Proceedings of the 2014 10th International Pipeline Conference

IPC2014
September 29 - October 3, 2014, Calgary, Alberta, Canada

IPC2014-33015

Comparison of Degradation of two Different Gas Turbine Engines in

Natural Gas Compressor Stations

C. Hartloper, K.K. Botros H. Golshan, D. Rogers, Z. Samoylove

NOVA Research & Technology Center TransCanada Pipelines Ltd.
Calgary, Alberta, Canada Calgary, Alberta, Canada
colin.hartloper@novachem.com
kamal.botros@novachem.com

ABSTRACT NOMENCLATURE

Gas Turbine (GT), like other prime movers, undergoes wear c0-c2 coefficients of objective function F1
and tear over time which results in performance drop as far as F vector of objective functions
available power and efficiency are concerned. In addition to H engine health parameter vector, or air-compressor
routine wear and tear, the engine also undergoes corrosion, enthalpy rise (equation 8)
fouling etc. due to the impurities it breathes in. It is standard k isentropic exponent
procedure to ‘wash’ the engine from time to time to revive it. m measured engine parameter vector
However, it is important to establish a correct schedule for engine-air-compressor air inlet flow rate (kg/s)
the wash to ensure optimal maintenance procedure. This calls N1 engine-air-compressor low shaft speed (rpm)
for accurate prediction of the performance degradation of the P2 engine-air-compressor suction pressure (kPa)
engine over time. In this paper, an error-in-variables based P3 engine-air-compressor discharge pressure (kPa)
methodology is applied to evaluate the performance Pr engine-air-compressor pressure ratio
degradation of two GT engines between soak washes. These R gas constant (kJ/mol-K)
engines are LM2500+ (single spool) and RB211-24G (twin Sh.P shaft output power (kW)
spool). The engine-air-compressor isentropic efficiency and T2 engine-air-compressor suction temperature (K)
air inlet flow rate as well as the engine heat rate and specific T3 engine-air-compressor discharge temperature (K)
work are analyzed for both engines. For both engines, the u external input parameters
compressor isentropic efficiency is found to degrade over V error covariance matrix
time, while the engine heat rate correspondingly increases. y true engine parameters
The compressor air inlet flow rate and engine specific work Z average compressibility factor
remain mostly constant. Through a comparison between the ε random error or noise
time-history of the engine health parameters, it is found that ηis engine-air-compressor isentropic efficiency
the LM2500+ degrades at a much faster rate than the RB211- θ vector of fitting parameters
24G. However, the degradation of the LM2500+ is found to σ standard deviation in measured parameters
be fully recoverable by offline washes, while the degradation Ψ compressor head coefficient
of the RB211-24G is only slightly recovered by offline
washes. The RB211-24G engine is found to be running near
its maximum efficiency at all times, which is likely the cause
for the observed non-recoverable degradation that the engine
experiences. The engine’s site location is also found to
contribute to the degradation that the engine experiences.

