CONDITION-BASED MAINTENANCE FOR GAS TURBINES PLANTS

Christoforos Romesis Yi-Guang Li

Laboratory of Thermal Turbomachines Department of Power and Propulsion
National Technical University of Athens, Cranfield University, Bedford
Greece England

Abstract
Condition-Based Maintenance (CBM) is a maintenance approach for machinery and
industrial equipment, where decisions are made based on the current condition of the asset,
the operation of which is monitored. A CBM program consists of three parts: Data
Acquisition, Data Processing and Decision Making, where the last one can further be divided
into a Diagnostic part and a Prognostic part. This paper attempts an overview of the concept
of CBM in the field of Gas Turbines. A review of certain key methods and techniques is
provided, along with some existing commercial tools. Through this review, the benefits of
incorporating CBM policies into Gas Turbine plants are revealed and the areas where future
research should focus on are highlighted.

1 INTRODUCTION
One of the major challenges for Gas Turbine users is to ensure high level of engine
availability and reliability, and efficient operation during their complete life-cycle. For this
purpose, various maintenance approaches have been introduced over the years.
Historically, the earliest maintenance approach of machinery equipment is the so-
called Breakdown Maintenance or Run to Failure, according to which maintenance actions
are taken only after breakdown. A later and more advanced maintenance approach, the
Preventive Maintenance or Scheduled Maintenance, involves maintenance actions after
specific time intervals of operation, regardless the condition of the engines. Nowadays, due to
increased complexity of Gas Turbine plants along with higher safety standards and lower
profit margins, a move from traditional maintenance approaches to more reliable and cost-
effective maintenance approaches, is required. This leads to Condition-Based Maintenance
(CBM), where maintenance actions are taken according to the actual condition of the
operating engines, which is assessed through appropriate condition monitoring procedures.
According to Jardine et al. [1]: CBM is a maintenance program that recommends
maintenance actions based on the information collected through condition monitoring. CBM
attempts to avoid unnecessary maintenance tasks by taking maintenance actions only when
there is evidence of abnormal behaviors of a physical asset.
In general, a CBM program consists of three main parts [1], [2]:
The Data Acquisition part, where data are acquired from engines under monitoring;
The Data Processing part, where the acquired data are validated and transformed
properly according to the requirements of the decision-making techniques that follow;
The Decision Making part, where a number of methods and techniques are applied to
 the available data to obtain the current health condition of the engine and recommend
maintenance plans.

The aim of the paper is to provide a comprehensive review of existing CBM tools,
techniques and procedures. This review tries to reveal both the importance of incorporating
CBM approaches into gas turbine plants and the fields where relevant future research should
be focused on. Thus, the structure of the paper is as follows. A brief review of existing tools,
methods and techniques related to CBM is given in Section 2. Methods developed by the
research teams in the Laboratory of Thermal Turbomachines of National Technical University
of Athens (LTT/NTUA) and the Department of Power and Propulsion of Cranfield University
(CU) are also described in the same Section. Some existing commercial CBM programs are
described in Section 3. Finally, a summary, discussions, conclusions and potential future
research in the area are provided in Section 4.

2 BRIEF REVIEW ON CBM METHODS AND TECHNIQUES

This section provides a brief review of existing methods and practices covering all
aspects of CBM, organized into sections representing the three parts of a CBM program. A
more in-depth review can be found in [1], [2] and [3].

2.1 DATA ACQUISITION

The first part of a CBM program is related to Data acquisition, which is the process of
collecting and storing useful information from engines for CBM purposes. There are two
types of data that can be acquired; Event Data and Condition Monitoring Data. Event data
include all historical data and records of the targeted gas turbine engines, such as history of
faults, breakdowns, repairs, overhauls etc., while Condition Monitoring Data consist of all
acquired measurement along the operating engines, indicative of its current operating and
health condition.
Although the majority of available CBM methods and techniques rely only on
Condition Monitoring Data, Event Data are equally important to efficient CBM, as stated by
Jardine [1]. However, Event Data are not acquired usually in an automated way and therefore,
manual data entry is required.
Condition Monitoring Data are normally collected by using available instrumentation
on gas turbine engines: vibration measurements through accelerometers, velocity pick-ups
and displacement sensors; pressure measurements at various gas path stations of the engines
through pressure transducers; temperature measurements through thermocouples and/or
resistance temperature detectors; shafts speed measurement; lube oil measurements (pressure,
temperature, corrosion through corrosion probe); fuel system measurements through water
capacitance probe, corrosion probe, and Btu detector; exhaust gas analysis; torque
measurements; ambient conditions measurements, namely ambient pressure, temperature and
relative humidity. A more detailed description on gas turbine condition monitoring
measurements, sensors and sensing strategies are given in [1] and [4].

2.2 DATA PROCESSING

The second part of a CBM program is data processing, which can be divided into two
main processes; Data Cleaning and Data Analysis.

2
 2.2.1 Data Cleaning
Data cleaning is the first step of data processing. At this step the acquired
measurements are filtered to reduce measurement noise and validated, through appropriate
sensor validation techniques.
Measurement noise can be significant and can alone lead to wrong assessment of
engine health condition. Thus, data processing to reduce measurement noise, called noise
filtering, should always be applied. A good summary of state-of-the-art noise filtering
techniques and related signal processing processes is given in [2].
Existence of a sensor malfunction may mislead a CBM program. Thus, application of
appropriate sensor validation techniques is common to CBM. These techniques aim at
detecting and isolating any possible sensor fault, while more modern techniques allow also an
estimation of the underlying sensor bias. Over the years, several types of Sensor Validation
techniques have been proposed by many researchers, such as analytical redundancy methods
[5], approaches related to Kalman filter [6] or Fuzzy Logic methods [7] and Artificial Neural
Networks [8]-[12].

2.2.2 Data analysis

The second step of data processing is Data analysis. At this step acquired
measurements are further processed in order to extract useful information that can be used at
the Decision Making part of CBM for diagnostic and/or prognostic purposes. For this reason,
this step can also be found in literature under the term feature extraction. The main
principles of data analysis and signal processing techniques usually applied are described in
[2].
Data analysis approaches strongly rely on the nature of available data. Condition
monitoring data are versatile and can be grouped into two main categories [1]:
a. Value type, where data of a condition monitoring variable is acquired in the form of
single values. For example, temperatures and pressures along the engine gas path during
operation are all value type data.
b. Waveform type, where data of a condition monitoring variable is acquired in the form of
time series, also called time waveform. For example, vibration data and acoustic
emissions data are waveform type data.

Value Type data

Feature extraction on Value Type data depends mainly on the problem in concern. A
typical approach applied on aerothermal data which is the majority of Value Type data of
Gas Turbines condition monitoring variables is the estimation of the percentage deviation of
a measurement from its nominal value, the co-called Delta of a measurement [13]. For an
acquired measurement of a condition monitoring aerothermal variable, where its nominal
value during a fault-free operation of the Gas Turbine at the same operating conditions is ,
the Delta () is defined as follows:

=
100% (1)

The use of Deltas has advantages over the use of raw measurements for two reasons.
Firstly, the dependence of Deltas on the operating point is very weak when neighboring
operating conditions are considered. Secondly, specific Gas Turbines fault conditions result in
distinct measurements Deltas and therefore Deltas can act as patterns of specific engine health
conditions.

