Open Access Journal

Journal of Power Technologies 96 (3) (2016) 194–199

journal homepage:papers.itc.pw.edu.pl

Mathematical modeling of an axial compressor in a gas turbine cycle

Marcin Plis∗, Henryk Rusinowski

Institute of Thermal Technology, Silesian University of Technology, Konarski st. 22, Gliwice 44-100, Poland

Abstract
Contemporary thermal diagnostic systems of power units require advanced computational tools, including mathematical mod-
els. These models should have a simple structure and short computing time. These conditions are satisfied by models that
include mass and energy balances as well as additional empirical functions whose coefficients are estimated by using the
measurement results. This paper presents a simulation model of an axial compressor which forms part of a gas turbine
unit with a rated output of 125.4 MW. The model was developed with the use of a generalized compressor map, describing
the relationship between corrected air mass flow rate, pressure ratio, isentropic efficiency and corrected rotational speed.
The model encompasses the empirical formula for compressor internal efficiency, which additionally considers the effect of
variation of the angle of inlet guide vanes. The unknown values of empirical coefficients which appear in this equation were
estimated on the basis of operating data. The calculation results obtained were compared with the measurement results. The
quality of the model prediction was evaluated and conclusions were drawn.

Keywords: mathematical modelling, gas turbine, axial compressor, empirical functions, inlet guide vanes of compressor

1. Introduction
Current thermal diagnostics systems of power plants require
mathematical models of thermal processes. The mathe-
matical model can be developed on the basis of physical
laws (analytical model) and as an approximation of mea-
sured data (empirical model). The advantage of using an-
alytical models is the ability to accurately understand the
process mechanism. These processes are often complex,
which makes it impossible to develop a model using only the
process laws of physics. In such cases, empirical models are
frequently used, which are easier to develop than analytical
models [1]. However, their scope of applicability is limited Figure 1: Schematic diagram of the axial compressor
to the operating parameters for which the model was cali-
brated. Good results are obtained by combining analytical
and empirical models. The model can be used to calculate mass flow rate, thermal
This article presents a simulation model of an axial com- parameters of compressed air and energy assessment in-
pressor which forms part of a gas turbine unit with a rated dexes, e.g., compressor efficiency.
output of 125.4 MW.
The model encompasses a generalized compressor map 2. Mathematical model of an axial compressor
and empirical equations describing isentropic efficiency of
the compressor and pressure drop in an air filter. The un-
known values of empirical coefficients were estimated on the Fig. 1 presents a schematic diagram of the modelled axial
basis of measurement results. As a result, the current tech- compressor.
nical condition of the modeled machine is taken into account. The mathematical model of the axial compressor encom-
passes:
∗ Corresponding author
Email address: marcin.plis@polsl.pl (Marcin Plis) • generalized compressor map,
 Journal of Power Technologies 96 (3) (2016) 194–199

• empirical equation describing isentropic efficiency, • the flow characteristics of the compressor, which de-
scribe the relationship between the corrected air mass
• empirical equation describing pressure drop in an air flow rate ṁcorr the pressure ratio PR the corrected speed
filter. Ncorr and the auxiliary variable β (the lower part of the
drawing),
The generalized compressor map is used to calculate air
mass flow rate and air pressure downstream of the com- • the characteristics of the internal efficiency showing the
pressor. The temperature of the compressed air is computed relationship between the internal efficiency of the com-
from the empirical equation describing isentropic efficiency. pressor ηC the corrected air mass flow rate ṁcorr the
corrected speed Ncorr and the auxiliary variable β (the
2.1. Compressor map upper part of the drawing).
In order to calculate the mass flow rate and thermal pa-
2.2. The flow characteristics of the compressor
rameters of compressed air, access to a real compressor
map is required. However, this information is often inacces- The generalized compressor map which is presented in
sible [2]. Fig. 2 cannot be directly used in the simulation program for
The problem of the lack of precise maps of the compres- two main reasons [3]:
sor may be solved by adopting generalised maps, which are
• in the area of high corrected speed, the speed lines
presented in a dimensionless system and with the assump-
Ncorr are almost vertical, which prevents precise deter-
tion that the geometry of the blade system is fixed [2, 3].
mination of the compressor operating point,
Moreover, the analyzed compressor is additionally equipped
with an inlet guide vanes (IGV) system, which changes the • at low rotational speeds, these lines are almost horizon-
geometry of the blade system during operation. tal.
The generalized map was used to develop the mathemati-
cal model of the compressor. The parameters present in the
compressor map were defined as follows [2–6]: Therefore, the additional, variable β is introduced. The β
lines are parallel to the surge line and vary from 0 to 1. The
• corrected air mass flow rate: line β = 0 indicates the maximum capacity of the compres-
√
ṁ1 · T 1
sor, and the line β = 1 fits in with the surge line. Fig. 3 shows
p1 the course of β lines on the flow characteristic.
ṁcorr =  √  (1)
ṁ1 · T 1 An analogous approach was introduced into the efficiency
p1 characteristics.
nom
The modeled compressor is also equipped with IGVs (in-
where: ṁ1 – air mass flow rate, kg/s, T 1 – compressor air
let guide vanes), which are designed to regulate the com-
inlet temperature K, p1 – compressor air inlet pressure kPa,
pressed air mass flow rate into the combustion chambers at
the index “nom” refers to nominal parameters.
partial loads of the gas turbine unit [3]. Varying the angle of
• pressure ratio: the IGVs modifies the fluid-dynamic behavior of the compres-
p  sor. This leads to changes in the flow characteristics [2, 6].
2
π2 p
The compressor map presented in Fig. 2 was developed
PR = =  p 1 (2) at the maximum IGV angle. The factor describing the influ-
π2 nom 2
p1 nom ence of IGV angle variation on air mass flow rated is called
where: p2 – compressor air inlet pressure, kPa. VACF (Vane Angle Correction Factor) [5]. This parameter de-
termines the percentage change of air mass flow rate when
• corrected rotational speed: inlet guide vanes are opened (or closed) by about 1◦ [7].
 