1 Copyright © 2014 by ASME

INTRODUCTION
Degradation in gas turbines (GT) is of great importance to
industrial users for engine performance reliability and
operating costs. Degradation refers to a drop in the
performance of the GT engine indicated by a loss of available
power and increased heat rate (or reduced thermal
efficiency). In essence, degradation of the overall engine is a
combined effect of the degradation contribution from each of
its components, although each component of the engine
degrades at a different rate [1]. Significant development has
been made in determining the mechanisms of engine
degradation [1-3]. Fouling, corrosion, and impingement of
foreign objects on both the compressor and turbine blades are
some of the mechanisms that cause engine degradation [1].
These factors contribute towards reduced tip clearance,
change in blade geometry and surface roughness [2,3], which
are manifested in a reduction in shaft power output and
increase in the overall heat rate of the engine.
From a maintenance and operational standpoint, engine wash
is a standard procedure to both prevent the engine from
permanent damage and aid in performance recovery [3].
Correct scheduling of the engine wash is extremely important
to ensure optimal operation and maintenance of the engine
[4], and this is only possible if the degradation behavior of
the engine can be accurately tracked over time. It is also
important to establish the effectiveness of each wash and its
corresponding performance recovery during the operational
period.
There are two main objectives of the present paper. The first
objective is to extend the current methodology of degradation
analysis, which was previously applied to a single-spool GE
LM2500+ engine [5], to a twin-spool RR RB211-24G engine
in order to quantify the degree of engine-air-compressor
degradation. The engine-air-compressor degradation is
tracked through changes in both the isentropic efficiency as
well as the air inlet flow rate. The second objective is to
contrast the effect of both engine-air-compressor degradation
and offline soak washes on the overall performance of the
two engines, where the overall engine performance is tracked
by the engine heat rate as well as the engine specific work.
The operational data on which the analysis is performed was
acquired between April 2009 and January 2013 for the
LM2500+ engine, and between March 2010 and April 2013
for the RB211-24G engine.
The following sections are organized as follows. First, the
techniques used to analyze both the operational data and the
engine health parameters are introduced. Next, the results for
the RB211-24G engine are presented, followed by the results
for the LM2500+ engine. Finally, the results from the two
engines are compared and conclusions on the performance
degradation of GT engines are drawn.
2 Copyright © 2014 by ASME
DEGRADATION PREDICTION MODEL
Gas Path Analysis (GPA) and an Error-in-Variables Model
(EVM) were used to track the engine degradation over time
in a manner similar to that described by Botros et al. [5].
These methods will be described in the following
subsections.
Gas Path Analysis
GPA refers to fundamental thermodynamic calculations to
evaluate the properties (such as pressures, temperatures, fuel
flow rate, etc.) of the air and the products of combustion as
they pass through the various components of the engine [6].
For the GPA, three types of parameters are defined, namely:
1. Engine Parameters (y), such as pressure, temperature,
fuel flow rate, shaft power etc. However, what is read by
the sensors at the plant are the Measured Engine
Parameters (m), which are the true engine parameters (y)
distorted with noise (ε) due to instrumentation
inaccuracies. (m = y + ε )
2. External Input Parameters (u), such as properties of the
gas path, Lower Heating Value (LHV) of the fuel, spool
speeds, mechanical efficiencies of the components, etc.
3. Health Parameters (H), which are the indicators that
define the health condition of the engine, and hence the
degree of degradation of its components. The air
compressor health parameters are the isentropic
efficiency and inlet air mass flow rate, while the engine
health parameters are the heat rate and the specific work.
These health parameters are calculated from the
measured parameters (m) for a given set of input
parameters (u). In order to track the health parameters as
the inlet conditions vary, the parameters must be
corrected to standard inlet conditions. The formulae for
the corrected shaft speed (Ncorrected), mass flow ( corrected),
shaft power (Sh.Pcorrected) and specific work (Sp.Wcorrected)
are given below as [7]:
(1)
√
̇√
̇ (2)
(3)
√
 (4)
̇ ̇
where θ and δ are the dimensionless ambient temperature and
pressure and have values of θ=T 2/288.15 and δ=P2/101.325.
Figure 1: Schematic representation of a GT engine, with
engine parameters and health parameters identified.
It is recognized that the noise in the measured parameters
will propagate to the calculated health parameters, resulting
in a noisy scatter (smearing) of the health parameter values.
A summary of the Engine Parameters (y) and the Health
Parameters (H) is schematically depicted in Figure 1, where
mf and Sh.P represent the fuel flow rate and the shaft power
respectively. The other symbols have their usual meanings.
The Gas Path Analysis was performed using the commercial
software GasTurb11 [8,9]. The ambient temperature,
pressure, measured compressor exit pressure and
temperature, temperature at the combustor exit and the spool
speed are specified as inputs. Although the RB211-24G is a
twin-spool engine, there is aerodynamic coupling between
the shaft speeds of the two shafts [10] and therefore the
health parameters for this engine are calculated assuming a
single-spool arrangement. Mechanical efficiencies of the
components are also fed in as fixed external input parameters
along with LHV of the fuel. The procedure completes the
thermodynamic cycle of the engine operation and calculates
the health parameters. However, these health parameters, as
