Proceedings of

ASME TURBO EXPO 2002

June 3–6, 2002, Amsterdam, The Netherlands

GT-2002-30275

GAS TURBINE PERFORMANCE PREDICTION BY USING GENERALIZED

PERFORMANCE CURVES OF COMPRESSOR AND TURBINE STAGES

P.R. Spina
DIEM - University of Bologna
Viale del Risorgimento, 2 – 40136 Bologna, Italy

ABSTRACT nm number of measured (or available) data

nop number of operating points
The paper presents a method for gas turbine performance nx number of "X" unknown parameters
prediction which uses compressor and turbine performance P power
maps obtained by using generalized stage performance curves p pressure
matched by means of the “stage–stacking” procedure. In Q computed or measured parameter
particular, the overall multistage compressor performance is R gas constant
predicted using generalized relationships between stage ∆h rot
R = degree of reaction
efficiency, pressure coefficient and flow coefficient, while the (∆h 0 )stage
multistage turbine performance is predicted by modeling each
turbine stage by a series of two nozzles, a fixed one (stator) and SF Shape Factor
a moving one (rotor). T temperature
The characteristic of the proposed method is that the U blade speed at the mean radius
unknown parameters defining the generalized stage V absolute flow velocity
performance curves are determined by combining a Cycle W relative flow velocity
Program with the compressor and turbine performance maps w weight
obtained using the “stage–stacking” procedure, and by X unknown parameter
searching for the values of the unknown parameters which Y total pressure loss coefficient
better reproduce, by means of the Cycle Program, the overall β pressure ratio
performance and thermodynamic data measured on a gas V
turbine. φ = a flow coefficient
U
1 T dT
Φ =
RT
∫ c p (T) T + Φ (Tr )
NOMENCLATURE r
∆h 0s
A area ψ = pressure coefficient
cp specific heat at constant pressure U2
cv specific heat at constant volume ∆h 0
Ψ = stage aerodynamic loading coefficient
h enthalpy U2
i incidence angle η efficiency
k = cp/cv
M T0
M mass flow rate µ = corrected mass flow rate
N rotational speed p0
n number of stages

1 Copyright © 2002 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 02/01/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 N In both uses, it is required that the Cycle Program accurately
ν = corrected rotational speed
T0 reproduces the gas turbine performances, since, otherwise, both
thermodynamic and thermoeconomic analyses and gas turbine
∆h 0 ψ operating state determination would be affected by errors due to
ζ = = temperature coefficient
U 2 η Cycle Program inaccuracy. In particular, the case of gas turbine
operating state determination is very critical, since the errors in
Subscripts and Superscripts characteristic parameter estimation due to Cycle Program
* normalized value with respect to the reference value inaccuracy may be of the same order as parameter variations
0 total physical state due to an actual loss in gas turbine performance (Pinelli and
a ambient, axial Spina, 2002). The major problem in Cycle Program set up is the
C compressor lack of component data, and, in particular, of compressor and
c computed value turbine performance maps. These data are in fact proprietary to
f fuel the gas turbine manufacturers and they are not usually available.
GT gas turbine Different methods exist to predict compressor and turbine
i inlet conditions to i-th stage, inlet section performances, as, for example, the use of non-dimensional
i+1 outlet conditions from i-th stage = inlet conditions to component maps and scaling techniques (Saravanamuttoo and
(i+1)-th stage Mac Isaac, 1983; Kurzke and Riegler, 2000), or the use of
m measured value “stage–stacking” procedures to obtain the overall multistage
o outlet section compressor and turbine maps by using generalized stage
r reference value performance curves (Stone, 1958; Doyle and Dixon, 1962;
rel relative physical state Robbins and Dugan, 1965; Howell and Calvert, 1978). This
rot rotor second method seems more general and permits a detailed
stage-by-stage one-dimensional analysis of the flow through
s isentropic transformation
compressor and turbine. Several authors, for example, have
stat stator
used this second method to investigate the effects of compressor
T turbine
stage deterioration on compressor and gas turbine performance
th throat
(Saravanamuttoo and Lakshminarasimha, 1985; Aker and
Saravanamuttoo, 1988; Seddigh and Saravanamuttoo, 1990;
Tabakoff et al., 1990; Massardo, 1991; Cerri at al., 1993;
INTRODUCTION
Lakshminarasimha et al., 1994; Procacci and Rispoli, 1995).
The paper presents a method for the prediction of
In order to perform thermodynamic and thermoeconomic
compressor and turbine performance maps which uses
analyses on gas turbines, programs for thermodynamic cycle
generalized stage performance curves matched by means of the
calculation (Cycle Programs) have been widely used for many
“stage–stacking” procedure. In particular, the overall multistage
years to predict machine performance and key thermodynamic
compressor performance is predicted using generalized
parameters (such as power output, heat rate, turbine outlet
relationships between stage efficiency, pressure coefficient and
temperature, fuel and exhaust mass flow rates, etc….) at
flow coefficient (each of them normalized with respect to a
different loads and boundary conditions (Glassman, 1974;
reference value). The stage-by-stage evaluation of the outlet
Waters and Associates Inc., 1983; Stecco et al., 1985; El-Masri,
conditions starting from the knowledge of those at the inlet is
1988; Consonni and Macchi, 1988; Thermoflow Inc., 1989,
then possible. The multistage turbine performance is instead
1990). In the last two decades, Cycle Programs have also been
predicted by modeling each turbine stage by a series of two
used more and more frequently for gas turbine operating state
nozzles, a fixed one (stator) and a moving one (rotor).
determination (Saravanamuttoo and Mac Isaac, 1983; Aker and
As only a little qualitative information is usually available
Saravanamuttoo, 1988; Muir et al., 1989; Ping Zhu and
(or estimable) about stage design, generalized stage
Saravanamuttoo, 1992; Benvenuti et al., 1994; Bettocchi et al.,
performance curves depending on few parameters are used. The
2000, 2001; Tsalavoutas et al., 2000). This can be performed by
characteristic of the proposed method is that the unknown
using Gas Path Analysis (GPA) techniques which use
parameters are determined by combining a Cycle Program with
measurements taken on the machine to calculate the
the compressor and turbine performance maps obtained using
characteristic parameters that are indices of the gas turbine
the “stage–stacking” procedure, and by searching for the values
health state (such as efficiencies and characteristic flow passage
of the unknown parameters which better reproduce, by means of
areas of compressor and turbines, combustor efficiency and
the Cycle Program, overall performance and thermodynamic
pressure drops along the gas path). These techniques frequently
data (such as compressor pressure ratio, fuel consumption,
make use of a Cycle Program solved in inverse mode to
turbine outlet temperature, etc….) measured (or available) on a
estimate the characteristic parameters (Stamatis et al., 1990;
gas turbine. In particular, the unknown parameters are
Benvenuti et al., 1993; Bettocchi and Spina, 1999).
determined in an iterative way by using a constrained

