344 1.3848 0.1351 1.1142

Non-dimensional curves for new set of inlet conditions

3.90
Mach
3.40 number
new
Head coefficient (Ψ) [-]

2.90 conditions
0.65
2.40
0.62
1.90 The red curves are the non-
dimensional curves obtained 0.56
1.40
0.50
0.90
0.44
0.40
0.050 0.070 0.090 0.110 0.130 0.150
Volume flow coefficient (Ф) [-]

Figure 4 The interpolated non-dimensional curves of new head coefficient (Ψnew) and volume flow
coefficient (Фnew) for varying Mach number at new suction conditions.
In Figure 4, the black curves with different markers represent the non-dimensional curves for
varying Mach numbers for the new set of inlet conditions. The red curves are the non-dimensional
curves for design suction conditions from Figure 3.In the last calculation step the coordinates of non-
dimensional curves consisting of Ψnew and Фnew from Table 4 are recalculated back to obtain the actual
polytropic head curves i.e. new polytropic head (Hp_new) vs. volumetric flow rate (Qnew) for desired set
of process conditions and shaft speeds. It is done with the use of equations ( 5) and ( 6)
2
U new
H pnew   new ( 5)
2
 ( 6)
Qnew   newU new Dimp
2
4
To validate the methodology, the newly obtained compressor map was compared to the available
map that was generated by the OEM at the same suction conditions during the machine’s performance
test. These curves are shown in Figure 5. The accuracy achieved in polytropic head prediction with the
use of described methodology reaches c.a. 2-3% of error when predicted curve is compared to the
original OEM’s characteristic. The specific results regarding the accuracy of the method are depicted

6
9th International Conference on Compressors and their Systems IOP Publishing
IOP Conf. Series: Materials Science and Engineering 90 (2015) 012042 doi:10.1088/1757-899X/90/1/012042

in Table 5. Such level of accuracy is regarded as satisfactory.

Comparsion of OEM Vs Mapping tool. Polytropic head curves for different

rotational speeds
160 The black
curves-
Mapping tool.
140
Red markers-
OEM
120
Polytropic Head [kJ/kg]

10167
100
9683
80 8715
7746
60
6778
10167
40
9683
20 8715
7746
0
10000 15000 20000 25000 30000 35000 6778
Volume flow [m3/h]
Figure 5 Comparison of compressor map generated by compressor mapping tool and map provided by
the OEM. The red markers are the polytropic head curves from the OEM. The black curves are
obtained from the compressor mapping tool.

Table 5 Validation of Compressor Mapping prediction tool with the values obtained from the OEM
for the Rotational speed (N) = 9683 rpm.
Flow Polytropic head OEM Polytropic head Deviation in Polytropic
[m3/h] [kJ/kg] Mapping tool[kJ/kg] head[%]
1 15991.0 136.0 138.7 2.0%
2 19277.3 133.3 135.2 1.4%
3 22563.7 124.1 126.7 2.2%
4 25850.0 109.1 111.3 2.0%
5 30779.5 71.7 70.4 1.8%

This method is only applicable when each stage of a centrifugal compressor is treated separately. Each
stage refers to the number of impellers stacked together between the suction and discharge of a
compressor.

7
9th International Conference on Compressors and their Systems IOP Publishing
IOP Conf. Series: Materials Science and Engineering 90 (2015) 012042 doi:10.1088/1757-899X/90/1/012042

2.2. Method limitations

One should remember that the accuracy of the method depends on the range of the operating envelope
of the invariant non-dimensional curve varying with Mach number. Higher inaccuracies in curve
approximation are seen when curve extrapolation is used.
The following convention is employed to define the upper and lower boundary of Mach numbers at
design inlet conditions for which the presented methodology renders accurate results.
Mulow - Mach number at design inlet conditions for lowest N;
Muhigh - Mach number at design inlet conditions for highest N;
Extrapolation limits the accuracy of presented method. Therefore the following constraints are
imposed at newly calculated Munew. It is checked whether the Munew is within the range of Mach
numbers calculated for the design conditions as shown below. There are three possible cases:
1. Mulow<Munew<Muhigh – the newly obtained Mach number is between boundary values of lowest
and highest Mach numbers, therefore the interpolation method will be used to calculate
coordinates of new head and volume flow coefficient curve. Experience has shown that
interpolation allows to obtain non-dimensional curve approximation with high accuracy;
2. Munew<Mulow – the newly obtained Mach number is smaller than the lowest Mach number
obtained for design inlet conditions, therefore the extrapolation method will be used to calculate
coordinates of new head and volume flow coefficient curve;
3. Munew>Muhigh – the newly obtained Mach number is greater than the highest Mach number
obtained for design inlet conditions, therefore the extrapolation method will be used to calculate
coordinates of new head and volume flow coefficient curve;
In 2nd and 3rd case the following condition must be met to ensure an acceptable accuracy of the curve
approximation:
 Mu low  Mu new 
 Check if    5% (case 2)
 Mu new 

 Mu new  Mu high 
 Check if    5% (case 3)

 Mu new 

The three conditions are visualised in Figure 6. The red curves are the invariant non-dimensional
curves for design inlet conditions. The dotted curves are the extrapolated curves corresponding to case
2 and 3. The dashed curve is the interpolated curve corresponding to case 1.
Extrapolation or interpolation of the new non-dimensional curve of Ψnew vs. Фnew for the new Munew
is done depending on which of the three aforementioned conditions is met. The accuracy of the
extrapolation largely deteriorates and the method is rejected for deviations in new Mach number
greater than 5% from the highest or lowest design Mach number, as it was mentioned earlier.
It must be also mentioned that this method is only applicable when each stage of a centrifugal
compressor is treated separately. Each stage refers to the number of impellers stacked together
between the suction and discharge of a compressor.