calculated by GasTurb11, are also contaminated with noise
that has propagated from the measured parameters.
Therefore, additional steps must be taken before data analysis
is possible.
3 Copyright © 2014 by ASME
Parameter Estimation – Bayesian Error-in-Variables
Model (EVM)
EVM is commonly used in data modeling for obtaining point
estimates of unknown parameters where all the involved
variables have inherent uncertainty in them [11]. EVM finds
direct relevance to the problem in hand in which all the
measured variables and the health parameters from the GPA
model are subject to varying degrees of inaccuracies.
Ordinary Least Squares method would not be suitable for the
present problem as it only corrects the dependent variable
assuming the other variables to be error free [12]. Also,
unlike in Ordinary Least Squares, EVM models are
dependent on prior knowledge of the error structure in each
and the vector of coefficient parameters used in the objective
functions as of the variables involved [11] allowing more
accurate estimates. Detailed discussions on EVM can be
found in Keeler and Reilly [11] and Seber and Wild [13].
The EVM analysis involves defining the vector of variables
involved, which must be conditioned off the noise, as:
̅
(5)
( )
where P2 and P3 are the compressor suction and discharge
pressure, T2 and T3 are the compressor suction and discharge
temperature, N1,corrected is the corrected compressor shaft
speed and ηis is the compressor isentropic efficiency.
Similarly vector Ytrue contains the true values of the variables
in Y in the same order. Next, a vector of objective functions
is defined as:
[ ] (6)
[( ⁄ ) ]
(7)
where Ψ is the compressor head coefficient [10] and is
defined as:
(8)
[( ) ]
where H, Z, R and k are the air-compressor enthalpy rise,
average compressibility factor, gas constant and isentropic
exponent, respectively. The vector of the coefficient
parameters used in the objective function is:
̅ ( )
(9)
In addition to these three coefficients, the model also re uires
defining the error covariance matri (V). n the error
covariance matrix, the diagonal elements are the average
uncertainty of each variable involved while the non-diagonal
elements are the error of one variable with respect to the
uncertainty in the others. For the present functions model, the
covariance matrix is selected on the basis of the expected
uncertainty associated with the measurement of the
individual variables, implying that the errors are uncorrelated
(i.e., “zeros” in the non-diagonal elements) [14]. Thus:
̅
(10)
( )
Given the range of values for the variables in question as well
as the instrumentation used, the average standard deviations
were chosen to be:
Further details of the EVM method are identical to those
given by Botros et al. [5].
RESULTS FOR RR RB211-24G
Figure 2 shows the timeline of operation for which engine
performance data were collected on the RR RB211-24G
engine employed at a compressor station in Alberta, referred
to as Station 1. Data acquisition began on March 1, 2010.
This engine was then replaced with a new engine on August
18, 2011. An offline wash was performed on the new engine
on June 14, 2012. The Intermediate Pressure (IP) stator was
4 Copyright © 2014 by ASME
replaced on April 23, 2013, and another offline wash was
performed on August 19, 2013. Subsequently, data continued
to be collected until October 25, 2013. To track the effect the
engine and IP stator replacements as well as the offline
washes, the timeline has been separated into five distinct
periods. Period A spans from the beginning of data analysis
(March 1, 2010) until the engine is replaced on August 18,
2011. Period B1 spans from the installation of the new
engine until the first offline wash on June 14, 2012, and
Period B2 spans from the first offline wash until the IP stator
replacement on April 23, 2013. Period C1 spans from the IP
stator replacement until the second offline wash on August
23, 2013, and period C2 spans from the second offline wash
until the end of data acquisition on October 25, 2013. To
allow for the tracking of engine performance throughout the
periods, each period is further divided into equally spaced
sets of 500 operating hours. Periods A, B1, B2 C1 and C2
have six, five, seven, three and two data sets, respectively, for
a total of 23 data sets. The naming convention for the data
sets from this engine is illustrated by the following example:
A_500_3 is the third set of 500 operating hours from period
A. Figure 2 also shows the total operating hours for each of
the five periods.
Figure 2: Timeline of data acquisition for the RR RB211-
24G engine.
In the following paragraphs, the performance of the engine
air compressor is examined and compared to the overall
engine performance. First, the degradation and recovery of
the engine air compressor isentropic efficiency is examined
and related to the overall engine heat rate. It is expected that
a decrease in the isentropic efficiency of the compressor, due
to degradation, will lead to an increase in the overall engine
heat rate. Second, the degradation of engine air compressor
inlet flow rate is examined and related to the overall engine
specific work. In the following sections, comparisons will be
drawn between four pairs of two data sets. The first pair is
the last data set from period A (A_500_7) and the first data
set from period B1 (B1_500_1). Comparison between these
two data sets will highlight the impact of the engine
replacement. The second pair is the last data set from period
B2 (B2_500_7) and the first data set from period C1
(C1_500_1). Comparison between these two data sets will
highlight the impact of the IP stator replacement. The third
pair is the last data set from period B1 (B1_500_5) and the
first data set from period B2 (B2_500_1), while the fourth
pair is the last data set from period C1 (C1_500_3) and the
first data set from period C2 (C2_500_1). Comparison
between these last two pairs of data sets will signify the
impact of the offline washes on the newly installed engine.