2 Copyright © 2002 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 02/01/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 minimization algorithm, which repeatedly calls the Cycle 1 − η*(ψ / φ )  ψ*
3.5
ψ *  ψ *  
Program and modifies the parameters defining the generalized η* = 1 − 1 −
min
 , ∈   , 1 (2)
3.5  φ *  φ *  φ *  min 
stage performance curves, in order to minimize the sum of the   ψ*  
square differences between performance and thermodynamic 1 −   
data measured (or available) on the gas turbine and computed   φ *  min 
by the Cycle Program.
1 − η*(ψ / φ )
 ψ* 
2
ψ*   ψ* 
η* = 1 −  − 1 ,
max
∈ 1 ,    (3)
  φ*  φ *   φ *  max 
2
 ψ * 
COMPRESSOR PERFORMANCE PREDICTION   − 1
 φ *  max 
The procedure to obtain the performance maps of a Fig. 2 shows the generalized stage efficiency curve obtained
multistage compressor, which link together overall pressure by using equations (2) and (3) compared to data obtained from
ratio (βC), efficiency (ηC), corrected mass flow (µC) and the curve proposed by Howell and Bonham (1950). This single
corrected rotational speed (νC), is based on the fact that the curve, together with the curves ψ* = Fψ(φ*, SF), allows the
overall performance of a multistage compressor depends on the obtainment of relationships η* = Fη(φ*, SF) for all types of
performance of its stages. Therefore, by starting from stage compressor stages.
characteristics, which link together pressure coefficient (ψ), The equations (1) to (3) are similar to that used by
stage efficiency (η) and flow coefficient (φ), the stage-by-stage Lakshminarasimha et al. (1994), but fit the data of Muir et al.
evaluation of the outlet conditions from the knowledge of those (1989) and Howell and Bonham (1950) better.
at inlet is possible. At this point the use of a "stage-stacking" Once the generalized stage characteristics are known, the
procedure allows the evaluation of the overall multistage following relationships are used for the stage-by-stage
compressor performance. evaluation of the outlet conditions starting from the inlet ones:
In order to model each compressor stage, generalized
relationships between ψ*=ψ/ψr, η*=η/ηr and φ*=φ/φr were
used. These relationships allow the complete evaluation of stage 1.2

characteristics once the stage reference point (ψr, ηr, φr) is ψ∗

known or estimated. 1.0

The first generalized relationship, ψ* = Fψ(φ*), was set up

0.8 Muir et al. (1989)
based on the work of Muir et al. (1989), who obtained a
generalized curve by fitting experimental data points of a
0.6 SF = 0.0
number of compressor stages. First, the curve of Muir et al.
(1989) was approximated using the following relationship:
0.4 SF = 1.0
ψ *max − 1  * 2
ψ* = ψ *max − ⋅φ − φ * (1)
2  ψ max  0.2
 φ* − 1 SF = -0.5
 ψ max 
0.0
Secondly, in order to account for different stage characteristics 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5 φ∗ 1.6
based on different stage types, a number of generalized
Figure 1 – Generalized stage characteristics ψ* = Fψ(φ*, SF)
ψ* = Fψ(φ*) curves covering all experimental data points and experimental data points (Muir et al., 1989)
reported by Muir et al. (1989) were used. These curves were
obtained by substituting, for each ψ* value calculated by means 1.2
of the equation (1), the respective abscissa φ* with η∗
(φ* + (φ* - 1) SF), where "SF" is the "Shape Factor" introduced 1.0