√n In order to calculate the value of VACF for the analyzed
T1
Ncorr =   (3) compressor, the characteristics of air volume stream as
√n a function of actual value of IGV angle was elaborated. The
T 1 nom
characteristics presented in Fig. 4 were developed on the
where: n– rotational speed of the machine, rpm. basis of operating data.
On the basis of the characteristics shown in Fig. 4 it was
The generalized map of the compressor operating in the calculated that the value of VACF for the analyzed compres-
gas turbine unit is shown in Fig. 2. Corrected rotational sor is 0.81%.
speed lines Ncorr present in the generalized map of the com- The compressed air mass flow rate, as a function of the
pressor were developed for a range from 81% to 106.2% angle of IGVs, can be calculated from the following equa-
of the corrected speed value in ISO conditions (T ambient = tion [5]:
15◦ C).
ṁ1 = ṁmax · [1 − V ACF · IGV max − IGV ]


The generalized map of an axial compressor includes [2, 5]: 1 (4)

— 195 —
Journal of Power Technologies 96 (3) (2016) 194–199

Figure 2: Generalized map of the analyzed axial compressor

Figure 3: Course of β line on flow characteristics

— 196 —
 Journal of Power Technologies 96 (3) (2016) 194–199

Table 1: Values of the estimated coefficients bi present in the empirical

equation for the internal efficiency of the compressor
Coefficient Value
b0 32.967
b1 97.489
b2 -28.028
b3 -91.749
b4 55.686
b5 -78.199
b6 -0.854
b7 -0.001
b8 0.422
b9 0.726

R2 99.5%
δ 0.6 K
Figure 4: Air volume stream variation as a function of IGV angle

where: ṁ1 – compressed air flow depending on the angle where: the superscript mod relates to the value obtained
of IGVs, kg⁄s; ṁmax
1 – compressed air mass flow read from the from the model and the superscript meas relates to the value
generalized characteristics of the compressor for the maxi- based on measurement data.
mum IGVs angle, kg/s; VACF – Vane Angle Correction Fac- Table 1 shows estimates of the coefficients bi . The quality of
tor which determines the effect of a variation in the angle model prediction was evaluated on the basis of determina-
of the inlet guide vanes on the compressed air mass flow, tion coefficient R2 and the mean square error δ [1].
VACF=0.81%; IGV max – the maximum value of opening the
inlet guide vanes, IGV max = 84◦ . 2.4. Algorithm for the calculation of the pressure drop in the
The generalized flow characteristics of the compressor air filter
presented in Fig. 2 were developed at the maximum IGV an- Before the air enters the compressor, it must be cleaned
gle. For the analyzed compressor, this angle is 84◦ . The of any contaminants, e.g. dust in the air filter, [2, 10]. The
variation of IGV angle does not change the shape of the lines empirical equation which describes the pressure drop in the
Ncorr , but they shift horizontally to the left when the IGVs air filter ∆pa f , as a function of corrected air mass flow rate
are being closed and to the right when the IGV are being takes the following form:
opened. The shape of the compressor surge line and the
line β remain unchanged [2, 3, 5]. This is illustrated in Fig.5. ∆pa f = 0.032 · ṁcorr1 − 0.908 (7)
The continuous lines relate to compressor operation with the
The equation (7) is characterized by the high determina-
inlet guide vanes closed at an angle of IGV=60◦ .
tion coefficient R2 =98.70%.
The effect of shifting the line Ncorr after changing the IGV
angle is included in the calculation algorithm. 3. Numerical calculation results

