Proceedings of ASME Turbo Expo 2004
ASME Turbo Expo: Power
DETECTION AND LOCALIZATION OF FOULING IN A GAS TURBINE
COMPRESSOR FROM AEROTHERMODYANMIC MEASUREMENTS
Knox T. Millsaps
Director, Marine Propulsion Laboratory
Mechanical and Aerospace Engineering
Naval Postgraduate School
Monterey, CA 93943 USA
millsaps@nps.navy.mil
(831) 656-3382
Jeffrey S. Patterson
Head, Emerging Technologies
Naval Surface Warfare Center,
Carderock Division
Philadelphia, PA
ABSTRACT
This paper presents a new method for condition assessment of
axial flow compressors that provides a tool for specifying the
magnitude and location of degradation due to fouling. A
simple, meanline, stage-stacking analysis is developed, which
includes the impact of blade roughness on the mass flow, work
coefficient, and efficiency. The performance of a baseline,
three-stage compressor with hydrodynamically smooth blades
is calculated. Using the baseline geometry, the influence of
roughness of the blade surfaces in the front, middle and rear
stages are calculated. Empirical data for the increased total
pressure losses and greater turning deviation that occurs due to
rough blades are used. This analysis indicates that airfoil
fouling in different stages, produce characteristic
aerothermodynamic signatures, and hence the faults can be
localized by the magnitudes of the various influence
coefficients. This analysis also predicts that the most sensitive
parameter for predicting fouling in the front stages is the
percentage change in mass flow and the most sensitive
parameter for predicting fouling in the rear stages is the
adiabatic efficiency.
1 1
Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 02/02/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Powerfor
forLand,
Land,Sea,
Sea,and
andAir
Air
June14-19,
June 14-17,2004
2004,Vienna,
Vienna,Austria
Austria
GT2004-54173
GT2004-54173
LT Jon Baker
United States Coast Guard
Headquarters, Naval Engineering
Washington D.C.
INTRODUCTION
The power, output, and fuel consumption of gas turbines are
sensitive to degradation in compressor performance. A
compressor’s adiabatic efficiency and pumping capacity will
deteriorate significantly if the airfoil surfaces become fouled or
eroded. Suspended liquids or solids in the air going through
the machine can adhere to the airfoil surfaces causing the shape
to change or the surfaces to become rough. This roughness can
either disturb the flow inducing earlier transition to turbulence
in the boundary layers or increase skin friction on turbulent
portions of the blade, if the roughness elements protrude
through the viscous sublayer.
Generally, softer particles adhere to the leading edge,
pressure or suction surface of an airfoil creating increased
surface roughness, yielding greater skin friction losses and less
turning, which is equivalent to greater trailing edge deviation.
The harder particles tend to erode the leading edges.
Even a lightly fouled compressor can have a reduction of
more than 1% in both efficiency and mass flow. The combined
effect can lead to a nearly 2% increase in fuel consumption and
more than a 3% drop in power output. There is also a reduction
in the compressor’s stall margin. In addition, for certain naval
warship applications, the compressor performance deterioration
can significantly increase the starting times of ship service gas
turbines, used for critical electrical power generation.
Copyright ©©2004
2004by
byASME
ASME
 In naval and other marine applications, fouling is generally
caused by salt in the air. While extensive filtration systems are
used to reduce the moisture and salt content to less than 3 parts
per million (ppm), as described in [1], typically several
kilograms of salt are ingested during a 24-hour period.
Other contaminants, such as dust, oily smoke, oil mist
from bearing compartments, or hard particulates may also be
ingested.
Compressors are cleaned by various methods, depending
on the location and nature of the fouling. Pure water sprays, or
soaking the compressor in solvents/detergents are widely used.
However, this does not always work and some foulants adhere
so strongly to blades that abrasive substances, such as ground–
up nutshells, are used to scrape away deposits. However, these
abrasives are used sparingly, since they can erode blades and
cannot be used on blades that employ coatings.
The United States Navy and Coast Guard as well as many
industrial users are shifting their approach to gas turbine
compressor cleaning. Preventative Maintenance Schedules
(PMS), also called Regularly Scheduled Maintenance (RSM)
prescribe maintenance based on hours of operation or a
scheduled time interval. PMS guidelines were developed with
field experience and conservatism in mind. The drawback this
system is the tendency to clean components more often than
required. The goal of Condition Based Maintenance (CBM) is
to perform maintenance only when there is evidence of need.
As stated in OPNAV 4790.16 [2], “unnecessary maintenance
contributes to inflated ownership costs and reduced readiness
for deployable assets”.
In order to implement a CBM strategy for compressor
cleaning, reliable indicators of fouling are required. It may not
be obvious from normally observed parameters that a
compressor is fouled. Specifically, a gas turbine with a fouled
compressor, operating at a low power level, will show only a
small increase in fuel consumption. However, this increase in
SFC could be due to any one or a combination of other factors,
such as increased leakage, reduced inlet temperature resulting
from combustor problems or turbine fouling, to name just a few
possible causes. Therefore, a direct, sensitive, and location
specific measure of compressor fouling is required. A simple
performance measure may indicate that something is wrong,
but it will lack the specificity to allow the operator to make a
decision leading to the proper corrective action(s).
As an example, one may observe a drop in compressor
efficiency and hence spray-wash it, only to find afterwards the
performance is not recovered. It could be that the problem is
due to fouled rear stages from oil leakage, airfoil damage, or
erosion, instead of the customary front stage salt deposits. The
spray wash will not be effective in cleaning rear stages or
repairing damaged or otherwise eroded airfoils. These different
causes of non-recoverable compressor performance each
require different corrective actions. Therefore, it may not be
sufficient to only categorize the degree of deterioration.
Indicators that suggest the type and location of a fault with the