3
 Another issue related to Value Type data, is the Data Reduction. Value type data,
although sound simpler than Waveform Type data, usually involve large number of variables,
each of them having different degree of correlation with the phenomena of interest. In these
cases, Data Reduction Techniques, such as Principal Component Analysis (PCA) and its
extension, Independent Component Analysis (ICA), can be applied. A review of Data
Reduction techniques can be found in [14], while an example of application is given in [15].

Waveform Type data

There are three main categories of Waveform Type data analysis in the literature:
Time-Domain analysis, Frequency-domain analysis and Time-frequency analysis.
In Time-Domain analysis features are extracted directly from the waveform time-
series. More traditional approaches rely on the principles of statistical analysis and involve
the estimation of parameters, such as: mean values, standard deviations, peak-to-peak
intervals, skewness, kurtosis etc. [1] . More advanced approaches involve time-series models,
such as Auto-Regressive (AR) models and Auto-Regressive Moving Average (ARMA) models
[16] .
In Frequency-Domain analysis features are extracted through signal processing in the
frequency domain. The most popular analysis is the Spectrum Analysis by means of Fast
Fourier Transform (FFT). The most commonly used technique for Spectrum Analysis is
Power Spectrum, although other techniques have also been developed, such as Cepstrum,
High-Order Spectrum, etc. [17], [18].
Time-Frequency analysis allows feature extraction from Waveform Type data in both
the time and frequency domain. This is particularly useful for transient data analysis, which is
very common during Gas Turbines operation at the presence of faults [1]. The most
commonly used techniques, among others, are: Short-time Fourier Transform (STFT) and
Spectrogram [19], which is the power of STFT, WignerVille distribution [20] and Wavelet
Analysis [20], [21] .

2.3 DECISION MAKING

The third part of a CBM program is Decision Making. This is the part of CBM where
the assessment of the health condition of gas turbine plant takes place and also maintenance
plans are proposed. This part of CBM covers two areas: Diagnosis and Prognosis. A brief
review of available methods and techniques on both areas is given in this paragraph, along
with a review on approaches related to Information Fusion. Fusion technique is highly
associated with both Diagnosis and Prognosis. In the last few years it has attracted a lot of
interests of researchers in CBM practices and therefore it is mentioned separately.

2.3.1 Diagnosis
Diagnosis can be described as a procedure of reasoning to interpret the heath
condition of machinery equipment using data acquired during its operation. As stated in [2],
there is no consensus on the terminology related to Diagnostics. However the following
definitions have been adopted by the military and other sectors of industry:
Fault diagnosis is defined as the procedure of detecting, isolating and identifying an
impending or incipient failure condition, during which the affected component is still
operational even though at a degraded mode.
Failure diagnosis is defined as the procedure of detecting, isolating and identifying a
component that has ceased to operate.

4
 Fault (failure) detection, isolation and identification can further be given the
following meaning:
- Fault (failure) detection: An abnormal operating condition is detected and reported.
- Fault (failure) isolation: Determining which component is failing or has failed.
- Fault (failure) identification: Estimating the nature and extent of the fault (failure).

There is an extent bibliography on gas turbine diagnostics, containing a large number

of methods that have been proposed by many researchers over the years. Existing diagnostic
methods can be grouped into three main categories and are reviewed in the following:

Model-based approaches
This category includes diagnostic methods where the core tool for diagnostic
reasoning is based on engine thermodynamic models, where the relationship among the
involved parameters is determined through explicit mathematical and thermodynamic
equations. In the field of gas turbines gas path diagnostics, model-based diagnostic
approaches rely on engine thermodynamic performance models. Different model-based gas
path diagnostic methods have been developed since GPA was first introduced by Urban [22]
around 1967. For example, GPA and many of its derivatives were introduced by Urban et al.
[23], Voponi [24], Doel [25] and [26], Stamatis et al. [27], Escher and Singh [28], Aretakis et
al. [29], Li [30], etc.

Statistical approaches
In this category all methods in which diagnostic conclusions are derived mainly
through the use of the principles of statistics. These methods are suitable in cases where a
large amount of historical records are available or where various health conditions of the
engines can be reproduced (e.g. through an engine performance model). In all these cases
useful diagnostic information can be extracted through appropriate application of statistics.
Moreover, statistical approaches are also able to handle the inherent stochastic nature of the
diagnostic problem.
The simplest statistical approaches can deal with fault detection through a classic
Hypothesis Test, with null hypothesis H0: Fault A is present, where the alternative
hypothesis is H1: Fault A is not present. In actual diagnostic problems, however, more
advanced statistical approaches are required. The most popular approach is Statistical Pattern
Recognition methods, where the acquired set of measurements from an operating engine is
compared with measurements sets that result from known engine faults. This is a
classification method where an examined case is classified into one of a number of predefined
known health conditions of the engine. A typical Pattern Recognition method is described in
[29]. Another popular approach that fall into this category is Kalman Filters, where engine
health condition is estimated through a recursive statistical processing of measurements time
series. A good example of Kalman Filter-based diagnostic methods is described in [31]. Other
statistical approaches used for gas turbine diagnostics are Hidden Markov Models (HMM)
[32], which is a method based on the Markov Process of probability theory and Support
Vector Machines (SVM) [33], which is a classification method.

Artificial Intelligence (AI) approaches

AI methods for gas turbine diagnosis have become very popular since more than a
decade ago due to many reasons. For example, many AI methods are capable of reasoning
and revealing the cause-effect relationship among observable measurement and non-

5
observable health parameters when the functional relationship between them is unknown or
incomplete. Another reason is their ability to deal efficiently with large systems where lots of
parameters may be involved and vast amount of calculations may be required, but the whole
calculations may be divided into smaller actions taking place in parallel. Additionally, some
AI methods offer higher efficiency in handling stochastic phenomena like those encountered
in diagnostic problems.
There are many AI methods that have been developed for gas turbine diagnosis. The
most traditional approach is the Expert Systems (ES) where an inference engine consisting of
a set of IF-THEN type of rules [34] is used. One of the most popular AI methods is the
Artificial Neural Networks (ANN). Different architectures and algorithms of ANNs have been
introduced for gas turbine gas path diagnostics, for example, diagnostics using Feed Forward
Back-Propagation (FFBP) ANN [35], Auto-Associative Neural Networks (AANN) [36],
Radial Basis Function (RBF) networks [37], Probabilistic Neural Network (PNN) [12] and
Self-Organizing Maps (SOM) which are unsupervised ANN [38]. A nested neural network
approach was introduced by Ogaji and Singh [39]. In addition, Genetic Algorithms (GA) that
mimics natural evolution processes is also a popular AI method [40]. Stochastic AI methods
widely used are Bayesian Belief Networks (BNN) [41], which are a combination of
probability Theory with Graph theory and Fuzzy-Logic that measure the degrees of truth of
the involved parameters value [42].