Contour maps of N1,corrected vs. pressure ratio (Pr=P3/P2)

colour-coded by engine-air-compressor isentropic efficiency
are used to track the compressor degradation between data
sets. The data sets directly before and after both the engine
replacement and IP stator replacement are presented in
Figure 3 and Figure 4, respectively. The data sets directly
before and after the first and second offline washes are
presented in Figure 5 and Figure 6, respectively. In Figure 3
it is evident that the engine replacement significantly
improves the isentropic efficiency, while in Figure 4 it is
observed that the IP stator replacement only slightly
improves the isentropic efficiency. In Figure 5 and Figure 6 it
can be observed that both the first and second offline washes
have little effect on improving the isentropic efficiency. A
trend noticed in all of the contour maps is the very narrow
relationship observed between Pr and N1,corrected. This trend is
further exemplified in Figure 7, where it can be seen that the
N1,corrected vs. Pr curves collapse onto a single line for all data Figure 3: Contour map demonstrating the significant
sets from all three periods. From the typical compressor improvement in engine-air-compressor isentropic
characteristic map shown in Figure 8, it can be discerned that efficiency after the replacement of the RB211-24G engine.
along the maximum efficiency line there is only one
N1,corrected for a given Pr. Therefore, the tight relationship
between Pr and N1,corrected for this engine air compressor
implies that the engine is being run very close to its
maximum efficiency. Additionally, from this tight
relationship it is evident that when choosing a value for
N1,corrected implicitly chooses a value for Pr as well.

Figure 4: Slight improvement in engine-air-compressor

isentropic efficiency after the replacement of the RB211-
24G IP stator.

5 Copyright © 2014 by ASME

Figure 5: Minimal improvement in engine-air-compressor
isentropic efficiency after the first offline wash of the
RB211-24G.

Figure 7: Scatter plot exhibiting the linear relationship

between pressure ratio and N1,corrected during all five
periods identified in Figure 2 for the RB211-24G.

Figure 8: Typical compressor characteristic map [15].

Figure 6: Minimal improvement in engine-air-compressor
isentropic efficiency after the second offline wash of the
RB211-24G.

6 Copyright © 2014 by ASME

To quantitatively measure the degradation during periods as
well as the improvement (or lack of improvement) between
periods, the isentropic efficiency is tracked over time for
constant N1,corrected and Pr. Figure 9 shows the time-history of
engine air compressor isentropic efficiency for selected
N1,corrected of 6426, 6351 and 6276 RPM, where the following
general trends can be observed. In Period A, the compressor
isentropic efficiency degrades from the beginning of the data
acquisition until the engine replacement, after which the
isentropic efficiency improves by approximately 2
percentage points. In Period B1, the isentropic efficiency
begins to degrade immediately, and continues to degrade at a
constant rate through to the end of Period B2, decreasing by
2.5 percentage points in the 12,145 total operating hours of
B1 and B2. It is important to note that the first offline wash
only slightly improves the isentropic efficiency, with an
increase of 0.5 percentage points. The IP replacement
between periods B2 and C1 improves the isentropic
efficiency by approximately 0.5 percentage points. The
isentropic efficiency remains relatively constant throughout
the 3,594 total operating hours of C1 and C2, showing
negligible improvement after the second offline wash. Note
that the above-mentioned trends are consistent across all
three N1,corrected.

Figure 10 and Figure 11 show shaft power vs. power turbine

speed coloured by heat rate for the engine replacement and IP
stator replacement, while Figure 12 and Figure 13 show the
same for the first and second offline washes. For exemplary
purposes, the mean heat rate for each data set is calculated
for all data within a range of shaft powers. As expected, the
increase in compressor isentropic efficiency due to the engine
replacement, as observed in Figure 9, is reflected in a slight
decrease in engine heat rate, with the heat rate decreasing
0.4% due to the engine replacement.

Figure 9: Engine-air-compressor isentropic efficiency of

RB211-24G vs. time for N1,corrected=6426RPM, 6351RPM
and 6276RPM, from top to bottom.

7 Copyright © 2014 by ASME

 Figure 10: Coloured scatter plots illustrating the slight
decrease in heat rate after the replacement of the RB211-
24G engine. Average heat rate is calculated for all data in
the set with corrected shaft power between 22,500 and
23,500.
Figure 12: Significant decrease in heat rate after the first
engine wash of the RB211-24G. Average heat rate is
calculated for all data in the set with corrected shaft
power between 19,000 and 22,000.