fault can be useful in employing a CBM strategy.
2
2 Copyright © 2004 by ASME
Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 02/02/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
This paper briefly reviews the state of the art in
compressor fouling, detection and localization, provides a
simple meanline analysis that characterizes the
aerothermodynamic signatures resulting from surface fouling in
different stages, and recommends the best parameters for
assessing the degree and location of the fouling.
STATE OF THE ART IN COMPRESSOR FOULING
The state of the art in the aerodynamics of roughed airfoils, the
impact of roughness on cascade total pressure loss and turning,
and the previous work in the area of condition assessment due
to compressor fouling, is provided by Baker [3].
A very brief review is provided here to introduce the
reader to the necessary background for the development and
discussion of the results. If more detail is desired, the
comprehensive review by Baker should be consulted.
Roughness may adversely impact compressor
aerodynamic performance by increasing the frictional shear
forces the air experiences as it flows along the blade surfaces.
This increased “drag” can occur from the roughness creating
disturbances in the front portions of the airfoil in the laminar
regions that promote early transition to turbulence and/or from
increased local friction coefficients, due to the roughness
elements penetrating through the viscous sublayer in the rear
portions of the airfoils, after the flow in fully turbulent. The
basic fluid physics of these phenomena are well known and are
discussed in standard references on boundary layer theory and
turbulence, such as Schlitching [4] and Hinze [5], respectively.
The roughness will degrade performance if a blade is
hydrodynamically rough. The criterion that is generally
accepted for an airfoil to be smooth is based on the Reynolds
number of the roughness elements, and is calculated for a
compressor blade, using
U ⋅k
Re k = < 90 (1)
ν
where U is the free stream velocity, k is the roughness element
height (either equivalent sand grain or RMS can be used), and
ν is the kinematic viscosity of air. Schlichting [4] suggests
Rek < 100 → Smooth
100 < Rek < 1000 → Transitional
Rek > 1000 → Fully Rough
Since the suction surface of an airfoil sees greater local
velocities than the pressure surface, the suction side is more
sensitive to roughness. Rotor blades are similarly most
sensitive when roughness accumulates toward the leading edge.
Koch and Smith [6] suggest an alternative value of the
roughness Reynolds number of 90 as given in Equation 1.
The total pressure loss produced by the blades in
cascades, away from the endwall, is often called profile loss.
Discussions on losses and loss coefficients can be found in
Cumpsty [7] and Gostelo [8]. For compressible flow, the
sectional profile loss is generally defined as
 PT 2 − PT 1
ϖ=
PT 1 − P1
This is the decrease in average total pressure across an airfoil
divided by the difference between the upstream total and static
pressures. The main influence of blade roughness appears
around optimum incidence angles, while the “far-off-optimum
performance” is hardly affected as discussed be Bammert and
Woelk [9]. Kurz [10] provides an argument on why added
roughness on the pressure side has a very small effect
compared to the suction side. Milsch [11] found an increase in
profile losses for NACA 65(12)06 compressor cascades from
2% with ks/C = 0.3 ⋅ 10-3 to 10% with ks/C = 5 ⋅ 10-3, where ks is
the equivalent sand grain roughness element height and C is the
chord. This significant increase in profile loss was caused by
enhanced momentum boundary layer growth, earlier flow
separation, and quicker transition to turbulent flow.
The amount of turning in a cascade is limited by
boundary layer separation on the suction surface. Roughness
on an airfoil tends to increase the momentum thickness on the
suction side and consequently increase the deviation angle,
leading to reduced turning. The only quantitative data on the
impact of increased roughness on deviation that the authors
found in the open literature were due to Mal’tsev and Shakhov
[12], who performed measurements on plane compressor
cascades with solidities of 1.21 and 1.53, 28 degrees of camber,
and wind tunnel stream velocity 33 m/sec (Rec = 2.4 ⋅ 105).
The relative roughness, Rz , defined by
Rz = k
C
is the roughness height, k, divided by the blade chord, C, and
was varied from 1.3 ⋅ 10-3 to 16.7 ⋅ 10-3. Measurements were
made of change in flow lag angle, or deviation angle, δ, with
different solidity and variation of roughness and variation of
the derivative of flow turning angle with respect to the
incidence as a function of roughness. The paper presented the
results both in graphical form and an equation for a curve fit:
1 
1 2 1
δ = K1 + K 2 ( R z ) 2 + K 3 ( R z ) + ( K 4 + i )  ( )
1
 K + K Rz  − K 7  s − K 8 
2 superior to measurements of engine parameters in
 5 6
where C is the chord, s is the blade spacing (C/s = solidity),
K1=6.9°, K2=20.4°, K3=17.9°, K4=10.0°, K7=3.1°; K5=519.0,
K6=27.0, K8=1.2 are dimensionless coefficients. These
correlations of deviation versus roughness were used in the
analysis, which will be described.
Several authors have investigated the effects of
elevated levels of compressor airfoil roughness, most notably
Bammert and Woelk [9], Kurz [10], Lakshminarasimha and
Boyce [13], Boyce, et al [14], Meher-Homji [15], Diakunchak
[16], and Leipold and Boese [17]. The main effect of
compressor fouling is a reduction of mass flow rate. This
reduction increases with operational speed as shown by Doyle
[27], Saravanamuttoo [18], and Zaba [28].
Boyce, et al, [14], Meher-Homji, Boyce, et al, [18],
and Saravanamuttoo and his colleagues [19, 20, 21, 22, 23, 24,
Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 02/02/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
25, 26] investigated detection and classification methods for
(2) compressor fouling. Fouling also reduces the compressor
pressure ratio, and adiabatic efficiency, causing a decrease in
gas turbine power output and an increase in heat rate. Fouling
reduces the operating range on the compressor map and shifts
the operating point to lower mass flow rates, at a given
corrected speed. A reduction in airflow results in a drop in
power output. Roughness makes the constant speed lines
steeper on the compressor map, thus the range of mass flow
rates covered by the characteristic curves decreases. Bammert
[9] showed that the attainable pressure ratios are also reduced.