by Cerri et al. (1993). This allows the representation of different

types of compressor stages, included transonic and supersonic 0.8

stages (negative SF values). In Fig. 1 three curves

ψ* = Fψ(φ*, SF) obtained using SF ∈ [−0.5 , 1] are shown, 0.6

together with the experimental data points and the curve 0.4
reported by Muir et al. (1989). Howell and Bonham (1950)
The second generalized relationship, η* = Fη(ψ*/φ*), was 0.2
set up by substituting the abscissa ζ*/φ* of the generalized stage
efficiency curve η* = Fη(ζ*/φ*) proposed by Howell and 0.0
Bonham (1950) with ψ*/φ* = η* · (ζ*/φ*). The obtained curve 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 ψ∗/φ∗ 1.6

was then approximated using the following relationships: Figure 2 - Generalized stage efficiency curve

3 Copyright © 2002 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 02/01/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 ( )
h T0(i +1)s = h (T0i ) + U i2 ψ i = h (T0i ) + U i2 Fψ (φ i ) (4) • The splitting of the stage enthalpy drop between stator and
rotor based on the degree of reaction of each stage (Ri):
U i2 ψ i U i2 Fψ (φ i ) ∆h rot = Ri ⋅ (∆h 0 )i
( )
h T0(i +1) = h (T0i ) +
ηi
= h (T0i ) +
Fη (φ i )
(5)
∆h stat ≈ (∆h 0 )i − ∆h rot
(9)
(10)
p 0(i +1)
=e
[Φ(T0(i +1)s ) − Φ(T0i )] (6)
• The calculation of the absolute and relative flow angles as a
p 0i function of φ, Ψ and R, on the hypothesis that axial velocity
is constant through the stage.
Ui Ai p 0i T0(i +1) • The mathematical approximation of turbine cascade losses
φ i +1 = φ i (7)
U i +1 A i +1 p 0(i +1) T0i as a function of incidence angle "i" (which, in turn, depends
on flow and stagger angles) by means of the following
By stacking all the "n" stages it is then possible to evaluate: generalized relationships:
p 0(n +1) Yi*min − 1
• the overall pressure ratio β C = Y* = 1 + (i * −1)2 , i* ∈ [i *min , 1] (11)
( )
; 2
p 01 i *min − 1
• the overall enthalpy variation ∆h 0C = h T0(n +1) − h (T01 ) ; ( ) Yi*max − 1
( ) h T0(n +1)s − h (T01 ) Y* = 1 + (i * −1)2 , i* ∈ [1 , i *max ]
(12)
• the compressor efficiency η C =
h (T0(n +1) ) − h (T01 )
, where (
i *max
−1 ) 2

T0(n +1)s is determined by: Φ (T0(n +1)s ) = Φ(T01 ) + ln β C .

• The approximation of the mass flow characteristics of each
turbine cascade by means of non-isentropic converging
nozzle characteristics:
It can be noted that, in addition to the stage reference point p
(ψr, ηr, φr), the parameters which must be known or estimated in M = 2 ⋅ V2 ⋅ A th (13),
RT2
equations (1) to (3) in order to define the generalized
where p2, T2 and V2 are calculated, starting from the
characteristics of each compressor stage are:
physical state 01, pressure ratio (p01/p2) and losses (Y)
 ψ* , φ* , SF, η*(ψ / φ) min , (ψ * / φ*)min , η*(ψ / φ) max , (ψ * / φ*) max 
 max ψ max  across the cascade, by using the following relationships (see
Moreover, the "stage-stacking" procedure requires that the Fig. 3):
= e [Φ (T01 ) − Φ (T2s )]
blade speed Ui at the mean radius and the flow passage area Ai p 01
(14)
be known for each stage. Usually not all the listed parameters p2
are considered as problem variables, since, otherwise, the p 01 − p 02
search procedure of the unknown parameters would be too Y= (15)
p 02 − p 2
difficult. Therefore, some parameters are usually fixed, based
= e [Φ (T01 ) − Φ (T2 )]
on values found in literature or derived from experience, and p 02
(16)
some simplifications are introduced. The choice of parameters p2
to be fixed, their values and the adopted simplifications will be
shown below. V2 = 2(h 01 − h 2 ) , where V2 ≤ kRT2 (17)
Choking conditions are reached when the corrected mass
flow µ1 assumes its maximum value.
TURBINE PERFORMANCE PREDICTION It can be noted that stator and rotor cascades can be
mathematically modeled in the same manner, if absolute and
The procedure to obtain the performance maps of a relative velocities are considered for stator and rotor
multistage turbine, which link together expansion ratio (βT) (or respectively.
the equivalent parameter (∆h0s)T/T0iT), efficiency (ηT), corrected h
mass flow (µT) and corrected rotational speed (νT), is based on 01 02
the following main assumptions: 1
• The splitting of the overall turbine enthalpy drop among
V22
turbine stages based on the aerodynamic loading coefficient 2
(Ψi) of each stage:
U i2 Ψi 2
(∆h 0 )i =
n
⋅ ∆h 0 T (8) 2s