Figure 11: Slight decrease in heat rate after the RB211- Figure 13: Significant decrease in heat rate after the
24G IP stator replacement. second engine wash of the RB211-24G.

8 Copyright © 2014 by ASME

The increase in compressor isentropic efficiency due to the IP
stator replacement also results in a 0.4% decrease in heat
rate. Therefore, it can be concluded that improving the air
compressor isentropic efficiency via either an engine
replacement or an IP stator replacement resulted in a decrease
in engine heat rate, and thus an improvement in the overall
engine thermal efficiency. The first and second offline
washes both resulted in engine heat rate decreases of 1.1%.
Therefore, it appears that both engine washes have secondary
effects that improve the overall engine performance, rather
than just cleaning the air compressor.

Contour maps of N1,corrected vs. Pr colored by corrected engine-

air-compressor air inlet flow rate ( corrected) are used to track
the compressor degradation between data sets. The data sets
directly before and after both the engine replacement and IP
stator replacement are presented in Figure 14 and Figure 15,
respectively. The data sets directly before and after both the
first and second wash are presented in Figure 16 and Figure
17, respectively. ere it can be seen that there is no
significant change in corrected from the engine replacement,
the IP stator replacement or the offline washes. The time-
history of corrected at three specific N1,corrected (as an example)
presented in Figure 18 confirms that there is no noticeable Figure 14: Contour map demonstrating that, unlike the
change in compressor air inlet flow rate (i.e., engine engine-air-compressor isentropic efficiency, the engine-
breathing) throughout the data acquisition. This constant air-compressor air inlet flow rate of the RB211-24G
corrected is reflected in the engine specific work, which, as engine is unaffected by the engine replacement.
shown in Figure 19 through Figure 22, shows little variation
due to the engine replacement, IP replacement or offline
washes.

Figure 15: Similar to the engine replacement, the RB211-

24G IP stator replacement has little effect on the
compressor air inlet flow rate.

9 Copyright © 2014 by ASME

 Figure 16: The first offline wash has little effect on the
compressor air inlet flow rate of the RB211-24G.

Figure 18: Engine-air-compressor corrected air inlet flow

rate of RB211-24G vs. time for N1,corrected=6426RPM,
6351RPM and 6276RPM, from top to bottom.

Figure 17: The second offline wash shows similar effect as

the first on the air inlet flow rate of the RB211-24G.

10 Copyright © 2014 by ASME

Figure 19: Minimal effect on the engine specific work Figure 21: The first offline wash has minimal effect on the
from the replacement of the RB211-24G engine. RB211-24G specific work.

Figure 20: Minimal effect on the engine specific work Figure 22: Similar to the first wash, the second wash has
from the replacement of the RB211-24G IP stator. minimal effect on the RB211-24G specific work.

11 Copyright © 2014 by ASME

RESULTS FOR GE LM2500+
Figure 23 and Figure 24 show the timeline of operation of the
GE LM2500+ engine at a compressor station in
Saskatchewan, referred to as Station 2. The engine was
installed after a hot section repair on April 8th, 2009 and has
been running since then. Data was available until January
14th, 2013 and four offline washes have been performed
during this time. To track the engine performance, the data
were separated into distinct periods. Each period is separated
by an offline wash except in the case of periods 4 and 5,
which together span between a set of offline washes.
Furthermore, data sets consisting of the first, middle, and last
500 hours of each period are selected to track the engine
performance both throughout each period as well as between
periods. The data spanning the timeline shown in Figure 23
has already been analyzed and is reported in Botros et al. [5],
while the data spanning the timeline shown in Figure 24,
containing two additional offline washes, is new to this study.
The naming convention for the data sets from this engine is
illustrated by the following example: 3-L500 is the last 500
operating hours from period 3 while 4-F500 is the first 500
operating hours from period 4. The total operating hours of
each period are indicated under each period in the respective
figure.
Figure 23: Timeline of data acquisition for the GE
LM2500+ for the data that was originally presented in
Botros et al. [5].
Figure 24: Timeline of data acquisition for the new data
for the GE LM2500+.
12
Similar to the analysis of the RB211-24G engine, in the
following paragraphs the performance of the LM2500+
engine air compressor is examined and compared to the
overall engine performance. First, the degradation and
recovery of the engine-air-compressor isentropic efficiency is
examined and related to the overall engine heat rate.
Subsequently, the degradation of engine air compressor inlet
flow rate is examined and related to the overall engine
specific work. In the following sections, comparisons will be
drawn between two pairs of two data sets. The first pair is the
last data set from period 3 (3-L500) and the first data set
from period 4 (4-F500). Comparison between these two data
sets will highlight the impact of the third offline wash. The
second pair is the last data set from period 5 (5-L500) and the
first data set from period 6 (6-F500). Comparison between
these two data sets will highlight the impact of the fourth
offline wash.
Similar to the case of the RB211-24G engine, contour maps
of N1,corrected vs. Pr colored by engine air compressor
isentropic efficiency are used to track the compressor
degradation of the LM2500+ engine between data sets. The
data sets directly before and after both the third and fourth
offline wash are presented in Figure 25 and Figure 26,
respectively. Both the third and fourth washes resulted in an
improvement in the engine air compressor isentropic
efficiency, although the improvement is more significant
after the third wash. This improvement is in contrast to the
case of the RB211-24G engine, where a small increase in
isentropic efficiency was observed after the first offline wash.
Also in contrast to the RB211-24G engine is the observation
that there is much more scatter about the N1,corrected vs.
pressure ratio curve, as is highlighted in Figure 27. Referring
back to Figure 8, this higher degree of scatter implies that the
LM2500+ engine air compressor is not operating at the
maximum isentropic efficiency.
Copyright © 2014 by ASME