Scott [29] indicated that intake depression is one of
the most accurate and sensitive of all parameters for measuring
fouling. Boyce [14], Diakunchak [16], and Haq, et al, [26] all
provide recommendations on useful parameters based on field
experience. Simulation studies by Zaba [28] with a 16-stage
compressor showed that the reduction in efficiency and mass
flow due to fouling is highly dependent on location (i.e. front
stages fouled, rear stages fouled. or uniform fouling
throughout). He introduced a fouling influence coefficient to
separate losses caused by deposits according to their location in
the compressor, or turbine.
Based on this extensive review of the open literature, the
following is a summary of the state of the art on the effects of
fouling on cascade aerodynamics, compressor performance, the
detection of fouling using measured performance, and the
ability to use these measurements to determine the location of
fouling.
1. Airfoil roughness, beyond a level where the
(3) roughness elements protrude through the viscous sub-
layer, increases the skin friction and hence momentum
boundary layer thickness, particularly on the suction
sides of the airfoils. This results in higher total
pressure loss and lower turning.
2. Fouling degrades all aspects of compressor
performance, including pressure ratio, mass flow,
efficiency, stall margin, and usable flow range.
3. Direct measurements of compressor performance are
C  (4)
  detecting and localizing compressor fouling.
4. The measurement of static pressure depression at a
compressor inlet, which is an indicator of mass flow,
appears to be one of the most sensitive measures of
fouling. Compressor pressure ratio and efficiency are
in order less sensitive measures. Finally, temperature
ratio is poor indicator of fouling.
5. Influence coefficients, which indicate ratios of
percentage changes in two parameters, have been
shown to provide information on fouling location.
3
3 Copyright © 2004 by ASME
 OBJECTIVES
To the best knowledge of the author, no analysis has been
published in the open literature that provides a basis for
determining the most effective physical quantities to measure to
detect and localize compressor fouling, which clearly accounts
for the two most important physical influences that arise from
airfoil hydrodynamic roughness, namely the increase in profile
total pressure loss and the decrease in blade element turning
and hence stage work. It is the objective of this study to supply
just such an analysis.
DEVELOPMENT OF ANALYTICAL MODEL
A three-stage, meanline analysis was chosen for this work,
because the authors believe it is the simplest model that would
provide a clear physical understanding and useful predictions
of the interactions between and among the stages, without
unnecessary complexity.
Undoubtedly, a full, multi-stage, multi-streamline analysis
that included the impact of tip leakage and other real flow
effects would yield a more accurate model. The goal is not to
have a very accurate model of a particular machine, but rather
to better understand how fouling changes multi-stage machines
and where to look to find the most sensitive and reliable
indicators of fouling.
A simple, meanline, aerothermodynamic model of a three-
stage axial compressor was developed. The model predicts the
effect of fouling, imposed as an increase in surface roughness,
on the fouled stage, upstream stages, downstream stages, and
on the overall performance.
The model was written as a simple excel spreadsheet. The
geometry of the compressor was chosen to approximate the
inlet guide vanes and the first three stages of the Allison 501
compressor. It was a ‘first-order’ prediction of the trend of
performance degradation due to fouling at different compressor
stages.
Figure 1. Three-Stage Compressor Model Schematic.
In the development of this model the following assumptions
were used: Ambient conditions were taken to be ISO conditions
(P=101,325 Pa and T=15°C, 288.15 K); Constant specific heat
(Cp=1.004 kJ/kg-K) and ratio of specific heats (γ = Cp/Cv = 1.4)
was assumed for air; The kinematic viscosity of air was
4
4 Copyright © 2004 by ASME
Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 02/02/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
assumed constant at 1.47 ⋅ 10-5 m2/s; Meanline analysis – the
characteristics of the streamline at mid-span was calculated; All
airflow velocity profiles were assumed 2-D and uniform;
Axisymmetric flow was assumed at the inlet to the bellmouth;
and the processes were assumed adiabatic.
The model only approximates the effects that airfoil
roughness has on profile losses and turning at mid-span.
Secondary flow, tip losses, or other loss mechanisms are not
included.
BASELINE GEOMETRY
The baseline geometry was chosen to roughly match that of an
Allison 501k-34 compressor. The first two rotors have
relatively low camber (γ ≅ 23°) and stagger (λ ≅ 26°). The
third stage has increased camber and stagger. Since the lightly-
cambered inlet guide vanes impart very little swirl on the inlet
air stream, the first few stages operate at very ‘high’ flow
coefficients at design inlet mass flow. NACA 65(12)10 series
blades were used throughout and airfoil staggers were adjusted
to attain reasonable flow coefficients (e.g. φ ≅ 0.65).
The area entering each, successive rotor and stator row
was reduced so the axial velocity remained approximately
constant through the compressor (Cz ≅ 136.7 m/s). The inlet
guide vanes were staggered for zero incidence. Rotor and
stator staggers, λr and λs, were adjusted to attain two degrees of
positive incidence for all rows at the baseline, clean condition.
The geometry remained constant throughout all simulations.
For a clean compressor at a constant shaft speed of N=14,340
RPM, this corresponds to an air mass flow of approximately
15.377 kg/s (34.4 lbm/s) and a compressor pressure ratio of
12.5. A throttle was placed downstream to backpressure the
stages to obtain the correct incidence and mass flow (i.e. the
same as design). Point (0) is assumed to be well upstream of
the bellmouth where stagnation and static temperature and
pressure are equal. The bellmouth radius goes from 0.3810 m at
point (0) to 0.1868 m at the point (1), the throat. The discharge
coefficient, CD, is assumed to be 0.98. Total temperature and
pressure are assumed to be constant across the bellmouth. The
static pressure at the throat must be calculated. Since the Mach
number at the throat is about 0.4, a compressible flow equation
is utilized to account for variable density across the bellmouth.
Accordingly,
1
 γ −1