2.3.2 Prognosis
Prognosis can be considered as the procedure of estimating the future health condition
of an engine based on past and present information acquired from the engine and/or
assessment of its current health condition. Prognosis is practically expressed through the
estimation of the Remaining Useful Life (RUL), which is the remaining safe operating time
before failure. RUL estimation requires estimation of the fault propagation process and
knowledge of the failure mechanism. The fault propagation process is usually estimated
through forecasting models and techniques given available historical measurements and
health conditions of the engine. The failure mechanism, on the other hand, can be expressed
in many ways varying from a simple set of hard limits of selected engine parameters to more
sophisticated models (e.g. mathematical model of crack growth).
Compared to Diagnosis, the literature of Prognosis is much more limited.
Nevertheless, existing prognostic methods are grouped into three categories, i.e. Model-based
approaches [43], Statistical approaches [44], [45] and AI approaches [46].
The ultimate goal of any Prognostic approaches is to suggest maintenance actions and
optimize maintenance plans. Maintenance plan optimization is practically a problem of
optimizing some key factors, such as safety, risk, cost, reliability, availability etc. The concept
of incorporating prognostics into maintenance plans is discussed in [1] while some early work
in this area is described in [47] and [48].

2.3.3 Lifing Analysis

Life assessment has always been a major concern to the gas turbine users for both
safety and economy considerations. Overestimating the engines life could lead to
catastrophic incidents and economic losses while underestimating the engines life will cause
the premature removal of the engine. The large safety margin imposed by Original Equipment
Manufacturers (OEM) to ensure safety operations makes the engines life estimation too

6
conservative, which leads to a waste of engine life during engine life span. Moreover, as the
life limits provided by the OEMs are normally calculated on the basis of a design envelope of
expected base load, calculated mechanical and thermal stresses as a function of the operating
condition and the capability of the materials within those conditions, the guidelines do not
always address the specific operating environment and requirements of each operator.
Therefore, having the knowledge of how engine life responds to changes in the
ambient, operating and health conditions becomes very important as these changes will affect
the engine performance parameters and hence alter the engine life. If the changes can be
quantified, it will help the users to make informed maintenance decisions, maximise operation
effectiveness and thus cut down operating costs.
Many studies have been conducted in order to understand how these changes affect
the creep life. While most of them focus on investigating creep life from the material point
of view, less information is available in the open literatures that discuss the creep life from the
performance, operation or change of engine health points of view.
Some research focusing on engine creep life consumption due to varying engine
operation and health conditions has been published. For example, Tinga et al. [49] examined
the degradation effect on the high pressure (HP) turbine blades creep life of a turbofan F100-
PW-220 engine using an integrated lifing model. Naeem et al. [50] used a lifing model to
investigate the degradation effect on the HP turbine blades creep life of a turbofan F404-GE-
400 engine. Abdul Ghafir et al. have introduced a Creep Factor to estimate the creep life
consumption of HP turbine of a helicopter engine using a model based approach [51] and an
artificial neural network approach [52].

2.3.4 Information Fusion

In the last few years more and more researchers recognize the need of combining
multiple sources of information for the CBM needs as relying on a single source of
information may lead to inaccurate conclusions (e.g. inaccurate diagnostic conclusions made
from a specific diagnostic method) or unreliable condition monitoring (e.g. measurement
acquired from a single sensor that may fail or malfunction). Additionally, combining many
sources of information may expand the capability of a CBM program (e.g. combining many
diagnostic methods, each perform better than the others on diagnosing specific fault
conditions).
This combination of information sources is described by the term Information Fusion.
Fusion is not a specific part of a CBM program, but a methodology that can be applied to all
parts of it. Since it gains a lot of interest lately by the research community, it is described as
follows.
Techniques allow Information Fusion can be grouped into three categories depending
on the part of CBM they are applied. These are Data-Level Fusion, Feature-Level Fusion and
Decision-Level Fusion. Data-Level Fusion refers to techniques allowing fusion of the
readings of multiple sensors and resulting in a single value of the measured quantity. A
description of the techniques on Data-Level Fusion is given in [2]. Feature-Level Fusion
refers to techniques that combine different signal processing approaches and result in a
feature extraction incorporating valuable information for the Decision making part of CBM.
An example of a Feature-Level Fusion technique is presented in [53]. Decision-Level Fusion,
on the other hand, refers to techniques that combine diagnostic/prognostic methods aiming at
a more accurate and/or more reliable diagnostic/prognostic conclusions. At the Decision-
Level Fusion, there are techniques that combine the results of independently acting methods

7
(see, for example, [54]) and techniques where one method feeds another, as in [55] for
example. Another example of a Decision-Level Fusion approach applied to both diagnosis
and prognosis is [56] where an information fusion system for engine diagnostics and health
management developed for the P&W F117 engine is described [57].
Information fusion techniques are based on methods like Dempster-Schafer theory
[54], Artificial Neural Networks [58], Fuzzy Logic [59] and so on.

2.4 CBM performance metrics

One question that the CBM community should answer is how the performance and
effectiveness of CBM systems can be assessed. This question is important since neither
commercial nor military standards exist to support a systematic and consistent approach to
assessing the performance and effectiveness of CBM technologies and this is recognized as
one of the main barrier between research community and engine users, who are discouraged
in investing and incorporating CBM systems into their plants [2].
Performance metrics involve well established, generally accepted and independently
acting procedures that are able to assess both the technical and economic feasibility of CBM
systems. Such procedures should not only allow validation and evaluation of incorporated
methods and techniques at all parts of a CBM program (ranging from data acquisition and
processing, to fault diagnosis, prognosis, lifing and maintenance planning), but also be able to
measure the combined impact of the CBM program, including evaluating both costs and
performance of the whole CBM system.
So far, efforts are limited in a number of few benchmark cases proposed by several
researchers to allow the evaluation of available diagnostic methods [60], [35], [61].
Evaluation of diagnostic methods has been studied in a more systematic way during the
OBIDICOTE project conducted by European Research Community, where a set of
benchmark fault cases have been considered and used by many researchers for diagnostic
methods evaluation [62], [40], [55]. This idea has been further supported by the Engine
Health Management Industry Review (EHMIR). The EHMIR provided references or theme
problems to aid technology development and evaluation. A gas path diagnostic theme and
associated metrics for benchmarking the performance of diagnostic solutions resulted from
the EHMIR effort is described in [63].
A more detailed discussion on the concept of CBM performance metrics is
attempted in [2] where the lack of standards is mentioned and also a number of approaches
that can be used as starting point in filling the gap are described. These approaches involve:
- Feature-evaluation metrics
- Fault diagnosis performance metrics
- Prognosis performance metrics
- Effectiveness metrics (diagnosis, prognosis)
- Cost/benefit analysis of CBM