Figure 25: Contour map demonstrating the significant
improvement in engine-air-compressor isentropic
efficiency after the third soak wash of the LM2500+.
Figure 26: Similar to the third wash, the fourth wash
significantly improves the isentropic efficiency of the
LM2500+ engine air compressor.
13
Figure 27: Unlike the RB211-24G, the operating
conditions of the LM2500+ show significant variance.
Figure 28 shows the time-history of compressor isentropic
efficiency for N1,corrected of 9669, 9594 and 9519 RPM, where
the following general trends can be observed. Between period
3 and 4, after the third offline wash, the isentropic efficiency
increases by 2.5 percentage points. Across periods 4 and 5,
the efficiency then degrades by 1 percentage point, and
subsequently recovers 1 percentage point from the fourth
offline wash. Then, the isentropic efficiency drops sharply,
decreasing 1 percentage between the fourth offline wash and
the end of data acquisition, a time span of merely 1,200
operating hours. Note that all of the above mentioned trends
are consistent across all three N1,corrected.
Figure 29 and Figure 30 show shaft power vs. turbine speed
color-coded by heat rate for the third and fourth offline wash,
respectively. Here, the mean heat rate is calculated for
exemplary purposes for all data with shaft powers ranging
from 24500 – 25500 kW. From both of the offline washes,
the increase in air compressor isentropic efficiency is
reflected in a decrease in engine heat rate. The engine heat
rate decreased 2.6% and 3.0% from the third and fourth
offline washes, respectively. Thus, improving the air
compressor isentropic efficiency via an offline wash resulted
in an improvement in the overall engine thermal efficiency.
Note that the improvement in overall engine thermal
efficiency as a result of the offline washes was much more
significant for the LM2500+ than it was for the RB211-24G.
Copyright © 2014 by ASME
 Figure 29: Coloured scatter plots illustrate the significant
decrease in heat rate after the third offline wash of the
LM2500+. Average heat rate is calculated for all data in
the set with corrected shaft power between 24500 and
25500.

Figure 28: Engine-air-compressor isentropic efficiency of

LM2500+ vs. time for N1,corrected=9669RPM, 9594RPM
and 9519RPM, from top to bottom.

Figure 30: Significant decrease in heat rate after the

fourth offline wash of the LM2500+.

14 Copyright © 2014 by ASME

Contour maps of corrected N1,corrected vs. Pr colored by
corrected are used to track the compressor degradation
between data sets. The data sets directly before and after the
third and fourth offline wash are presented in Figure 31 and
Figure 32, respectively. Here it can be seen that, similar to
the case of the RB211-24 engine, there is no significant
change in corrected from both of the offline washes. The time-
history of air compressor inlet flow rate presented in igure
confirms that there is no noticeable degradation or
improvement in corrected throughout the data ac uisition. As
e pected, this constant corrected is reflected by a time-
independent scatter in specific work, as observed in Figure
34 and Figure 35.

Figure 31: Contour map demonstrating the minimal

effect of the third wash on engine-air-compressor air inlet
flow rate for the LM2500+.

Figure 32: Similar to the third offline wash, the fourth

offline wash has minimal effect on the air inlet flow rate
of the LM2500+.