C D PT 1 A1  P1  γ 2γ   P1  γ  (5)
m& a = C D ρ1 A1cz 1 =    1 −   
RTT 1  PT 1  γ − 1   PT 1  
 
is solved for P1. Mass flow is initially assigned the design
value of 15.377 kg/s. An iteration on the mass flow
computation is used until convergence is reached. The
bellmouth is assumed to be nearly isentropic to calculate the
static throat temperature, T1. The ideal gas law is used to
determine the density at the throat, ρ1, and the inlet axial
velocity is calculated from
m& a = ρ1 A1cz1 (6)
 As a final calculation at the throat, the Mach number at point and
(1) is found from γ

c z1  γ −1  γ −1
M1 = (7) PT 2, rel = P2  1 + M 2, rel 2  (14)
γ RT1  2 
The total pressure loss coefficient, ϖ, was initially assumed to
The inlet guide vanes impart a small amount of turning on the be 0.02 for a hydrodynamically smooth airfoil. The relative
inlet air prior to entering the first stage. There is a small total airflow Mach number for all airfoils was approximately 0.6.
pressure loss across the guide vanes. In the absence of The initial mean roughness height for the clean rotor airfoils
empirical loss data for the very thin, lightly cambered inlet
was assumed to be k = 3 µm. This resulted in a
guide vane profile, the total pressure loss coefficient was
hydrodynamically smooth blade. The outlet relative total
assumed to be 0.02. The total pressure at the guide vane outlet,
pressure is the inlet minus the loss in total pressure:
PT2, was calculated by Equation 2.
PT 3 = PT 2, rel − ϖ ( PT 2, rel − P2 ) (15)
The absolute air inlet angle, α1, is measured from axial.
The outlet air does not follow the extended mean camber line. The initial estimate was used to find the first approximation of
There is a certain amount of mean flow deviation associated outlet static temperature. The work was calculated from
with an airfoil. Since the inlet guide vanes are staggered for Euler’s Turbine Equation:
zero incidence, the deviation angle is subtracted from the (
w = U ( Cθ 3 − Cθ 2 ) = c p TT 3, abs − TT 2, abs ) (16)
camber to find the outlet absolute flow angle. Deviation angle,
Hence, the absolute exit temperature, TT3, was calculated. The
δc, was calculated with Carter’s Rule,
n
isentropic total pressure (without losses) is
s γ
δ C = mγ   (8)  T  γ −1
C PT 3 s = PT 2  T 3  (17)
where n = 1 for IGVs and n = 0.5 for cascades. Carter’s  TT 2 
empirical deviation constant, ‘m’, was obtained for the circular The actual total exit pressure is
arc airfoils from a graph of ‘m’ over a range of stagger angles. PT 3 = PT 3 s − ∆Ploss (18)
It was assumed the axial velocity is constant through the row,
where
so
∆PT ,loss = PT 2, rel − PT 3, rel (19)
cos(α1 )
C2 = C1 (9) The exit static pressure was finally found from the absolute exit
cos(α 2 )
Mach number and total. The airflow is subject to an area
Total temperature is 288.15 K at point (2). The Mach number change between the rotor outlet and stator inlet. To set the
was first calculated by Equation 11, using total temperature to conditions at the first stage stator inlet, the spreadsheet iterates
find the static temperature with P3 in the compressible flow equation until the mass flow equals
 γ −1 2 the mass flow at the bellmouth, m& 1 = m& 3 . The tolerance was
T2 = TT 2 / 1 +  M2 (10)
 2  set at +/- 0.01% for continuity of mass flow. Next, the static
then refined by iteration to find T2. Lastly, static pressure was pressure at the first stage stator inlet was calculated from the
calculated with isentropic relationship,
γ γ −1
 γ −1 2  γ −1
 P  γ
P2 = PT 2 /  1 + M2  (11) T3 = TT 3  3  (20)
 2   PT 3 
The IGV outlet conditions provide the first stage rotor inlet Lastly, the axial velocity was calculated from Equation 10.
conditions. The first stage rotor calculations will now be These conditions were applied to the first stage stator inlet.
described. The method and equations are the same for all three The stator calculations were carried out in the same manner as
rotors. Constant axial velocity is assumed across the rotor. The the IGVs with several differences. Because the stators and
relative inlet and outlet angles were checked with the rotors have the same geometry, the same turning chart was used
appropriate turning chart to ensure symmetric loading. The to ensure symmetric loading.
rotor deviation was estimated with Carter’s Rule. The relative The initial mean roughness height for the clean stator
inlet Mach number for the rotor was found by
airfoils was assumed to be the same as the rotors (k = 3 µm)
W2 and therefore hydrodynamically smooth. As with the rotor, the
M 2, rel = (12)
γ RT2 first stage stator Mach number was approximately 0.6, so the
same curve was used to estimate the total pressure loss
The relative inlet temperature and pressure were found by
coefficient, ϖ. The stator deviation was estimated with Carter’s
 γ −1  Rule. The first stage pressure and temperature ratios were
TT 2, rel = T2  1 + M 2, rel 2  (13)
 2  calculated with