2.5 Contribution of the LTT/NTUA research group

The research group of the Laboratory of Thermal Turbomachines of the National
Technical University of Athens (LTT/NTUA) is active for nearly 25 years in the field of gas

8
turbines condition monitoring and diagnostics. The research activities of LTT/NTUA group
expand to the whole range of the CBM area and they have resulted in the development of a
large number of methods and techniques, as well as complete monitoring and diagnostic
systems that are on use.
Suggestively, LTT/NTUA has developed a number of methods allowing sensor
validation and sensor fault diagnosis. Among them, there are Pattern Recognition Methods
[29], Model-based methods through appropriate optimization techniques [64] and
Probabilistic Neural Network based methods [12] that allow the detection, isolation and
identification of sensor faults, even at the simultaneous presence of engine component faults.
Additionally, LTT/NTUA has developed a number of model-based diagnostic
methods, varying from Adaptive Performance Modeling methods [65], [66], Deterioration
Tracking methods [13] allowing the estimation of engine performance parameters during
gradual deterioration of its components performance, a combinatorial approach [67] and an
optimization technique [60], [68] that both handle efficiently the problem of estimating
engine performance parameters from a limited set of measurements, along with a performance
model zooming for in-depth component fault diagnosis [69]. A number of developed
diagnostic methods have been applied in cases where Waveform Type data were available,
including Pattern recognition Methods [70], Wavelet analysis [21] and Stochastic approaches
[71].
A wide range of developed methods falls into the category of AI approaches, like
methods based on Bayesian Belief Networks [40], Probabilistic Neural Networks [72], Fuzzy
Logic [73] and Neural networks based methods for diagnosis through engine emissions [74].
In recent years fusion approaches have also been developed incorporating principles of the
Dempster-Schafer theory [54], [75], Probabilistic Neural Networks [76] and Fuzzy logic [77].
In the area of prognostics incorporating maintenance policies, LTT/NTUA has recently
proposed a procedure for compressor washing economic analysis and optimization [78].
The aforementioned methods have been applied successfully for the diagnosis of
benchmark fault cases [41] and real implanted faults [79] and a number of CBM software
have been developed by LTT/NTUA. Among them, is EGEFALOS software [80] that allows
condition monitoring and fault diagnosis and so far is in operation on a FIAT TG-20 and an
ABB-GT10 gas turbine. A condition monitoring software has also been developed for the
Hellenic Air force, implemented in the J-79 jet-engines test cell. Currently, CBM software
allowing condition monitoring, diagnosis and prognosis for a 334MW CCGT comprising two
PG9171 GE gas turbines and one SST-900 Siemens steam turbine, is under development.

2.6 Contribution of the CU research group

Research in gas turbine gas path diagnostics at Cranfield University was initiated by
Prof. Riti Singh and has been carried out around 30 years. It has been supported and funded
by several industrial partners, including Rolls-Royce, Manx Electricity Authority (MEA) in
the UK and China Aviation Powerplant Research Institute of AVIC in China. Different gas
path diagnostic, prognostic and lifing techniques have been developed by the Cranfield team,
a large number of technical papers have been published, several patents [81]-[83] have been
filed and computer software has been developed. More details of the relevant technologies
developed at Cranfield are described as follows.
One of the key technologies for gas turbine gas path diagnostics is the capability of
accurate performance modeling of gas turbine engines. Rigorous gas turbine performance
modeling techniques, including design and off-design performance modeling [84] with the

9
consideration of a variety of gas turbine configurations, varying ambient conditions, variable
geometries, multiple fuels, water injection, degradation, etc. all implemented in computer
program Turbomatch have been developed. More recently, design point performance
adaptation [85] and off-design performance adaptation techniques [86], [87] have also been
developed in order to adapt engine thermodynamic model to real engine performance due to
the very high requirement of accurate performance models for gas path diagnostics.
In gas path diagnostics, different model based and non-model based techniques have
been developed over the years, including linear and nonlinear GPA based on Influence
Coefficient Matrix [88], [89], Genetic Algorithms based approach [90]-[93], diagnostics using
nested Artificial Neural Networks [94], diagnostics using Fuzzy Logic [95], Adaptive GPA
[30], diagnostics using non-linear weighted least squares approach [96], diagnostics using
Rough Set theory [97], etc. Apart from the above gas path diagnostics where steady state gas
path measurement data are used, gas path diagnostics using transient measurement data based
on Genetic Algorithms [98] and Artificial Neural Networks [99] have also been developed.
Selection of gas path measurements for gas path diagnostics plays an important role
in order to achieve satisfactory diagnostics results and obtain the best possible observability
of engine health and performance conditions. The importance of the measurement selection
was analyzed in [100] and a measurement selection approach to improve the quality of gas
path diagnostics has been developed [101].
Gas path sensors provide essential performance information of gas turbine engines
and therefore detection and recovery of failed sensors is one of the critical steps in gas path
diagnostics. Different sensor diagnostic approaches have been developed, such as those using
Genetic Algorithms [90], [102], Artificial Neural Networks [94], [103], [104], Influence
Coefficient Matrix [89], etc.
Gas turbine lifing analysis is also an important part of engine condition monitoring.
There are significant amount of research available on engine lifing analysis based on detailed
m material information and using finite element approaches. However, those approaches are
too complicated and very time consuming, which is not convenient for quick estimation of
engine life consumption of gas turbine engines. Therefore, with a focus on creep life, a
concept of Creep Factor [51] was introduced to assess the speed of creep lifing consumption
and unique gas turbine creep life analysis approaches using a model based method [50], [51]
and Artificial Neural Networks [52] with the consideration of varying ambient,
environmental, operation and health condition have been developed.
Unique software for gas turbine thermodynamic performance simulation and gas path
diagnostics, Turbomatch/Pythia [89], [84], has been developed at Cranfield. It is modular
based so performance models for gas turbine engines with different configurations can be
easily set up with excellent flexibility. Special versions of the software have been developed
for industrial partners to meet their special needs in their performance and diagnostic analysis.
Application of the developed gas path diagnostic system and the software to real gas turbine
engines has been carried out and proved to be successful.

3 CBM ON GAS TURBINES APPLICATIONS

In this section a few existing commercial solutions that allow the use of CBM for gas
turbine applications are described, along with some examples that demonstrate the benefits of
incorporating a CBM approach.

10
3.1 EXISTING CBM COMERCIAL TOOLS
In gas turbines industry there are a very limited number of commercial tools available
for CBM applications. Two of them developed by engine manufacturers are briefly described
here.
GE Plant Maintenance Operations Software Bently Nevada System1 is a patented
condition monitoring platform for real-time optimization of equipment and selected
processes, condition monitoring, and event diagnostics of GE gas turbine plants. An extension
of this platform is Bently PERFORMANCE that offers online thermodynamic performance
monitoring of machinery. An additional capability of System 1 is the Predictive Emissions
Monitoring System (PEMS) that is able to predict the level of stack emissions generated by
gas turbines based on ambient conditions, fuel composition and machine operating conditions
while taking into account real time degradations. More details about Bently Nevada System1
and associated solutions and services can be found in GEs measurement and control official
site: http://www.ge-mcs.com/
Siemens Power Diagnostics Services is a group of products providing CBM related
services for gas turbine plants. One of these products includes an automated expert system for
power plant operational data analysis, monitoring, fault diagnosis and prognosis. Other
products in the group provide the capability of Fiber Optic Vibration Monitor (FOVM), the
Blade Vibration Monitoring System (BVM4), and the Radio Frequency Monitor (RFM). A
more detailed description of Siemens Power Diagnostics Services is given in [105].