15 Copyright © 2014 by ASME

 Figure 34: The third wash has minimal effect on the
specific work of the LM2500+.

Figure 33: Air inlet flow rate of LM2500+ vs. time for
N1,corrected=9669RPM, 9594RPM and 9519RPM, from top
to bottom.

Figure 35: The fourth wash also has minimal effect on the
specific work of the LM2500+.

16 Copyright © 2014 by ASME

DISCUSSION
From the data for the RB211-24G engine, it is clear that this
engine is base-loaded and therefore its air compressor is
being operated on a tight range along the maximum
compressor efficiency line. This result contrasts the results
for the LM2500+, where the engine is subject to variable
loading and therefore its air compressor is operating in a
wider parameter space. This contrast in compressor
operation may be due to the higher demand placed on Station
1 compared to Station 2.
Additionally, Station 1 is located in a forested region close to
the mountains in the province of Alberta, where the air
quality is very good and air-borne particulates will be at a
minimum. This contrasts Station 2, which is located on the
prairies, where farming activity can potentially stir up dust in
the air that could subsequently be breathed in by the engine
air compressor.
For the RB211-24G engine, there was a 2 percentage point
improvement in engine air compressor isentropic efficiency
(ηis) from the engine replacement. There were 4,984
operating hours (OH) between the engine replacement and
the first offline wash, followed by 7,161 OH until the IP
stator replacement. During this time, a total of 12,145 OH,
ηis degraded at a modest rate, dropping approximately 2
percentage points with only slight improvement after the first
offline wash. The IP stator replacement resulted in a 0.5
percentage point improvement in ηis. There were 3,594 OH
between the IP stator replacement and the end of data
acquisition, including a second offline wash. During this time
ηis remained constant. The lack of significant improvement
after both the first and second offline washes suggests that
the degradation of the RB211-24G engine air compressor is
unrecoverable. For the LM2500+ engine, an offline wash
was conducted every 3000 – 4000 OH. From the results
contained in the present paper, as well as previous papers [5],
it is apparent that between each offline wash ηis degraded
approximately three to four times faster than it did for the
RB211-24G engine. However, this degradation was found to
be fully recoverable after each offline wash.
The rate of recoverable and unrecoverable degradation for
the two engines is found to be correlated to the engine
location and operating mode, respectively. The RB211-24G
is located in a forested area, where the air quality is good.
The RB211-24G therefore is not subject to recoverable
degradation since air-borne dust is not breathed into the
engine air compressor. However, since the RB211-24G is
constantly running near capacity (i.e., base-loaded), the
engine parts are likely to wear out. The loss in efficiency due
to the part wear is likely unrecoverable. The LM2500+, on
the other hand, is located in the prairies, where the air quality
may be hampered by farming activities in the surrounding
areas. The dust that is stirred into the air from farming
activities potentially increases the rate of degradation in the
17
LM2500+ engine air compressor. However, as this
degradation is caused by air-borne dust, it appears to be fully
recoverable by offline washes. Additionally, since the
LM2500+ is subject to a loading cycle with periods of
decreased loading, the engine parts do not experience as
much wear and tear. Therefore, unrecoverable degradation
seems minimal.
An alternate explanation for the type of degradation that each
engine experiences is based on the operating hours between
offline washes. For the RB211-24G engine, there are more
operating hours between engine washes. If there is air-borne
dust breathed into the engine air compressor, operating the
engine for too long between offline washes may result in
prolonged contact between the dust and engine parts, thus
resulting in the wearing down of engine parts (unrecoverable
degradation). However, if there are less operating hours
between offline washes, as is the case for the LM2500+, then
air-borne dust will be washed before it has a chance to wear
down the engine parts, thus avoiding unrecoverable
degradation.
or both engines, the degradation in ηis was reflected in
changes to the engine heat rate. Specifically, improvements
in ηis from the engine replacement and offline washes of the
RB211-24G engine and offline washes of the LM2500+
resulted in a decrease in heat rate and thus an improvement in
overall engine thermal efficiency. However, for the
LM2500+ engine ηis was found to have a stronger influence
on heat rate, while only a slight influence was displayed for
the RB211-24G engine. Both the RB211-24G and LM2500+
did not show any degradation in terms of their air breathing
capability, as indicated by the consistent relationship between
corrected specific work and pressure ratio throughout the
data acquisition for both engines.
CONCLUDING REMARKS
1. The RB211-24G engine installed at Station 1 is
operated on a tight range along the maximum
compressor efficiency line, while the LM2500+ engine
installed at Station 2 is operated over a wider range.
2. The RB211-24G engine degrades at a modest rate, with
the engine-air-compressor isentropic efficiency
decreasing approximately 2 percentage points over
12,145 OH. The LM2500+ engine degrades at a much
more rapid rate, with a maximum degradation in
engine-air-compressor isentropic efficiency of 1
percentage point over 1,200 OH between offline wash 4
and the end of data acquisition.
3. The degradation of the RB211-24G engine is not fully
recoverable by offline washes, while the degradation of
the LM2500+ engine is recoverable.
4. For both engines, the degradation in engine-air-
compressor isentropic efficiency is reflected by an
increase in engine heat rate. Changes in the compressor
Copyright © 2014 by ASME
 isentropic efficiency more strongly influence the heat
rate of the LM2500+ engine, while they more weakly
influence the heat rate of the RB211-24G engine.
5. Both engines do not show any sign of degradation in
terms of the air breathing capability or the corrected
specific work.
ACKNOWLEDGEMENTS
27- 4, 1991.
12. P. M. Reilly and . Patino-Leal, "A bayesian study of the
error-in-variables model," Technometrics, vol. 2 , no. , pp.
221-2 1, 1981.
13. . A. . Seber and C. J. Wild, Nonlinear Regression, New
York: John Wiley and Sons, 1989.
14. P. Valko and S. Vajda, "An e tended mar uardt-type
procedure for fitting error-in-variables models," Computer
Chem. Eng., vol. 11, no. 1, pp. 7-4 , 1987.
15. N. A. Cumpsty, Compressor Aerodynamics, John Wiley &
Sones, 1989.