5
5 Copyright © 2004 by ASME

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

 PT 4 The discharge coefficient, CD, was assumed to be 1 (no loss).
π1 = (21) The throat area, Athroat, is the area at the throat of the nozzle, β
PT 2
is the diameter ratio, where
and
D
TT 4 β = throat (24)
τ1 = (22) D8
TT 2 and Y is the expansion factor to account for compressibility,
The stage efficiency was calculated by Equation 6, which where
accounts for the loss measured in cascades, or profile loss
 2 γ +1
γ 
predicted by the model, as well as other losses, such as annulus  P  γ
 P 
(1 − β )   4 T 9
− T 9 
and secondary losses. It is common to quote the estimate for   PT 8  
 PT 8 
 
the different losses given by Howell [30], shown here as Figure
Y=
  (25)
2. In the region of design flow coefficient (about 0.60 to 0.66  2
γ 
 γ − 1   PT 9   4  PT 9 
 
for this 3-stage compressor), annulus and secondary losses may 1 − 1 − β
be considered constant. Howell’s loss relationship attributes  γ  P  
PT 8  
  T8 
  
30% of stage efficiency loss to profile (cascade) losses, 40% to
secondary losses, and 20% to annulus losses. Tip loss was not The diameter of the orifice was adjusted until the mass flow
included. through the nozzle equaled the mass flow into the compressor
When correlated to Howell’s loss relationship, the at the bellmouth. The bellmouth is set so m & a , design = 15.377
estimated stage efficiencies for this 3-stage compressor model kg/s. Fouling was introduced in the form of added roughness to
(Profile+Secondary+Annulus Loss) fit the curve within +/- the first stage baseline, and the simulation was run with only
0.14%. The rotor and stator calculations for each stage proceed the first stage roughened. Next, fouling was introduced as
through the compressor, maintaining continuity of mass flow, added roughness to the second stage baseline, and the
until the third stage outlet is reached, which is point (8) on simulation was run with only the second stage roughened.
Figure 1. Lastly, fouling was introduced as added roughness to the third
A simple flow nozzle as described by Lindeburg [31] was stage baseline, and the simulation was run with only the third
introduced at the third stage exit to backpressure the 3-stage stage roughened. Roughness elements were not actually
compressor, enabling it to take in the design mass flow. The ‘placed’ on the airfoils. A level of roughness was added
throttle discharges to ambient conditions at point (9), which are uniformally to the airfoil surfaces that was assumed to double
at the same ISO conditions as point (0). The mass flow through
the total pressure loss coefficient, ϖ, and increase the deviation,
the throttle is given by,
δs, according to Equation 4. The mean roughness height for the
C A Y  clean, smooth airfoils was assumed to be k = 3 µm. The
m& a ,throttle =  d throat
 2 ρ8 ( P8 − PT 9 ) (23) corresponding roughness parameter was found to be
 1− β 4 
  C/k = 10, 000 for the clean airfoils. The rotor Reynolds
number was approximately Re rotor ≅ 430,000 , and the stator
Reynolds number was approximately Re stator ≅ 340,000 , for all
Additional Losses
1.00
stages. With these values of airfoil Reynolds number and
0.98 roughness parameter, the flat-plate analog of the Moody
0.95 diagram was used to estimate the drag coefficient, CD. It was
Stage Efficiency

0.93
found that C D ≅ 0.004 . It was assumed that doubling the drag
0.90
0.88 coefficient, CD, is similar to doubling the total pressure loss
0.85 coefficient, ϖ. Doubling the drag coefficient made
0.83 Annulus Loss
0.80
Only C D ≅ 0.008 , which corresponds to an approximate roughness
Annulus+S e c o
0.78 nda r y Loss parameter of Chord / k = 1, 000 on the Moody diagram with a
0.75 P rof ile +S e c .+
0.73 Annulus Loss standard chord length of 0.03 m, the resulting mean roughness
0.70 height was k = 30 µm, or ten times the initial, smooth airfoil
0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60 0.65 0.70 0.75 0.80
roughness height. The Reynolds number of the roughness
Flow Coefficient
elements for the fouled airfoils was Re k ≅ 430 for rotors and
Re k ≅ 340 for stators. The reduction in airfoil turning (work)
Figure 2. Howell’s Estimate of Additional Losses
due to added roughness was accounted for by an increase in
deviation, δ, by Equation 4. Figure 3 compares the deviation

6
6 Copyright © 2004 by ASME

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

 found by several methods for airfoils with 30 degrees of
camber and solidity of 1.25. RESULTS
Table 1 presents the effect of fouling of each stage separately
on the overall compressor performance.
Airfoil Deviation Angle vs. Incidence
( S o lid it y = 1, C a m b e r = 3 0 d e g re e s )
9.00 Table1. Compressor Model Overall Performance.
8.75
%∆ %∆ %∆
Deviation Angle [degrees]

8.50
CLEAN (1st stage (2nd stage (3rd stage
8.25 fouled) fouled) fouled)
8.00 Press. Ratio,
πc 1.880 -1.575% -1.517% -1.274
7.75

7.50
Temp. Ratio,
7.25 Car t er - Smoot h
τc 1.206 -0.219% -0.213% -0.151%
Boyce - Smoot h
7.00 Efficiency,
Shakhov - Smoot h
ηc 89.802% -1.580% -1.510% -1.430%
6.75 Shakhov - Rough