∑ U i2 Ψi
i =1 s

Figure 3 – Nozzle thermodynamic transformation

4 Copyright © 2002 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 02/01/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 The overall performance maps of the multistage turbine are PROCEDURE FOR THE DETERMINATION OF THE
obtained by matching the mass flow characteristics of each PERFORMANCE CURVE PARAMETERS
turbine cascade (Saravanamuttoo and Mirza-Baig, 1993)1. An
example of stator and rotor matching is sketched in Fig. 4. The The objective of the proposed method is the set up of a
stator characteristic is plotted as: Cycle Program which accurately reproduces gas turbine
(A th )stat ⋅ p 01  p  performances, when only a little information is available about
M= ⋅ F  
 (18). the machine. This information usually consists of qualitative
T01  p 01  information about gas turbine component design, and few
For each "M" value it is possible to evaluate the static pressure measurements (such as compressor outlet pressure and
at the stator exit (p2) and, from the velocity triangle, the total temperature, turbine outlet temperature and fuel mass flow rate)
relative physical state at the rotor inlet (T02rel, p02rel). This state at different operating points.
is used to evaluate the rotor mass flow characteristic (which, In order to set up the Cycle Program, it is necessary to
reported in (p/p01, M) coordinates, appears as in Fig. 4) which, evaluate the parameters which define the compressor and
in turn, allows the evaluation of the static pressure at the rotor turbine performance maps. The characteristic of the proposed
exit (p3). method is that the unknown parameters are determined by
In order to evaluate the overall performance maps of the combining the Cycle Program with compressor and turbine
multistage turbine the following parameters must be known or performance maps obtained using the “stage–stacking”
estimated: procedure previously presented, and by searching for the values
− flow coefficient, aerodynamic loading coefficient, degree of of the unknown parameters which better reproduce, by means of
reaction and blade speed at the mean radius of each stage the Cycle Program, the overall performance and thermodynamic
(φ, Ψ, R, U); data measured on the gas turbine. In particular, the unknown
− parameters which define the loss curve of each cascade parameters are determined in an iterative way by using the
 Y , i , Y* , i* , Y* , i*  . following procedure (Fig. 5):
 r r i min min i max max  1. guess values of the parameters are initially assumed;
As previously observed for the compressor, usually not all the 2. the compressor and turbine performance maps obtained by
listed parameters are considered as problem variables. The using the “stage–stacking” procedure are used in the Cycle
choice of parameters to be fixed, their values and the adopted Program to calculate estimates of the performance and
simplifications will be shown below. thermodynamic data measured on the gas turbine;
3. the sum of the square differences between computed and
measured data is evaluated;
Initial X guess values
M
rotor Updat ed
X values
Compressor and Turbine
performance maps

stator
Operating poin t
Cycle Program

Computed data
(Qc )

0 p/p01 p3/p01 p2/p01 p02rel/p01 1

Figure 4 – Stator and rotor matching MINIMIZATION
Measured ALGO RIT HM
1
data (Qm )
Saravanamuttoo and Mirza-Baig (1993) approximate the mass flow min Fob (X 1 , ..., Xnx)
characteristics of multistage turbines by nozzle characteristics, i.e. a single
mass flow curve with a constant turbine efficiency is assumed for each
turbine. These curves are used to match different turbines (such as HPT
and LPT) each other. In this paper instead, the behavior of each turbine
cascade was approximated by nozzle characteristics and using a X1 , ..., Xnx
generalized relationship for cascade losses. The characteristics of stator w hic h minimize F ob
and rotor cascades were then matched to obtain the turbine stage
performance curves, which, in turn, were matched each other to obtain the
performance curves of the multistage turbine. Figure 5 – Search procedure of the unknown parameters