8
9th International Conference on Compressors and their Systems IOP Publishing
IOP Conf. Series: Materials Science and Engineering 90 (2015) 012042 doi:10.1088/1757-899X/90/1/012042

Extrapolated Invariant non-dimensional curve Interpolated

2
1.8
1.6
Head coefficient (Ψ) [-]

1.4
1.2
1 Muhigh
0.8
0.6
Mulow
0.4
0.04 0.05 0.06 0.07 0.08 0.09 0.1 0.11 0.12
Volume flow coefficient (Φ) [-]
Figure 6 Interpolated (dashed) and extrapolated (dotted) non-dimensional curves for new set of inlet
conditions from invariant non-dimensional curves

2.2.1. Curve approximation method

Mathematical method used to interpolate or extrapolate curves which was employed in the presented
curve prediction tool is the Cubic Spline method. This technique provides smooth continuous
approximated curves compared to piecewise linear interpolation. Cubic spline interpolation formula is
continuous in the first and second derivatives and both within the intervals and interpolating nodes. It
also provides an interpolating polynomial which has a smaller error compared to other interpolating
polynomials such as Lagrange or Newton’s A brief description of the original cubic spline
interpolation method employed in the presented performance map prediction tool is provided
hereunder.
f=CubicSpline(xinput, x, y)=Si(x) (7)

 xinput is the new Mach number for new set of inlet conditions;
 x is the dataset comprising known values of Ф obtained at design inlet conditions;
 y is the dataset comprising known values of Ψ obtained at design inlet conditions;

Spline interpolation is divided into small sub-interval and each of these sub-intervals is interpolated
by using a third degree polynomial. According to [4], for a set of m+1 data points (xi, yi) and where no
two xi are the same on a given interval, a cubic spline (Si ) can be defined on each sub-interval (xi,xi+1)
by
S i x   ai (xinput  xi )3+bi (xinput  xi )2+c i (xinput  xi )+d i (8)

For m piecewise polynomial, four times m boundary conditions are needed to determine the
coefficients ai, bi, ci, and di. The boundary conditions are described in the equations outlined below:

Si  xi   yi for i=0 : m  1 (9)

S m1  y m (10)
S i xi 1   S i 1 xi 1  for i  0 : m  2 (11)

9
9th International Conference on Compressors and their Systems IOP Publishing
IOP Conf. Series: Materials Science and Engineering 90 (2015) 012042 doi:10.1088/1757-899X/90/1/012042

S i' xi 1   S ' i 1 xi 1  for i  0 : m  2 (12)

S i" xi 1   S "i 1 xi 1  for i  0 : m  2 (13)

S 0" x0   S "m1 xm   0 (14)

A new variable σ for the second derivative of the polynomial is introduced. Another variable h is
defined for equidistant spacing.
 i  S " xi  (15)

h  xi 1  xi

By using the equations (8), (9), (10), (11), (12), (13), (14) and (15)the coefficients ai, bi, ci, di are
obtained.

i  i 1   i y i 1  y i 2   i 1 (16)
d i  y i ; bi  ; ai  ; ci  h i
2 6h h 6
3. Conclusion and Limitations
In this paper, a generic methodology is proposed, which enables correction of the original polytropic
head curves for actual operating conditions based on the equivalence of non-dimensional parameters.
The general procedure and validation of the method has been described. With minimal inputs
of process conditions, first tip impeller diameter and a design performance map from the
manufacturer, a compressor performance map for a new set of inlet conditions can be predicted within
an accuracy of 2-3%.
The effective implementation of this method in remote condition monitoring system enables actual
compressor performance assessment with the use of corrected polytropic head curves, which more
accurately reflect the theoretical behaviour of a healthy machine at given conditions. Additionally, it
can aid the performance review of the machine for new operating conditions and advise the engineers
on site if the compressor can operate at the new conditions. Another key feature of this method is the
accurate locations of the operating points on the compressor maps, which is essential for surge and
choke protection. The presented feature can eventually facilitate isolation of undesirable compressor
behaviour at an early stage and, therefore, help identify potential failure modes, thus reducing
downtime and maintenance costs of the compressor and surrounding equipment. The method has been
validated for 20 different centrifugal compressors, thus proving its readiness.

References
[1] Aungier R H 2000 Centrifugal Compressor- A Strategy for Aerodynamic Design and
Analysis (New York, USA: ASME Press).
[2] Cumpsty N A 1989 Compressor Aerodynamics (Harlow, Essex, UK: Longman Scientific
&Technical).
[3] Lüdtke K H 2004 Process Centrifugal Compressors (Berlin: Springer)
[4] Plato R. 2003 Concise Numerical Mathematics vol 57 (Providence R.I: American
Mathematical Society).
[5] Swain E 2005 Improving a one-dimensional centrifugal compressor performance prediction
method Proc of the Institution of Mechanical Engineers Part A: Journal of Power and
Energy 219(8) pp.653-659.

10