6.50
-10 -8 -6 -4 -2 0 2
Incide nce [de gre e s]
Figure 3. Airfoil deviation by various methods.
Although Shakov’s experimental conditions were not exactly
the same as for this simulation, the formula was a relationship
that produced the expected increase in deviation for all fouling
cases; the trends are correct and that was deemed most
important for this first order approximation.
FOULING SIMULATIONS
For the model simulation, each stage was assumed to have
uniform roughness applied over the rotor and stator surfaces,
one stage at a time. While this may not be exactly the
distribution of fouling observed in a real machine, where one
might expect, for example more accumulation of foulants at the
leading edge and other attractive locations for particle
adhesion, this was done for simplicity and is believed to
provide a generic roughened stage.
While the magnitude of the fouling is somewhat
arbitrary, and strongly impacts the compressor performance, the
ratios of changes are nearly independent of the level of
roughness, at least in the first approximation.
Since the absolute level of roughness is somewhat
arbitrary, a roughness that resulted in a doubling of the cascade
profile loss was chosen.
The total pressure loss coefficient and deviation for
stage 2 and stage 3 was estimated from Robbins [35] plot and
Carter’s Rule, as was done for the baseline, clean condition. All
other calculations were performed as previously described for
the clean condition. The ‘macro’ was run for the fouled first
stage condition and the results were examined to see the effect
on performance of the first stage, second stage, third stage, and
overall compressor performance. The same procedure was
conducted for uniform roughening of the second stage rotor
and stator, then finally for the third stage.
Mass Flow
m& a , [kg/s] 15.377 -1.367% -1.147% -0.958%
4 6 8 10
Intake
Depression 11,425 -3.140% -2.640% -2.210%
∆Pintake , [Pa]
The compressor pressure ratio, πc, was reduced 1.58% by first
stage fouling, 1.52% by second stage fouling and 1.27% by
third stage fouling. The compressor temperature ratio, τc,
dropped 0.22% due to first stage fouling, 0.21% due to second
stage fouling, and 0.15% due to third stage fouling.
Compressor efficiency, ηc, had only a slight dependence on
fouling location. Overall efficiency reduced 1.58%, 1.51% and
1.43% due to fouling of the first, second and third stages,
respectively. This corresponds to the fouling localization
studies conducted by Zaba [28], where his fouling simulations
showed very little dependence of compressor efficiency on
fouling location. The reduction in compressor mass flow, m&a ,
reduced 1.37%, 1.15% and 0.96% for fouling of the first,
second and third stages, respectively. The reduction in mass
flow is therefore dependent on the location of fouling. This
dependence agrees with Zaba’s [28] theoretical work and field
experiences, which showed a much greater reduction in mass
flow for early stage fouling as compared to fouling of the latter
stages. Intake depression, ∆Pintake , proved to be the most
sensitive parameter to monitor fouling regardless of location in
the compressor. Intake depression dropped 3.14%, 2.64% and
2.21% due to fouling of the first, second and third stages,
respectively. The superior sensitivity of intake depression to
fouling was demonstrated in several field applications by Scott
[29] and Saravanamuttoo [18].
Table 2 presents various influence coefficients,
indicating ratios of percentage changes in two parameters due
to fouling of each stage separately (3 simulations) on the
overall compressor performance.

7
7 Copyright © 2004 by ASME

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

 Table 2. Influence Coefficients for Overall Compressor reduction in efficiency, is the best to localize fouling.
Performance. The reduction in I m& a ,η c was approximately 12% per
stage as fouling location moved downstream, which is
1st Stage 2nd Stage 3rd Stage
Influence Coefficient a significant, and hopefully measurable decrease in a
Fouled Fouled Fouled
real machine.
% ∆ m& a 3. To a first approximation the influence coefficients are
I m& a ,η c , 0.864 0.762 0.669 independent of the degree of fouling. Doubling the
% ∆η c roughness will increase the mass flow reductions and
% ∆ m& a efficiency degradation in nearly the same ratios.
I m& a ,π c , 0.868 0.756 0.752
% ∆π c The model and predictions that were presented have not yet
% ∆ m& a been experimentally validated. However, it is believed that the
I m& a ,τ c , 6.242 5.385 6.344
results, based on the most complete physical model reported to
% ∆τ c
date, which include both losses and work decrement due to
I ηc ,π c % ∆η c 1.003 0.995 1.122
deviation in turning, are predictive at least qualitatively. A test
, program is underway to provide corroboration of this theory.
% ∆π c