5 Copyright © 2002 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 02/01/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 4. the parameter values are updated; temperatures (i.e. at different corrected rotational speeds).
5. the steps 2 to 4 are repeated until the sum of the square In order to improve the search procedure convergence the
differences between computed and measured data reaches its following simplifications were assumed:
minimum value. − it was considered that all the compressor stages and all the
The above procedure is realized by using a non-linear turbine stages have the same mean radii respectively;
minimization algorithm to solve the problem: − at ISO design conditions it was assumed that all the
1 nop nm  Q c − Q m  
2 compressor stages work at the same reference point ((ψ)r,
min Fob (X 1 ,..., X nx ) = ∑ ∑ w i   (19), (η)r, (φ)r);
nop j=1 i =1  Q m  i  − the same generalized stage performance curves were used
 j
for all the compressor stages;
where Xi are the unknown parameters, Qc and Qm the computed
− the same generalized loss curve was used for all the turbine
and measured data respectively, "nop" the number of operating
cascades.
points and "nm" the number of Qm data for each operating
These simplifications reduce the number of parameters,
point. In order to improve the minimization algorithm
which must be known or estimated to define compressor and
convergence, constraint equations can be added by imposing
turbine performance maps, to the following:
that, in some operating points, computed data be equal to the
corresponding measured data (Bettocchi et al., 1994): − for compressor:
ψr, ηr, φr, the mean blade speed UC common to all stages,
 Qc − Q m 
  = 0 , i = 1,..., nm , j = 1,..., nc (20), ψ*max , φ*ψ max , SF, η*(ψ / φ) min , (ψ * / φ*)min , η*(ψ / φ) max , (ψ * / φ*)max
 Q m  ij
and the flow passage areas Ai (i = 1,…, 18) of all stages;
where nc ≤ nop. − for turbine:
The constrained minimization algorithm which was used is φi, Ψi, Ri (i = 1,…, 3), the blade speed at the mean radius UT
contained in the IMSL math library (Visual Numerics, 1994)
and was successfully utilized by the author to solve in direct and common to all stages, Y , i , Y * , i * , Y * , i * .
r r i min min i max max
inverse mode the gas turbine mathematical model (Benvenuti et From the cross-sectional layout of the gas turbine the mean
al., 1994; Bettocchi et al. 1994; Bettocchi and Spina, 1999; radius and the inlet area of compressor and turbine (AiC, AiT)
Bettocchi et al., 2000, 2001). were estimated. These geometrical data, together with shaft
rotational speed, were used to calculate the blade speed at the
mean radius of compressor and turbine stages (UC, UT).
PROCEDURE APPLICATION Note that:
− for the compressor the stage reference point (ψr, ηr, φr) and
The search procedure for the evaluation of the unknown the flow passage areas Ai (i = 1,…, 18) are a function of the
parameters defining compressor and turbine performance maps inlet area AiC and of mass flow rate, pressure ratio and
was applied to a single shaft industrial gas turbine equipped compressor overall efficiency (ηC)r at ISO design
with a fixed geometry eighteen stage axial flow compressor and conditions;
a three stage turbine. The main gas turbine features are reported − for the turbine the reference values of the flow coefficients
in Table 1. φi (i = 1,…, 3) are a function of turbine inlet area AiT
through the mass balance equation, while the reference
Table 1 - Gas turbine main features (ISO ambient conditions, value of the pressure losses Yr is a function of turbine
natural gas fuel, no input-output pressure losses) overall efficiency (ηT)r and of Ψi, Ri (i = 1,…, 3) values.
Therefore, the parameters which remain to be estimated are:
Compressor mass flow rate [kg/s] 159 − for compressor:
Compressor pressure ratio 11
ψ*max , φ*ψ max , SF, η*(ψ / φ) min , (ψ * / φ*)min , η*(ψ / φ) max , (ψ * / φ*)max ,
Compressor stages 18
Turbine inlet temperature [K] 1371 AiC, (ηC)r;
Turbine inlet pressure [kPa] 1080 − for turbine:
Turbine stages 3
Ψi, Ri (i = 1,…, 3), i r , Yi*min , i *min , Yi*max , i *max , AiT, (ηT)r.
Shaft rotational speed [rpm] 4918
Some of these parameters were fixed based on values found
in literature (Ainley and Mathieson, 1951; Muir et al., 1989),
The available data were the following: while the others were considered as problem unknowns and, so,
− a cross-sectional layout of the gas turbine; determined using the search procedure. In Table 2 the values of
− shaft power, overall gas turbine efficiency (and so fuel mass the fixed parameters, the initial guess values of the unknown
flow rate), pressure and temperature at the compressor outlet parameters and their final values determined using the search
and turbine outlet temperature at some loads and ambient procedure, are reported.

