This paper introduces a data-driven framework for combustor-focused, performance-based condition monitoring of gas turbines. Commercial condition monitoring systems typically generate huge amounts of data that make efficient onboard monitoring challenging. This paper focuses on quantifying combustor component degradation, using premixer centerbody degradation in a swirl stabilized combustor as a case study. The input for these analyses is acoustic pressure measurements acquired at various locations on the combustor. The diagnosis methodology is based on a classification framework and consists of 3 steps: 1) Data curation, 2) Feature Engineering, and 3) Diagnosis. Data curation ensures good quality of the data that is passed through the algorithm. Feature engineering deals with the extraction of the most informative features, from the most informative sensors, that can accurately capture the introduced fault. To perform diagnosis, the classification model is trained using experimentally acquired data and is then tested on a separate data set. The framework was able to achieve high classification accuracy (>99%) for training size as low as 30% of the total recorded observations. The low number of features required to achieve this accuracy suggests high potential for integration into existing onboard condition monitoring systems.

Since the 1960’s the maintenance concepts of gas turbines have remarkably changed. Nowadays, condition monitoring systems supervise modern gas turbines. These systems are capable of avoiding failures that are developing slowly. The actual enhancement of early fault detection is predictive maintenance where big data from different sources is used to predict the behavior of gas turbines. In this paper eight different actual gas turbine failures are presented, beginning with the damage event and the root cause and followed by a discussion, whether the failure could have been avoided by predictive maintenance. In some cases, e.g. during commissioning, a successful early fault detection with predictive maintenance cannot be achieved, since the failure often occurs suddenly and is unforeseen. In others cases, especially during operation, the failure could have been prevented by predictive maintenance. However, it has to be stated, in all these cases significantly more measuring devices would have been necessary. The potential exists that more failures can be avoided by predictive maintenance in the future. However, there are also damage scenarios that cannot be prevented by any early fault detection system. Thus, the Possible Maximum Loss, an important figure for technical insurances, remains unchanged independent of the application of predictive maintenance.

In the last decade, the development of machine connectivity has made possible early fault detection with remote analysis of operating data. Solutions aiming to reduce maintenance costs and production losses due to unplanned downtimes were brought to market. These solutions provide with a model of the equipment in healthy conditions using machine learning techniques applied on historical data. During operation a warning is issued when expected and actual measurements do not match. Although these solutions have proven their value to detect abnormal behaviors, they generate a large number of alarms that require resource to be analyzed. Moreover, these solutions rely on a large number of sensors that need to work correctly both for the learning and the monitoring phase. This generates additional maintenance even though these sensors are often not essential to operate the machine. Lastly the solutions are expensive: their application is usually limited to critical machines with risks of production loss. Indeed, they are not economic for a Transmission System Operator that has ensured the availability of its network with redundancy. The objective of the authors was to focus on the monitoring of radial vibrations of centrifugal compressors. Experience proves this is one of the most critical data for early fault detection. The goal was to develop a smart modelling based on historical data using essential parameters influencing rotor-dynamics. As a result, a clear correlation was found between the operating point and the vibration level. That can be easily shown on a centrifugal compressor map. A second-degree polynomial equation was successfully tested. The model equation relies only on two compressor physics parameters: flow coefficient and speed. We discuss in the paper the impact of other essential parameters. The method has been applied on different type of centrifugal compressors, with different bearing technology (magnetic...) or shaft driving equipment (gas turbine, electric motor drive). A fault detection case study using this method is described, eg: vibration variation due to abnormal opening of an anti-surge control valve. In conclusion this method is a simple alternative to usual condition monitoring solutions. Similarly to what was described in the GT2014-25242 for a Predictive Emission Monitoring System [1], equations based on physical parameters prove to be an efficient modelling technique. Moreover, it helps monitoring teams to better understand the underlying relation between parameters. Indeed, to achieve a complete monitoring of a centrifugal compressor health, this method can be combined with first-principle performance models that use the same physical parameters.

This paper presents the development of an artificial neural network (ANN) Gas Path Diagnostics (GPD) technique applied to pipeline compression system for fault detection and quantification. The work detailed the various degradation mechanisms and the effect of such degradations on the performance of natural gas compressors. The data used in demonstrating the ANN diagnostics is so derived using an advanced thermodynamic performance simulation model of integrated pipeline and compressor systems, which has embedded empirical compressor map data and pipeline resistance model. Implantation of faults within the model is in such a way to account for faults degradations caused by fouling, erosion and corrosion, of various degrees of severities, to obtain wide range of corresponding simulated “true” measurements. In order to account for uncertainties normally encountered in field measurements, Gaussian noise distribution was combined with simulated true measurements, which depends on the instrument’s tolerances. Furthermore, since judicious measurements selection are crucial in ensuring flawless GPD predictions, a sensitivity and correlation analysis of the available measurements revealed that discharge temperature, rotational speed and torque are the most effective measurements for the diagnostics with acceptable degrees of accuracies. The measurements observability technique is a novel approach in pipeline compressor diagnostics. Analytical case studies of the developed method show that, a selected ANN architecture can detect and quantify faults related to degradation in efficiency and flow capacities in the presence of instrument error, varied operational and environmental conditions.