I π c ,τ c , % ∆π c 7.192 7.122 8.437

NOMENCLATURE
Symbol Definition _______Units
% ∆τ c
A Cross sectional area m2
B Roughness function (Nikuradse)
C Chord length m
An influence coefficient comparing the reduction in mass flow C Absolute velocity m/s
to the reduction in efficiency, I m& a ,η c , was seen to reduce CD Discharge coefficient [1]
significantly as the location of fouling moved from front to rear Absolute change in swirl m/s
(~12% reduction in I m& a ,η c per stage as fouling location moved Cf Friction coefficient [1]
from stage 1 to 3). While the influence coefficient does not cp Specific heat at constant pressure kJ/kg-K
match Zaba’s [28] results quantitatively, the trend is in
agreement. Zaba conducted his simulations on a 16-stage
cv Specific heat at constant volume kJ/kg-K
compressor. If more stages were stacked onto this 3-stage ∆hT Change in stagnation enthalpy kJ/kg
model, the range of the influence coefficient, I m& a ,η c , due to i Incidence Degrees
fouling at various stages would likely increase significantly. x ∂y
I Influence coefficient, ≡ ⋅ [1]
y ∂x
CONCLUSIONS k Mean height of roughness m
From the three-stage, meanline, compressor model that was
k1 Decrease in flow coefficient [1]
developed, which included both enhanced losses and increases
in the air flow turning deviation, the following conclusions k2 Decrease in stage efficiency [1]
were drawn: ks Equivalent sand roughness m
1. The static pressure depression at the inlet of the
M Mach number [1]
compressor, ∆Pintake, which is an indicator of mass
m Deviation constant [1]
flow, is the most sensitive parameter to monitor the
m& Air mass flow kg/s
degree of compressor fouling, regardless of location.
N Rotational speed RPM
However, it is a much more sensitive indicator of
n Stage number
typical, early stage fouling. Intake depression was
P Static pressure Pa
approximately twice as sensitive as any other
parameter to indicate early stage fouling. PT Stagnation pressure Pa
2. For uniform fouling of a single stage, the reduction in ∆P Intake depression Pa
mass flow decreases steadily as the fouling location Rair Gas constant for air (287) J/kg-K
moves downstream. The reduction in compressor Re Reynolds number [1]
efficiency has a slight dependence on fouling location. Relative roughness [1]
Hence, an influence coefficient, I m& a ,η c , comparing the S Solidity [1]
percent reduction in mass flow to the percent s Spacing m

8
8 Copyright © 2004 by ASME

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

 T Static temperature K NOTATION
TT Stagnation temperature K CBM Condition Based Maintenance
r EGT Exhaust gas temperature
U Free stream velocity m/s IGV Inlet guide vane
U Wheel speed m/s ISO International standard organization
r
W Relative velocity m/s NACA National Advisory Committee for Aeronautics
w Specific work kJ/kg PMS Preventative Maintenance Schedule
∆Wθ Relative change in swirl m/s
ACKNOWLEDGMENTS
Y Throttle expansion factor
The authors acknowledge the Naval Surface Warfare Center,
Carderock Division (NSWCCD), Gas Turbine Branch,
Greek Definition Units
Emerging Technologies group, Code 9331, in Philadelphia, PA,
α Absolute angle (relative to axial) Degrees
their continued support and technical contributions.
β Relative angle (relative to axial) Degrees
γ Airfoil camber angle Degrees REFERENCES
cp 1. MIL SPEC, MIL-E-17341C (ships), 1970, Engines,
γ= Ratio of specific heats [1] Gas Turbine, Propulsion and Auxiliary Naval
cv Shipboard. April 9, 1970.
δ Deviation (flow and TE metal angle) Degrees 2. OPNAV INSTRUCTION 4790.16, 2000, “Condition-
ηc Compressor adiabatic efficiency [1] Based Maintenance,” 2000.
λ Stagger angle Degrees 3. Baker, J. D., 2002, Analysis of the Sensitivity of
Multi-Stage Axial Compressors to Fouling at Various
ν Kinematic viscosity m2 / s Stages, Master’s Thesis, Naval Postgraduate School,
θ BL momentum thickness. m September 2002.
p −p 4. Schlichting, H., 1987, Boundary Layer Theory, 7th
ϖ = T 1 T 2 Total Pressure Loss Coefficient [1]
Edition, McGraw Hill, 1987.
pT 1 − p1
5. Hinze, J. O., 1987, Turbulence, 2nd Edition, McGraw
p Hill, 1987.
π c = T 2 Compressor total pressure ratio [1]
pT 1 6. Koch, C. C., Smith, L. H., 1976, “Loss Sources and
Magnitudes in Axial Flow Compressors”, Transactions
ρ Density kg / m3
of the ASME, Journal of Engineering for Power,
ψ = w / U 2 Pressure (or work) coefficient [1] Volume 98, 1976, pp. 411-424.
TT 2 7. Cumpsty, N. A., 1989, Compressor Aerodynamics,
τc = Total temperature ratio [1] Longman Singapore Publishers, 1989.
TT 1 8. Gostelow, J. P., 1984, Cascade Aerodynamics,
Pergamon Press, 1984.
SUBSCRIPTS 9. Bammert, K., Woelk, G.U., 1980, “The Influence of
abs Absolute the Blading Surface Roughness on the Aerodynamic
b Boyce (deviation) Behavior and Characteristic of an Axial Compressor,”
c Carter (deviation) Transactions from ASME, Journal of Engineering for
c Compressor Power, Volume 102, April 1980, pp. 283-287.
corr Corrected 10. Kurz, R., Brun, K., 2000, “Degradation in Gas
f Fouled Turbine Systems,” ASME Proceedings, 2000.
h Hub 11. Milsch, T. R., 1965, “Total Pressure Losses due to
k Roughness elements increased Roughness”, NACA, SP-36. 1965.
r Rotor 12. Mal’tsev, Y. N. and Shakhov, V. G., 1989, “Influence
rel Relative of Roughness of Deposits in Compressor Cascade on
s Shakhov (deviation) Flow Lag Angle,” Soviet Aeronautics, Volume 32,
s Stator Number 3, 1989.
st Stage 13. Lakshminarasimha, A. N., Boyce, M. P., et al, 1994,
t Tip “Modeling and Analysis of Gas Turbine Performance
z Axial Deterioration”, Transactions of the ASME, Journal of
Engineering for Gas Turbines and Power, Volume 116,
January 1994, pp. 46-51.