6 Copyright © 2002 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 02/01/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 It can be noted that the compressor inlet area AiC was not Results and discussion
used as problem unknown, but the AiC value estimated from the Figure 6 shows the curves obtained by interpolating the
cross-sectional layout of the gas turbine was used in the available data (overall gas turbine efficiency and fuel mass flow
calculations. This is because AiC does not influence compressor rate (a), pressure and temperature at the compressor outlet (b)
performance maps, since a variation on AiC causes a and turbine outlet temperature (c)) at some loads for an ambient
proportional variation of φr and φ, so that φ* remains the same. temperature of 288.15 K. Similar data for ambient temperatures
It can also be noted that, for the compressor and turbine of 278.15 K and 303.15 K were also available. All these data
efficiencies, the initial guess values and the final ones are quite were used in order to determine the unknown parameters
different from each other, while compressor and turbine defining compressor and turbine performance maps. In the same
efficiencies can usually be estimated with good accuracy from figure the points obtained by running the Cycle Program with
the available data. This was expressly done to test the search both the initial guess compressor and turbine performance maps
procedure robustness when the initial guess value is far from the (black points) and the final ones (white points) are reported.
real solution. These results were obtained using the working point at ISO
ambient condition and two other working points, at a different
Table 2 - Values of the fixed parameters and of the guess and load and ambient temperature, to form the objective function
final values of the unknown parameters (19). In fact, calculations performed to analyze the influence on
the algorithm convergence of the number of points seem to
Fixed Unknown Parameters
Parameters indicate as the best condition the case of three working points.
Parameters Guess Value Final Value This result is confirmed by the analysis performed by Gulati et
al. (2000). In addition, the constraint equations (20) for the only
ψ*max 1.115
working point at ISO ambient condition were used to improve
φ*ψ max 0.835 the minimization algorithm convergence.
Note that the agreement among the points obtained at the
SF 0.300 0.480 end of the search procedure (white points) and the available
curves is very good. However, an improvement of the results
η*(ψ / φ) min 0.200
can be obtained by restarting the search procedure from the
(ψ * / φ*)min 0.040 final values of unknown parameters reported in the last column
of Tab. 2. When this second calculation was performed, the
η*(ψ / φ) max 0.920 three points at the three available ambient temperatures
characterized by the lowest loads were used to form the
(ψ * / φ*)max 1.460
objective function (19). This is because these points are the
(ηC )r 0.850 0.889 ones estimated with the lowest accuracy. Moreover, the
compressor and turbine overall efficiencies were fixed at 0.889
Ψ1 1.300 1.240 and 0.880 respectively, since the computed compressor and
R1 0.200 0.246 turbine outlet temperatures at the design point are close to the
expected ones (see figure 6 (b) and (c)), and this should indicate
Ψ2 1.100 1.380 that the computed compressor and turbine efficiencies are right.
R2 0.500 0.440 This new calculation permitted a reduction of the errors in gas
turbine performance prediction at the lowest loads. The values
Ψ3 1.000 of the normalized Root Mean Square (RMS)2 between
computed and expected values on all the 12 computed operating
R3 0.500
points (4 for each of the three ambient temperatures), for each
i r [°] -7 available gas turbine thermodynamic and performance
parameter and for all the calculations performed are reported in
Yi*min 1.550 Table 3. In particular the expected values of gas turbine
thermodynamic and performance parameters were compared to
i *min [°] - 2.286
those obtained using the initial guess compressor and turbine
Yi*max 1.500 performance maps, the ones at the end of the first calculation
and the ones at the end of the second calculation.
i*max [°] 5.714
2
2 The normalized RMS between computed and expected values on all the
AiT [m ] 0.380 0.444 computed operating points was evaluated using the following relation:
(η T ) r 
0.920 0.880  
2
1 nop  Q c − Q m
RMSi =   
 nop j∑  Q
=1  m
  , i = 1,..., nm , j = 1,..., nop
ij 


7 Copyright © 2002 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 02/01/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Table 3 – Normalized RMS between computed and expected In Tab. 4 the final values of the unknown variable
values on all the computed operating points (12) parameters at the end of the second calculation are reported.
The initial, intermediate (at the end of the first calculation) and
RMS RMS RMS
final (at the end of the second calculation) performance maps
Parameters (start of the (end of 1st (end of 2st
calculation) calculation) calculation) used in the Cycle Program are reported in Figures 7 and 8 for
compressor and turbine respectively.
poC 1.423 x 10-2 7.787 x 10-3 1.049 x 10-4
Table 4 – Final values of the variable parameters
ToC 2.041 x 10-2 1.601 x 10-3 5.480 x 10-4 nd
(2 calculation)
ToT 2.050 x 10-2 1.850 x 10-3 6.883 x 10-5
Mf 2.986 x 10-2 5.807 x 10-3 2.969 x 10-4 SF 0.816
Ψ1 1.297
R1 0.253
Ψ2 1.274
R2 0.438
0.35 5
AiT [m2] 0.431
available data
initial solution
0.30 final solution 4
The analysis of the obtained results highlights the capability
of the proposed method in the set up of a Cycle Program which
0.25 η 3 accurately reproduces data available on a gas turbine, especially
GT
when recursive calculations are performed.
M [kg/s]
However, the high non-linearity of the problem involves
0.20 2
f difficulties in minimization algorithm convergence, in particular
when the number of unknown parameters is high, as, for
(a) example, in the case of variable geometry compressors. Better
0.15 1
1100 650 results may be obtained using a more robust minimization
algorithm, like a genetic algorithm.
p [kPa]
oC
1000
CONCLUSIONS
600
In the paper a method for gas turbine performance
900 prediction, which uses compressor and turbine performance
T [K]
oC maps obtained by using generalized stage performance curves
matched by means of the “stage–stacking” procedure has been
(b) presented. The characteristic of the proposed method is that the
800 550
unknown parameters defining the generalized stage
900
performance curves are determined by combining a Cycle
T [K] Program with the compressor and turbine performance maps
oT
obtained using the “stage–stacking” procedure, and by
800 searching for the values of the unknown parameters which
better reproduce, by means of the Cycle Program, overall
performance and thermodynamic data measured on a gas
700 turbine.
The method was applied and tested using data available on a
single shaft industrial gas turbine. The results of the application
(c) highlight the capability of the method in the set up of a Cycle
600 Program which accurately reproduces the available data,
0 2000 0 40000 60000
P [kW] especially when recursive calculations are performed. The high
GT
non-linearity of the problem involves, however, difficulties in
Figure 6 – Overall gas turbine efficiency, fuel mass flow rate
minimization algorithm convergence, in particular when the
(a), pressure and temperature at the compressor outlet (b) and
turbine outlet temperature (c) versus gas turbine power number of unknown parameters is high. Better results may be
(Ta = 288.15 K) obtained using a more robust minimization algorithm, such as a
genetic algorithm.