3.2 BENEFITS FROM INCORPORATING CBM SOLUTIONS

The benefits from incorporating CBM policies into gas turbines plants are many and
include increased safety, asset reliability and availability along with cost-efficient plant
operation. As stated by Boyce in [4], gas turbines operating costs amount to 7080% of the
life cycle costs of the facility. Operating a plant as close as possible to its design conditions
will guarantee that its operating costs will be reduced. For example, the fuel cost for modern
fossil fuel gas turbine power plants ranging from 6002800 MW can be between $72 million
and $168 million per annum. Savings of 13% of these costs, due to efficient condition
monitoring and CBM, can result in an overall cost reduction of up to $1 million per annum.
In addition, application of CBM brings the following benefits to gas turbine users.
From maintenances point of view, condition monitoring provided by CBM allows better
maintenance planning due to being able to order spare parts and arrange personnel and
necessary maintenance equipment in well advance, which would result in improved
availability and reduced operating costs. From gas turbine users point of view, since a CBM
program estimates the confidence level of plant operation, it allows also changing operating
profile based on the actual health condition of the engines. If, for instance, a fault at an early
stage has been detected and the development of this fault is accelerated later at specific
operating conditions, it becomes possible to ensure safe operation of the gas turbine before
the next scheduled maintenance by operating the engine at different operating points or at
lower power level.
A few success stories of incorporating CBM policies and related diagnostic
approaches into gas turbine plants are reported in [106] - [108].

11
4 CONCLUSIONS AND SUGGESTIONS FOR FUTURE RESEARCH
In this paper an overview of the concept, the technology and the importance of CBM
in the operation of modern gas turbine plants have been reviewed. The details of CBM and
related diagnostic, prognostic and lifing methods and techniques are also provided. In
addition, research work in this area contributed by the Laboratory of Thermal Turbomachines
of the National Technical University of Athens (LTT/NTUA) and the Department of Power
and Propulsion of Cranfield University are described. Furthermore, some commercial
applications of the CBM technology have also been presented.
This paper has revealed that the degree of penetration of CBM policies in the gas
turbines market is disproportional to the potential importance and benefits of CBM. Future
research should focus on bring advanced CBM technologies into gas turbine industry in order
to benefit gas turbines users. With that in mind, the following research topics are suggested:

o Further development/integration of gas turbine diagnostic approaches. Although a

large number of diagnostic methods have been proposed by researchers, each method
is applicable on a limited range of underlying health conditions. Integration of
methods into a diagnostic tool being capable of addressing the entire diagnostic
problem is required.
o Development of efficient prognostic approaches. Gas turbine prognostic approaches
should be further developed and evolve into tools covering all important aspects of
remaining useful life of gas turbine assets (i.e. safety, reliability, availability,
maintenance and operating cost).
o Development of gas turbine lifing approaches. Such approaches may include
prediction of gas turbine engine life, including creep life and low cycle fatigue life, as
a function of ambient, operating and health conditions of gas turbine engines. Such
capability will enhance the effectiveness of condition based maintenance.
o Development of efficient CBM validation approaches. Without CBM performance
metrics, qualitative and quantitative evaluation of any CBM policy or planning will
be subjective and its effectiveness over existing maintenance practices cannot be
justified.