9
9 Copyright © 2004 by ASME

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

 14. Boyce, M. P. et al, 1993, “Condition Monitoring and
Diagnostic Approaches for Advanced Gas Turbines,”
IGTI Publication, 1993.
15. Meher-Homji, Cyrus B. 1990, “Gas Turbine Axial
Compressor Fouling – A Unified Treatment of its
Effects, Detection, and Control”, ASME -
International Gas Turbine Institute Publication,
Volume 5, 1990.
16. Diakunchak, I.S., 1991, “Performance Deterioration in
Industrial Gas Turbines,” Proceedings from the 1991
ASME IGTI, June 1991.
17. Leipold, R., Boese, M., et al, 2000, “The Influence of
Technical Roughness Caused by Precision Forging on
the Flow Around a Highly Loaded Compressor
Cascade”, Transactions from ASME, Journal of
Turbomachinery, Volume 122, July 2000, pp. 416-424.
18. Saravanamuttoo, H. I. H.; MacGillivray, P. J., et al,
1984, Maclssac, B.D., “Development of Diagnostic
Model for Marine Gas Turbines”, Proceedings from
the 29th International Gas Turbine Conference and
Exhibit, Amsterdam, Netherlands, 1984.
19. Saravanamuttoo, H. I. H., Mirza F., 1997, “Off-Design
Performance Prediction of Single-Spool Turbojets
using Gasdynamics”, Journal of Propulsion and
Power, Volume 13, Number 6, November 1997.
20. Saravanamuttoo, H. I. H.; Muir, D.E.; et al, 1988,
“Health Monitoring of Variable Geometry Gas
Turbines for the Canadian Navy”, Proceedings from
the 1988 International Gas Turbine and Aeroengine
Congress, Amsterdam, Netherlands, ASME Paper #
88-GT-77.
21. Saravanamuttoo, H. I. H.; Aker, G. F., 1989,
“Predicting Gas Turbine Performance Degradation
Due to Compressor Fouling Using Computer
Simulation Techniques”, Journal of Engineering for
Gas Turbines and Power, Volume 111, Number 2,
1989.
22. Saravanamuttoo, H. I. H. and Lakshminarasimha, A.
N., 1986, “Prediction of Fouled Compressor
Performance Using Stage Stacking Techniques”,
Proceedings from the 4th Joint AIAA/ASME Fluid
Mechanics, Plasma Dynamics and Lasers Conference,
1986.
23. Saravanamuttoo, H. I. H. and Lakshminarasimha, A.
N., 1985, “Preliminary Assessment of Compressor
Fouling”, Turbomachinery International, Volume 26,
Number 7, September 1985.
24. Saravanamuttoo, H. I. H. and Seddigh, F., 1990,
“Proposed Method for Assessing the Susceptibility of
Axial Compressors to Fouling”, Proceedings from the
1990 International Gas Turbine and Aeroengine
Congress and Exposition, June 11-14, 1990, Brussels,
Belgium, ASME Paper #GT-348.
25. Saravanamuttoo, H. I. H., MacIsaac, B. D., 1983,
“Thermodynamic Models for Pipeline Gas Turbine
10
10
Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 02/02/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Diagnostics”, Journal of Engineering for Power,
Transactions of ASME, Volume 105, Number 4, 1983.
26. Haq, Inamul and Saravanamuttoo, “Detection of Axial
Compressor Fouling in High Ambient Temperature
Conditions,” Proceedings from the 1991 ASME IGTI,
June 1991.
27. Doyle, M.D. and Dixon, S.L., 1962, “The Stacking of
Compressor Stage Characteristics to Give an Overall
Compressor Performance Map,” Aeronautical
Quarterly, Nov 1962.
28. Zaba, T., 1980, “Losses in Gas Turbines Due to
Deposits on the Blading,” Brown Boveri Review,
Volume 67, Number 12, December 1980.
29. Scott, J. H., 1979, “Axial Compressor Monitoring By
Measuring Intake Air Depression,” 3rd Symposium on
Gas Turbine Operations and Maintenance, National
Research Council of Canada Symposium, September
1979.
30. Howell, A. R., 1948, “The Aerodynamics of the Gas
Turbine”, Journal of the Royal Aeronatical Society,
Volume 52, 1948.
31. Lindeburg, M. R., 2001, Mechanical Engineering
Reference Manual, 11th Edition, Professional
Publications, Inc, 2001.
32. Hill, P. G. and Peterson, C. R., Mechanics and
Thermodynamics of Propulsion, Second Edition,
1992.
33. Lindeburg, M. R., Mechanical Engineering Reference
Manual, 11th Edition, Professional, 2001.
34. Lofdahl, Jan-Olof, “35 Years of Operation With Rolls-
Royce Marine PROTEUS Gas Turbines in the
Swedish Navy Fast Surface Attack Ships,”
Proceedings of the 1996 IGTI and Aeroengine
Congress and Exhibition, 1998, ASME Paper #98-GT-
147.
35. Robbins, W H. and Dugan, J. F., 1965, “Prediction of
Off-Design Performance of Multi-Stage
Compressors,” NASA SP-36, 1965.
36. Saravanamuttoo, H.I.H. and Lakshminarasimha, A.
N., 1985, “Preliminary Assessment of Compressor
Fouling,” Turbomachinery International, Volume 26,
Number 7, Septmeber 1985.
37. Saravanamuttoo, H. I. H. and Seddigh, F., 1990,
“Proposed Method for Assessing the Susceptibility of
Axial Compressors to Fouling,” IGTI Proceedings,
1990.
38. Schaffler, R., 1980, “Experimental and Analytical
Investigation of the Effects of Reynolds Number and
Blade Surface Roughness on Multistage Axial Flow
Compressors”, Transactions of the ASME, Journal of
Engineering for Power, Volume 102, 1980.
39. Suder, K. et al, 1995, “The Effect of Adding
Roughness and Thickness to a Transonic Axial
Compressor Rotor,” Journal of Turbomachinery,
Volume 117, October 1995.
Copyright © 2004 by ASME