8 Copyright © 2002 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 02/01/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 0.90 ACKNOWLEDGMENTS
η
C The work was carried out with the support of the
M.U.R.S.T. (Italian Ministry of University and Scientific &
0.85
0.9 Technological Research).

0.80
ν* = 1.0 REFERENCES

1.1 Ainley, D. G., Mathieson, G. C. R., 1951, "A Method of

0.75
Performance Estimation for Axial Flow Turbines", Aeronautical
Research Council, R&M 2974.
20
Aker, G.F., Saravanamuttoo H. I. H., 1988, "Predicting Gas
βC initial
intermediate Turbine Performance Degradation due to Compressor Fouling
16 final Using Computer Simulation Techniques", ASME Paper
88-GT-206.
Benvenuti, E., Bettocchi, R., Cantore, G., Negri di
12
Montenegro, G., Spina, P. R., 1993, "Gas Turbine Cycle
Modeling Oriented to Component Performance Evaluation from
8 Limited Design or Test Data", Proceedings, 7th ASME COGEN
- TURBO, Bournemouth, UK, IGTI Vol. 8, pp. 327-337.
0.9 ν* = 1.0 1.1 Benvenuti, E., Bettocchi, R., Cantore, G., Negri di
4 Montenegro, G., Spina, P. R., 1994, "Experimental Validation of
0.6 0.8 1.0 1.2 µC* 1.4 a Gas Turbine Cycle Model Based on a Simultaneous Solution
Figure 7 – Compressor initial, intermediate (end of 1 calc.)
st Method", Proceedings, 8th ASME COGEN - TURBO, Portland,
nd
and final (end of the2 calc.) performance maps Oregon, USA, IGTI Vol. 9, pp. 245-255.
Bettocchi, R., Pinelli, M., Spina, P. R., Venturini, M.,
1.00 Sebastanelli, S., 2001, "A System for Health State
determination of Natural Gas Compression Gas Turbines",
ASME Paper 2001-GT-0223.
ηT
Bettocchi, R., Spina, P. R., 1999, "Diagnosis of Gas Turbine
Operating Conditions by Means of the Inverse Cycle
0.80 Calculation", ASME Paper 99-GT-185.
0.9
Bettocchi, R., Spina, P.R., Alliney, S., 1994, "Resolution
Method for Gas Turbine Mathematical Models", Proceedings,
1.0 8th ASME COGEN - TURBO, Portland, Oregon, USA, IGTI
ν* = 1.1 Vol. 9, pp. 361-369.
Bettocchi, R., Spina, P. R., Benvenuti, E., 2000, "Set-Up of
0.60
an Adaptive Method for the Diagnosis of Gas Turbine
1.01
µT* Operating State by Using Test-Bench Measurements", ASME
0.9
1.0 Paper 2000-GT-0309.
1.00
Cerri, G., Salvini, C., Procacci, R., Rispoli, F., 1993,
ν* = 1.1 "Fouling and Air Bleed Extracted Flow Influence on
0.99
Compressor Performance", ASME Paper 93-GT-366
Consonni, S., Macchi, E., 1988, "Gas Turbine Cycles
0.98 Performance Evaluation", Proceedings, 2nd ASME COGEN -
initial TURBO, Montreaux, Switzerland, pp. 67-77.
0.97 intermediate
Doel, D.L., 1994, "TEMPER – A Gas-Path Analysis Tool
final
for Commercial Jet Engines", ASME Journal of Engineering
0.96 for Gas Turbines and Power, Vol. 116, pp. 82-89.
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1
[(∆h ) /T ]* Doyle, M. D., Dixon, S. l., 1962, "The Stacking of
0s T 0iT
Figure 8 – Turbine initial, intermediate (end of 1 calculation)
st Compressor Stage Characteristics to Give an Overall
nd
and final (end of the2 calculation) performance maps Compressor Performance Map", The Aeronautical Quarterly,
pp. 349-367.
El - Masri, M. A., 1988, "GASCAN - an Interactive Code