12
5 REFERENCES
[1] Jardine, Lin, Banjevic, 2006, A review on machinery diagnostics and prognostics
implementing condition-based maintenance, Mechanical Systems and Signal
Processing 20 (7): 14831510
[2] Vachtsevanos G., Lewis F.L., Roemer M., Hess A., Wu B., 2006, Intelligent
Fault Diagnosis and Prognosis for Engineering Systems, John Wiley, New York,
2006.
[3] Marinai L., Probert D., Singh R., 2004, Prospects for aero gas-turbine
diagnostics: a review, Applied Energy, vol.79, pp.109126
[4] Boyce M.P., 2006, Gas Turbine Engineering Handbook-3rd Edition, Gulf
Professional Publishing, Linacre House, Jordan Hill, Oxford, UK
[5] Kelly R., 1996, Application of analytical redundancy to detection of sensor faults
on a turbofan engine, ASME Turbo Expo, 96-GT-3
[6] Caliskan F., Hajiyev C., 2000, Innovation sequence application to aircraft sensor
fault detection: comparison of checking covariance matrix algorithms, ISA
Transaction 39, p.47-56
[7] Healy T., Kerr L., Larkin L., 1997, Model based fuzzy logic sensor fault
accommodation, ASME Turbo Expo, 97-GT-222
[8] Kanelopoulos K., Stamatis A., Mathioudakis K., 1997, Incorporating neural
networks into gas turbine performance diagnostics, ASME Turbo Expo, 97-GT-
35
[9] Napolitano M., An Y., Seanor B., 2000, A fault tolerant flight control system for
sensor and actuator failures using neural networks, Pergamon editions, Aircraft
design 3, p. 103-128
[10] Mattern D., Guo T., Graham R., McCoy W., 1998, Using neural networks for
sensor validation, prepared for the 34th joint propulsion conference, NASA/TM-
1998-208483
[11] Doke R., Singh R., 1999, Neural networks for the detection of gas turbine sensor
faults, ISABE paper, ISABE 99-7255
[12] Romessis C., Mathioudakis K., 2003, Setting Up Of a Probabilistic Neural
Network for Sensor Fault Detection Including Operation with Component Faults,
ASME Journal of Engineering for Gas Turbines and Power, Vol 125, No. 3, July
2003, pp. 634-641
[13] Mathioudakis, K., Kamboukos, Ph., and Stamatis, A., 2002, Turbofan
performance deterioration tracking using non-linear models and optimization
techniques, ASME Turbo Expo, GT-2002-30026
[14] Fodor I.K., 2002, A Survey of Dimension Reduction Techniques, Lawrence
Livermore National Laboratory (LLNL) Technical Report UCRL-ID-148494,
University of California, Livermore, CA
[15] Garga A.K., Elverson B.T., Lang D.C., 1997, AR modeling with dimension
reduction for machinery fault classification, in Critical Link: Diagnosis to
Prognosis, Haymarket, 1997, pp. 299308
[16] Zhan Y., Makis V., Jardine A.K.S., 2003, Adaptive model for vibration
monitoring of rotating machinery subject to random deterioration, Journal of
Quality in Maintenance Engineering 9 (2003) 351375.
[17] Goldman S., 1999, Vibration Spectrum Analysis: A Practical Approach,
Industrial Press, New York.
[18] Harris C.M., Piersol A.G., 2002, Harris Shock and Vibration Handbook,
McGraw-Hill, New York.
13
 [19] Peng Z. K., Chu F. L., 2004, Application of the Wavelet Transform in Machine
Condition Monitoring and Fault Diagnosis: A Review with Bibliography,
Mechanical Systems and Signal Processing 18, pp.199221.
[20] Harrison G.A, Koren I., Lewis M.P., Taylor F.J., 1998, Application of wavelet
and Wigner analysis to gas turbine vibration signal processing, Proc. SPIE 3391,
Wavelet Applications V, 490 (March 26, 1998), doi:10.1117/12.304898
[21] Aretakis N., Mathioudakis K., 1997, Wavelet Analysis for Gas Turbine Fault
Diagnostics, ASME Journal of Engineering for Gas Turbine and Power, Vol. 119,
No. 4, , October 1997, pp. 870-876
[22] Urban, L. A., Gas Turbine Engine Parameter Interrelationships, HSD UTC,
Windsor Locks, Ct., 1st edition, 1967, 2nd edition, 1969.
[23] Urban L.A., Volponi A.J., 1992, Mathematical Methods of Relative Engine
Performance Diagnostics, S.A.E. technical paper 922048, Journal of Aerospace
Transactions vol.101.
[24] Volponi, A.J., 1982, 'Gas Path Analysis: An Approach to Engine Diagnostics', 35th
Symposium Mechanical Failures Prevention Group, 20-22 April 1982,
Gaitherburg, MD, USA
[25] Doel, D., 1993, Gas path analysis - Problems and solutions, 17th Symposium of
Aircraft Integrated Monitoring Systems, Bonn, Germany. 21-23 September 1993.
[26] Doel. D., 1994, TEMPER a Gas Path Analysis Tool for Commercial Jet
Engines, Transaction of the ASME, Journal of Engineering for Gas Turbines and
Power, Jan.1994, Vol. 116. pp.82-89
[27] Stamatis A., Mathioudakis K., Papailiou K.D., 1990, Adaptive Simulation of Gas
Turbine Performance, Journal of Engineering for Gas Turbine and Power, April
1990. vol.112, pp.168-175
[28] Escher P.C., Singh R., 1995, An Object-Oriented Diagnostics Computer Program
Suitable for Industrial Gas Turbines, United 21st International Congress of
Combustion Engines (CIMAC), Interlaken, Switzerland, 15-18, May 1995.
[29] Aretakis N., Mathioudakis K., Stamatis A., 2004, Identification of Sensor Faults
on Turbofan Engines Using Pattern Recognition Techniques, Control
Engineering Practice, Vol. 12, No. 7, July 2004, pp. 827-836
[30] Li Y.G., 2010, Gas Turbine Performance and Health Status Estimation Using
Adaptive Gas Path Analysis, ASME Journal of Engineering for Gas Turbines and
Power, Vol. 132, No. 4, April 2010.
[31] Dewallef P., Leonard O., 2003, On-line performance monitoring and engine
diagnostic using robust kalman filtering techniques, ASME Turbo Expo,
GT2003-38379
[32] Bunks C., McCarthy D., Ani T., 2000, Condition-Based Maintenance of
Machines Using Hidden Markov Models, Mechanical Systems and Signal
Processing, Oakland, CA, 14(4):597612, 2000.
[33] Matuck R. G., Barbosa R. J, Bringhenti C., Lima I., 2009, Multiple faults
detection of gas turbine by mlp neural network, ASME Turbo Expo, GT2009-
59964
[34] DePold R., Gass D., 1999, Application of expert systems and neural-networks to
gas-turbine prognostics and diagnostics, J.Eng Gas-Turbines Power, 1999,
pp.121612

14
[35] Volponi, A.J., DePold, H., Ganguli, R., Daguang, C., 2003, The Use of Kalman
Filter and Neural Network Methodologies in Gas Turbine Performance
Diagnostics: A Comparative Study, Journal of Engineering for Gas Turbines and
Power, Vol. 125, pp. 917-924.
[36] Pong J. L., Ming C. Z., Tzu C. H., Jin Z., 2000, An evaluation of engine faults
diagnostics using artificial neural networks, ASME Turbo Expo, 2000-GT-29
[37] Loboda I., Feldshteyn Y., Ponomaryov V., 2011, Neural Networks for Gas
Turbine Fault Identification:Multilayer Perceptron or Radial Basis Network?,
ASME Turbo Expo, GT2011-46752
[38] Yun-Peng C., Shu-Ying L., Shuang Y., Ning-Bo Z., 2012, Fault Diagnosis of a
Gas Turbine Gas Fuel System Using a Self-Organizing Network, Advanced
Science Letters, vol.8, pp.386-392(7)
[39] Ogaji, S.O.T., Singh, R., 2003, Gas Path Fault Diagnoses Framework for a 3-
Shaft Gas Turbine, Proceeding of the Institution of Mechanical Engineers:
Journal of Power and Energy, Volume 217, A3, pp.149-157
[40] Sampath S., Singh R., 2004, An integrated fault diagnostics model using Genetic
Algorithm and Neural networks, ASME Turbo Expo, GT2004-53914
[41] Romessis C., Mathioudakis K., 2006, Bayesian Network Approach for Gas Path
Fault Diagnosis, ASME Journal of Engineering for Gas Turbines and Power,
Vol. 128, No. 1, January 2006, pp. 64-72
[42] Eustace R.W., 2008, A Real-World Application of Fuzzy Logic and Influence
Coefficients for Gas Turbine Performance Diagnostics, ASME Turbo Expo,
GT2007-27442
[43] Li Y. G., Nilkitsaranont P., 2007, A Gas Path Diagnostic and Prognostic
Approach for Gas Turbine Applications, ASME Turbo Expo, GT2007-27275
[44] Pascovici, D.S., Kyprianidis K. G., Colmenares F., Ogaji, S.O.T., Pilidis, P, 2008,
Weibull Distributions Applied to Cost and Risk Analysis for Aero Engines,
ASME Turbo Expo, GT2008-51060
[45] Lipowsky H., Staudacher S., Bauer M., Schmidt K.J, 2009, Application of
Bayesian Forecasting to Change Detection and Prognosis of Gas Turbine
Performance, ASME Turbo Expo, GT2009-59447
[46] Studer L., Masulli F., 1996, On the structure of a neuro-fuzzy system to forecast
chaotic time series, International Symposium on Neuro-Fuzzy Systems, 29-31
Aug. 1996, pp. 103 110.
[47] Depold H., Siegel J., 2006, Using diagnostics and prognostics to minimize the
cost of ownership of gas turbines, ASME Turbo Expo, GT2006-91183
[48] Pascovici, D.S., Colmenares, F., Ogaji, S.O.T., Pilidis, P., 2007, An Economic
and Risk Analysis Model for Aircrafts and Engines, ASME Turbo Expo, GT2007-
27236
[49] Tinga T., Visser W.P.J., Wolf W.B., 2000, Integrated lifing analysis for gas
turbine components, National Aerospace Laboratory, Amsterdam, Tech. Rep.
NLR-TP-2000-632, 2000.
[50] Naeem M., Singh R., Probert D., 1998, Implications of engine deterioration for
creep life, Applied Energy, vol. 60, pp. 183-223, 8/1
[51] Abdul Ghafir M.F., Li Y.G., Singh R., Huang K., Feng X., 2010, Impact of
Operating and Health Conditions on Aero Gas Turbine Hot Section Creep Life
Using a Creep Factor Approach, ASME GT2010-22332, ASME Turbo Expo
2010, Glasgow, United Kingdom, June 2010