The authors wish to acknowledge the discussion and

technical dialogue with Ken Meszaros, Landen Stein and
Curtis Phinney on the operation of the two engines discussed
in this paper. The support of TCPL Technology
Management; Thomas Robinson, Anthony Tse and Tracy
Cairns during the course of this work is greatly appreciated.
This paper is part of a research program sponsored by
TransCanada Pipelines Ltd., and permission to publish it is
gratefully acknowledged.

REFERENCES

1. R. Kurz and K. Brun, "Degradation in gas turbine systems," J.

Eng. as Turbines Power - transactions of the ASME, vol.
12 , no. 1, pp. 70-77, 2001.
2. R. Kurz, K. Brun and M. Wollie, "Degradation Effects on
ndustrial as Turbines," J. Eng. as Turbines Power -
transactions of the ASME, vol. 1 1, no. 6, 2009.
3. M. Morini, M. Pinelli, P. R. Spina and M. Venturini,
"Numerical Analysis of the Effects of Nonuniform Surface
Roughness on Compressor Stage Performance," J. Eng. as
Turbines Power - transactions of the ASME, vol. 1 , no. 7,
2011.
4. M. ast, M. Assadi and S. De, "Condition based maintenance
of gas turbines using simulation data and artificial neural
network: A demonstration of feasibility," in Proceedings of
the ASME Turbo E po 2008, Berlin, ermany, 2008.
5. K. K. Botros, . oldshan and D. Rogers, "Effects of Engine
Wash re uency on T Degradation in Natural as
Compressor Stations," in Proceedings of ASME Turbo E po
201 , San Antonio, USA, 201 .
6. Y. Li, " as Turbine Performance and ealth Status
Estimation Using Adaptive as Path Analysis," J. Eng. as
Turbines Power - transactions of the ASME, vol. 1 2, no. 4,
2010.
7. A. J. Volponi, " as Turbine Parameter Corrections," J. Eng.
as Turbines Power, vol. 121, pp. 61 -621, 1999.
8. J. Kurzke, "Effects of inlet flow distortion on the performance
of aircraft gas turbines," J. Eng. as Turbines Power -
transactions of the ASME, vol. 1 0, no. 4, 2008.
9. J. Kurzke, "About simplifications in gas turbine performance
calculations," in Proceedings of the ASME Turbo E po 2007,
Montreal, Canada, 2007.
10. . . . Saravanamutto, . Rogers, . Cohen and P. V.
Straznicky, as Turbine Theory, 6th ed., Pearson Prentice
all, 2009.
11. S. E. Keeler and P. M. Reilly, "The error-in-variables model
applied to parameter-estimation when the error covariance
matri is unknown," Can. J. Chem. Eng., vol. 69, no. 1, pp.

18 Copyright © 2014 by ASME