The presented generalized compressor map, together with

2.3. Equation for the internal compressor efficiency
empirical equations describing internal efficiency of com-
Calculation of the compressed air temperature on the ba- pressor (equation (5)) and pressure drop in the air filter
sis of internal efficiency characteristics produced highly er- (equation (7)) create a simulation model of the axial com-
roneous results. The mean square error of the compressed pressor.
air temperature was δ = 4.8 K. Therefore, it was decided to Simulation calculations were performed for 38 sets of
develop an empirical relationship for the internal efficiency of measurement data.
the compressor based on the measurement data [8]. The calculation results are presented in the form of dia-
grams. The dashed line represents the calculation results
ηC = b0 + b1 · ṁcorr + b2 · ṁ2corr + obtained from the model, and the continuous line the results
+b3 · Ncorr + b4 · Ncorr
2
+ b5 · ṁcorr · of operational measurements. The values of the determi-
(5)
·Ncorr + b6 · IGV + b7 · IGV 2 + nation coefficient R2 and the mean square error δ are also
+b8 · ṁcorr · IGV + b9 · Ncorr · IGV included [1, 9].

The unknown values of empirical coefficients were es-

timated, on the basis of measurements, using the least Fig. 6 and 7 confirm the very good agreement between the
squares method [5, 9]. The criterion of estimation takes the calculation results and the measurement results.
following form [1]: The determination factor for prediction of compressed air
38 
temperature is 99.5%. The good prediction quality is gained
X 2 by using an empirical equation which describes internal effi-
mod meas
T 2i − T 2i −→ min (6)
i=1
ciency of the compressor as a function of corrected air mass

— 197 —
 Journal of Power Technologies 96 (3) (2016) 194–199

Figure 5: Effect of change in the IGV angle on the compressor flow characteristics

Figure 6: Comparison of the compressed air temperature, calculated from the measurement data and from the model

— 198 —
 Journal of Power Technologies 96 (3) (2016) 194–199

Figure 7: Comparison of the compressed air pressure, calculated from the measurement data and from the model

flow rate, corrected rotational speed and actual value of inlet References
guide vane angle. [1] H. Rusinowski, Thermal diagnostic systems in exploitation. (in Polish),
The prediction quality of compressed air pressure is also Polish Academies of Sciences, 2010.
[2] P. P. Walsh, P. Fletcher, Gas turbine performance, John Wiley & Sons,
very good. The determination factor is high (R2 = 99.1%) 2004.
and the mean square error is low at 0.009 MPa. [3] A. Lazzaretto, A. Toffolo, Analytical and neural network models for gas
turbine design and off-design simulation, International Journal of Ther-
modynamics 4 (4) (2001) 173–182.
[4] H. Cohen, G. F. C. Rogers, Saravanamuttoo, Gas turbine theory, Long-
4. Summary man Group Limited, 1996.
[5] J. Kalina, Fossil fuel savings, carbon emission reduction and economic
attractiveness of medium-scale integrated biomass gasification com-
This article presents a mathematical model of the ax- bined cycle cogeneration plants, Thermal Science 16 (3) (2012) 827–
ial compressor, which forms part of a gas turbine unit with 848.
a rated output of 125.4 MW. The analyzed axial compressor [6] F. Haglind, Variable geometry gas turbines for improving the part-load
performance of marine combined cycles–gas turbine performance, En-
is additionally equipped with inlet guide vanes.
ergy 35 (2) (2010) 562–570.
The developed simulation model encompasses the flow [7] GateCycleTM version 6.0 Manual.
characteristics of the compressor and empirical functions de- [8] P. Wirkowski, Modelling the characteristics of axial compressor of vari-
scribing the internal efficiency of the compressor and pres- able flow passage geometry, working in the gas turbine engine system,
Polish Maritime Research (2007) 27–32.
sure drop in the air filter. The model was calibrated on the [9] G. Szapajko, H. Rusinowski, Theoretical-empirical model of the steam-
basis of measurement results with the use of recorded mea- water cycle of the power unit, Acta Montanistica Slovaca 15 (1) (2010)
sure results. The prediction quality of compressed air ther- 24.
mal parameters is satisfying. This is proven by the high val- [10] M. P. Boyce, Gas turbine engineering handbook, second edition Edi-
tion, Gulf Professional Publishing, 2001.
ues of the determination coefficient R2 and the low values of
the mean square error δ.
Owing to its simple structure and short computing time, the
presented simulation model of the compressor can be used
in prediction control models, hierarchical intelligent control
systems OCL and systems for thermal diagnostics of power
plants.

Acknowledgements

This work is co-financed by the charter budget resources

of the Institute of Thermal Technology, The Silesian Univer-
sity of Technology.

— 199 —