15
 [52] Abdul Ghafir M.F., Li Y.G., Wang L., 2012, Creep Life Prediction for Aero Gas
Turbine Hot Section Component Using Artificial Neural Network Approach,
ASME GT2012-68856, ASME Turbo Expo 2012, Copenhagen, Denmark, 11-15
June 2012.
[53] Kyriazis A., Tsalavoutas A., Mathioudakis K., Bauer M., Johanssen O., 2009,
Gas Turbine Fault Identification by Fusing Vibration Trending and Gas Path
Analysis, ASME Turbo Expo GT2009-59942
[54] Romessis C., Kyriazis A., Mathioudakis K., 2007, Fusion of Gas Turbines
Diagnostic Inference: The Dempster-Schafer Approach, ASME Turbo Expo,
GT2007-27043
[55] Dewallef P., Romessis C., Leonard O., Mathioudakis K., 2006, Combining
Classification Techniques With Kalman Filters For Aircraft Engine Diagnostics,
ASME Journal of Engineering for Gas Turbines and Power, Vol. 128, No. 2, April
2006, pp. 281-287
[56] Orsagh R.F., Sheldon J., Klenke C.J., 2003, Prognostics/Diagnostics for Gas
Turbine Engine Bearings, ASME Turbo Expo, No. GT-2003-38075
[57] Volponi A.J., Brotherton T., Luppold R., Simon D.L., 2004, Development of an
information fusion system for engine diagnostics and health management,
NASA/TM-2004-12924
[58] Garbiras M., Goebel K., 2000, Fusing Diagnostic Information Without A-priori
Performance Knowledge, proceedings of the 3rd Int. Conference on Information
Fusion, Fusion 2000,1, pp.TuD5-9-TuD5-16
[59] Goebel K., Agogino A., 2001, Fuzzy Sensor Validation and Fusion for Gas
Turbine Power Plants, Technical Information series, 2001CRD033
[60] Kamboukos Ph., Mathioudakis K., 2005, Comparison of Linear And Non-Linear
Gas Turbine Performance Diagnostics, ASME Journal of Engineering for Gas
Turbine and Power, Vol. 127, No. 1, January 2005 pp 49-56
[61] DePold H., Siegel J., Hull J., 2004, Metrics for evaluating the accuracy of
diagnostic fault detection systems, ASME Turbo Expo, GT2004-54144
[62] Curnock B., 2000, Obidicote project - word package 4: Steady state test cases,
Technical Report DNS62433, Rolls-Royce
[63] Simon D.L., Bird J., Davison C., Volponi A., Iverson R.E., 2008, Benchmarking
Gas Path diagnostic methods: A public approach, ASME Turbo Expo, GT2008-
51360
[64] Kamboukos Ph., Mathioudakis K., 2006, Multipoint Non-Linear Method for
Enhanced Component and Sensor Malfunction Diagnosis, ASME Turbo Expo,
GT2006-90451
[65] Stamatis A., Mathioudakis K., Ruiz J., Curnock B., 2001, Real Time Engine
Model Implementation For Adaptive Control & Performance Monitoring of Large
Civil Turbofans, ASME Turbo Expo, 2001-GT-0362
[66] Stamatis A., Mathioudakis K., Smith M.K., Papailiou K.D., 1990, Gas Turbine
Component Fault Identification by Means of Adaptive Performance Modelling,
ASME Turbo Expo, 90-GT-376
[67] Aretakis N., Mathioudakis K., Stamatis A., 2003, Non-Linear Engine Component
Fault Diagnosis From A Limited Number Of Measurements Using A
Combinatorial Approach, ASME Journal of Engineering for Gas Turbine and
Power, Vol. 125, No. 3, July 2003, pp. 642-650

16
[68] Kamboukos Ph., Stamatis A., Mathioudakis K., 2004, Gas Turbine Component
Fault Detection from a Limited Number of Measurements, Proceedings Of The
Institution of Mechanical Engineers, PART A, Journal of Power and Energy, Vol.
218, No. A8, Dec 2004, pp. 609-618
[69] Aretakis N., Roumeliotis I., Mathioudakis K., 2011, Performance Model
Zooming For In-Depth Component Fault Diagnosis, ASME Journal of
Engineering for Gas Turbines and Power, March 2011, Vol. 133, No. 3, 031602-1
[70] Aretakis N., Mathioudakis K., 1998, Classification of Radial Compressor Faults
Using Pattern Recognition Techniques, Control Engineering Practice, Vol. 6, No.
10, October 1998, pp. 1217-1223
[71] Kyriazis A., Aretakis N., Mathioudakis K., 2006, Gas Turbine Fault Diagnosis
From Fast Response Data Using Probabilistic Methods and Information Fusion,
ASME Turbo Expo, GT2006-90362
[72] Romessis C, Stamatis A., Mathioudakis K., 2001, A Parametric Investigation of
the Diagnostic Ability of Probabilistic Neural Networks on Turbofan Engines,
ASME Turbo Expo, 2001-GT-0011
[73] Kyriazis A., Helmis I., Aretakis N., Roumeliotis I., Mathioudakis K., 2011, Gas
Turbines Compressor Fault Identification by Utilizing Fuzzy Logic-Based
Diagnostic Systems,9th European Turbomachinery Conference proceedings,
paper 121, March 21-25, 2011, Istanbul Turkey
[74] Romessis C., Mathioudakis K., 2007, Detection of Gas Turbines Malfunctions
From Emission Concentration Distributions, ASME Turbo Expo, GT2007-2710
[75] Romessis C., Mathioudakis K., 2009, Estimation of Gas Turbines Gradual
Deterioration Through a Dempster-Schafer based Fusion Method, ISABE paper
2009-1301
[76] Kyriazis A., Mathioudakis K., 2009, Enhanced Fault Localization Using
Probabilistic Fusion With Gas Path Analysis Algorithms, ASME Journal of
Engineering for Gas Turbines and Power, September 2009, Vol. 131, No. 5,
051601
[77] Kyriazis A., Mathioudakis K., 2009, Gas Turbine Fault Diagnosis Using Fuzzy-
based Decision Fusion, AIAA Journal of Propulsion and Power, Vol. 25, No. 2,
MarchApril 2009, pp. 335-343
[78] Aretakis N., Roumeliotis I., Doumouras G., Mathioudakis K., 2012, Compressor
Washing Economic Analysis and Optimization for Power Generation, Applied
Energy 95 (2012) 7786
[79] Tsalavoutas A., Mathioudakis K., Stamatis A., Smith M., 2000, Identifying Faults
in the Variable Geometry System of a Gas Turbine Compressor, ASME Turbo
Expo, 2000-GT-0033
[80] Tsalavoutas A., Aretakis N., Mathioudakis K., Stamatis A., 2000, Combining
Advanced Data Analysis Methods for the Constitution of an Integrated Gas
Turbine Condition Monitoring and Diagnostic System, ASME Turbo Expo, 2000-
GT-0034
[81] UK Patent for: Sensor Fault Diagnosis II GBA0126298 (30 August 2002)
[82] US Patent for: Sensor Fault Diagnosis II US Application No: 10/228,213
[83] UK Patent for: Fault Diagnosis UK Application number: 0301707.6Applied for
24th January 2003