9 Copyright © 2002 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 02/01/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 for Thermal Analysis of Gas Turbine Systems", ASME Journal Gasdynamics", Canadian Aeronautics and Space Journal, Vol.
of Engineering for Gas Turbines and Power, Vol. 110, pp. 201- 39, No. 1, pp. 16-22.
209. Seddigh, F., Saravanamuttoo, H. I. H., 1990, "A Proposed
Glassman, A. J., 1974, "Computer Program for Method for Assessing the Susceptibility of Axial Compressors
Thermodynamic Analysis of Open-Cycle Multishaft Power to Fouling", ASME Paper 90-GT-348.
System with Multiple Reheat and Intercool", NASA Technical Stamatis, A., Mathioudakis, K., Papailiou, K.D., 1990,
Note NASA TN D-7589. "Adaptive Simulation of Gas Turbine Performance", ASME
Gulati, A., Zedda, M., Singh, R., 2000, "Gas Turbine Engine Journal of Engineering for Gas Turbines and Power, Vol. 112,
and Sensor Multiple Operating Point Analysis Using pp. 168-175.
Optimization Techniques", Proceedings, 36th AIAA/ASME/ Stecco, S. S., Manfrida, G., Galletti, A., 1985, "Gas Turbines
SAE/ASEE Joint Propulsion Conference & Exibit, Huntsville, in Cogeneration: Overall Analysis and Numerical Predictions",
Alabama, USA, July 17 – 19, AIAA 2000-3716. Proceedings, IGTI Beijing International Gas Turbine
Howell, A. R., Calvert, W. J., 1978, "A New Stage Stacking Conference, Beijing, China.
Technque for Axial-Flow Compressor Performance Prediction", Stone, A., 1958, "Effects of Stage Characteristics and
Transactions of the ASME, Vol. 100. Matching on Axial Flow Compressor Performance",
Howell, A. R., Bonham, R. P., 1950, "Overall and Stage Transactions of the ASME, Vol. 80, pp. 1273-1293.
Characteristics of Axial Flow Compressors", Proceedings, Tabakoff, W., Lakshminarasimha, A. N., Pasin, M., 1990,
IMechE, Vol. 163, pp. 235-248. "Simulation of Compressor Performance Deterioration Due to
Kurzke, J., Riegler, C., 2000, "A New Map Scaling Erosion", ASME Journal of Turbomachinery, Vol. 112, pp. 78-
Procedure for Preliminary Conceptional Design of Gas 83.
Turbines", ASME Paper 2000-GT-0006. Thermoflow Inc., 1989, "GTMASTER: Interactive Software
Lakshminarasimha, A. N., Boyce, M. P., Meher-Homji, C. for Design & Performance Analysis of Gas Turbine Power &
B., 1994, " Modeling and Analysis of Gas Turbine Performance Cogeneration Systems", Wayland, MA, USA.
Deterioration", ASME Journal of Engineering for Gas Turbines Thermoflow Inc., 1990, "GTPRO: Interactive Software for
and Power, Vol. 116, pp. 46-52. Gas Turbine Power & Cogeneration System Design", Wayland,
Massardo, A., 1991, "Simulation of Fouled Axial Multistage MA, USA.
Compressors", IMechE Paper C423/048 Tsalavoutas, A., Aretakis, N., Mathioudakis, K., Stamatis,
Muir, D.E., Saravanamuttoo, H.I.H., Marshall, D.J., 1989, A., 2000, "Combining Advanced Data Analysis Methods for the
"Health Monitoring of Variable Geometry Gas Turbines for the Constitution of an Integrated Gas Turbine Condition
Canadian Navy", ASME Journal of Engineering for Gas Monitoring and Diagnostic System", ASME Paper 2000-GT-
Turbines and Power, Vol. 111, pp. 244-250. 0034.
Pinelli, M., Spina, P. R., 2002, "Gas Turbine Field Visual Numerics, Inc., 1994, "IMSL MATH/LIBRARY:
Performance Determination: Sources of Uncertainties", ASME FORTRAN Subroutines for Mathematical Applications",
Journal of Engineering for Gas Turbines and Power, Vol. 124, Houston, Texas, USA.
pp. 155-160. Waters, M., and Associates Inc., 1983, "Gas Turbine
Ping Zhu, Saravanamuttoo, H. I. H., 1992, "Simulation of an Evaluation (GATE) Computer Program. Thermodynamic
Advanced Twin-Spool Industrial Gas Turbine", ASME Journal Cycles, Methods and Sample Programs", EPRI Report AP-
of Engineering for Gas Turbines and Power, Vol. 114, pp. 180- 2871-CCM.
186.
Procacci, R., Rispoli, F., 1995, "Off Design Performance
Evaluation of Deteriorated Variable Geometry Axial Flow
Compressors", ASME Paper 95-CTP-35.
Robbins, W. H., Dugan, J. F., 1965, "Prediction of Off-
Design Performance of Multi-Stage Compressors", NASA SP-
36.
Saravanamuttoo, H. I. H., Lakshminarasimha, A. N., 1985,
"A Preliminary Assessment of Compressor Fouling", ASME
Paper 85-GT-153
Saravanamuttoo, H. I. H., Mac Isaac, B. D., 1983,
"Thermodynamic Models for Pipeline Gas Turbine
Diagnostics", ASME Journal of Engineering for Gas Turbines
and Power, Vol. 105, pp. 875-884.
Saravanamuttoo, H. I. H., Mirza-Baig, F. S., 1993, "Off-
Design Performance Prediction of Turbofans Using

10 Copyright © 2002 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 02/01/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use