17
 [84] Macmillan W. L., 1974, Development of a Modular Type Computer Program for
the Calculation of Gas Turbine Off Design Performance, Ph.D. Thesis, Cranfield
University, September 1974.
[85] Li Y.G., Pilidis P., Newby M.A., 2005, An Adaptation Approach for Gas Turbine
Design-Point Performance Simulation, ASME GT2005-68140, 2005. 11
[86] Li Y.G., Marinai L., Lo Gatto E., Pachidis V. and Pilidis P., 2009, Multiple-Point
Adaptive Performance Simulation Tuned to Test-Bed Data, AIAA Journal of
Propulsion and Power, Vol.25, No. 3, May-June 2009.
[87] Li, Y.G., Abdul Ghafir M.F., Wang L., Singh R., Huang K., Feng X., Zhang W.,
2012, Improved Multiple Point Non-Linear Genetic Algorithm Based
Performance Adaptation Using Least Square Method, ASME Journal of
Engineering for Gas Turbines and Power, Vol. 134, pp.031701, March 2012.
[88] Escher P.C., Singh R., 1995, An Object-Oriented Diagnostics Computer Program
Suitable for Industrial Gas Turbines, 21th (CIMAC) International Congress of
Combustion Engines, Switzerland, 15-18 May 1995.
[89] Li Y.G., Singh R., 2005, An Advanced Gas Turbine Gas Path Diagnostic System
PYTHIA, ISABE-2005-1284, the Seventeenth International Symposium on
Airbreathing Engines (17th ISABE), Munich, Germany, September 2005.
[90] Zedda M., Singh R., 2002, Gas Turbine Engine and Sensor Fault Diagnosis
Using Optimization Techniques, AIAA Journal of Propulsion & Power, Sept-Oct
2002, JPPOEL 18(5) 969-1128 (2002) Pages 1019-1026.
[91] Zedda, M., Singh, R., 1999, Gas Turbine Engine and Sensor Fault Diagnosis
Using Optimisation Techniques, AIAA-99-2530, 1999
[92] Gulati A, Zedda M, Singh R., 2000, Gas Turbine Engine and Sensor Multiple
Operating Point Analysis Using Optimisation Techniques, AIAA/ASME Joint
Propulsion Conference, Alabama, USA, July 2000. 13.
[93] Sampath, S., Singh. R, 2002, Engine Fault Diagnosis-An Optimisation
Procedure, Applied Energy, Vol 73, Issue 1 (Sept 2002) Pg 47-70
[94] Ogaji. S.O.T, Singh R., Probert S.D., 2002, Multiple-sensor fault-diagnoses for a
2-shaft stationary gas-turbine, Applied Energy 71 (2002) 321-339, February
2002.
[95] Marinai, L, Singh, R, Curnock, B, Probert D., 2003, Fuzzy-Logic-Based
Diagnostic Process for Turbofan Engines, 16th International Symposium of Air
Breathing Engines, Cleveland, Ohio September 2003
[96] Li Y.G., Korakianitis T., 2011, A Non-Linear Weighted-Least-Square (NLWLS)
Estimation Approach for Gas Turbine Diagnostic Applications, AIAA Journal of
Propulsion and Power, Vol. 27, No.2, March-April 2011.
[97] Wang L., Li Y.G., Abdul Ghafir M.F., Swingler A., 2011, A Rough Set Based
Gas Turbine Fault Classification Approach Using Enhanced Fault Signatures,
IMechE Journal of Power and Energy, Vol.225, pp.1052-1064, 2011.
[98] Li Y.G., 2003, A Gas Turbine Diagnostic Approach with Transient
Measurements, I.Mech.E Proceedings of the Institute of Mechanical Engineering,
Part A: Journal of Power and Energy, Vol.217, No. A2
[99] Ogaji S.O.T., Li Y.G., Sampath S., Singh R., 2003, Gas Path Fault Diagnosis of
a Turbofan Engine from Transient Data Using Artificial Neural Networks, ASME
Paper GT2003-38423, ASME TURBO EXPO 2003, Atlanta, Georgia, USA, June
16-19, 2003.
[100] Ogaji S.O.T., Singh R., 2002, Study of the Optimisation of Measurement Sets for
Gas Path Fault Diagnosis in Gas Turbines, ASME Turbo Expo, 3-6 June 2002,
Amsterdam, The Netherlands. GT- 2002-30050
[101] Jasmani M.S., Li Y.G., 2011, Measurement Selection for Multi-Component Gas
Path Diagnostics Using Analytical Approach and Measurement Subset Concept,
ASME Journal of Engineering for Gas Turbines and Power, Vol. 133, pp.111701,

18
 November 2011.
[102] Sampath S., Singh R., 2003, Evolution Strategy for Engine and Sensor Fault
Diagnosis, 16th International Symposium of Air Breathing Engines, Cleveland,
Ohio September 2003
[103] Zedda M., Singh R., 2001, Neural-network-based sensor validation for gas
turbine test bed analysis Proceedings Institution of Mechanical Engineers, Vol.
215 Part I, 2001.
[104] Xiradakis N., Li Y.G., 2004, Gas Turbine and Sensor Fault Diagnosis with a
Nested Artificial Neiral Networks, ASME TURBO EXPO 2004, GT-2004-53570
[105] Plotts K. and Diatzikis E., 2009, A Survey of New Technologies Used by Siemens
Energy for the Monitoring and Diagnosis of a Global Fleet of Power Generation
Systems, ASME Turbo Expo, GT2009-59967
[106] Customer Success Story, Bently Nevada Asset Condition Monitoring,
http://www.ge-mcs.com/download/case-studies/bently-nevada/GEA17148.pdf
[107] Customer Success Story, Bently Nevada Asset Condition Monitoring,
http://www.ge-mcs.com/download/case-studies/bently-nevada/GEA17062.pdf
[108] Customer Success Story, Bently Nevada Asset Condition Monitoring: Intelligent
Diagnostic Capabilities Avert a Costly Outage, http://www.ge-
mcs.com/download/case-studies/bently-nevada/GEA17169.pdf

19