The core of a helicopter drivetrain is a complex planetary main gearbox (MGB) which reduces the high input speed generated by the engines in order to provide the appropriate torque to the main rotors and to other auxiliary systems. The gearbox consists of various shafts, planetary gears and bearings and operates under varying conditions under excessive friction, heat and high mechanical forces. The components are vulnerable to fatigue defects and therefore Health and Usage Monitoring Systems (HUMS) have been developed in order to monitor the health condition of the gearbox, focusing towards early, accurate and on time fault detection with limited false alarms and missed detections. The main aim of a HUM System is by health monitoring to enhance the helicopters’ operational reliability, to support the maintenance decision making, and to reduce the overall maintenance costs. The importance and the need for more advanced and accurate HUMS have been emphasized recently by the post-accident analysis of the helicopter LN-OJF, which crashed in Norway in 2016. During the last few decades various methodologies and diagnostic indicators/features have been proposed for the monitoring of rotating machinery operating under steady conditions but still there is no global solution for complex structures. A new tool called IESFOgram has been recently proposed by the authors, based on Cyclostationary Analysis, focusing on the accurate selection of a filtering band, under steady and varying speed conditions. Moreover the Cyclic Spectral Coherence is integrated along the selected frequency band leading to an Improved Envelope Spectrum. In this paper the performance of the tool is tested on a complex planetary gearbox, with several vibration sources. The method is tested, evaluated and compared to state of the art methods on a dataset captured during experimental tests under various operating conditions on a Category A Super Puma SA330 main planetary gearbox, presenting seeded bearing defects of different sizes.

Gas turbine diagnostics that relies on gas path measurements is a well-developed area with many algorithms developed. They follow two general approaches, data-driven, and physics-based. The first approach uses deviations of monitored variables from their baseline values. A diagnostic decision is traditionally made in the space of these deviations (diagnostic features) by pattern recognition techniques, for example, artificial neural networks. The necessary fault classes can be constructed from deviation vectors (patterns) using the displays of real faults, and the approach has a theoretical possibility to exclude a complex physics-based model and its inherent errors from a diagnostic process. For the second approach known as a gas path analysis, a nonlinear physics-based model (a.k.a. thermodynamic model) is an integral part of a diagnostic process. The thermodynamic model (or the corresponding linear model) relates monitored variables with operational conditions and model’s internal quantities called fault parameters. The identification of the thermodynamic model on the basis of known measurements of the monitored variables and operational conditions allows estimating unknown fault parameters. The knowledge of these parameters drastically simplifies a final diagnostic decision because great values of these parameters indicate damaged engine components and give us the measure of damage severity. As the diagnostic decision seems to be simple, the studies following this approach are usually completed by the analysis of fault parameter estimation accuracy, and complex pattern recognition techniques are not employed. Instead, simple tolerance-based fault detection and isolation is sometimes performed. It is not clear from known comparative studies which of the two approaches is more accurate, and the issue of seems to be challenging. This paper tries to solve this problem, being grounded on the following principles. We consider that a key difference of the second approach is a transformation from the diagnostic space of the deviations of monitored variables to the space of fault parameters. To evaluate the influence of this transformation on diagnostic accuracy, the other steps of the approaches should be equal. To this end, the pattern recognition technique employed in the data-driven approach is also included in the physics-based approach where it is applied to recognize fault parameter patterns instead of a tolerance-based rule. To realize and compare the data-driven and modified physics-based approaches, two corresponding diagnostic procedures differing only by the mentioned transformation have been developed. They use the same set of deviation vectors of healthy and faulty engines as input data and finally compute true classification rates that are employed to compare the procedures. The results obtained for different cases of the present comparative study show that the classification rates are practically the same for these procedures, and this is true for both fault detection and fault isolation. That is, correct classification does not depend on the mentioned transformation, and both approaches are equal from the standpoint of the classification accuracy of engine states.

The ability to trend engine module performance and provide engine system fault detection and isolation are arguably, core capabilities for any engine Condition Based Maintenance (CBM) system. The origins of on-condition monitoring can be traced back nearly four decades, and a methodology known as Gas Path Analysis (GPA) has played a pivotal role in its evolution. Legacy Gas Path Analysis is a general methodology that assesses and quantifies changes in the underlying performance of the major modules of the engine (compressors and turbines), which in turn, directly affect overall performance measures of interest such as fuel consumption, power availability, compressor surge margins, and the like. Additionally, this approach is easily adapted to enable fault detection and identification of many engine system accessory faults such as variable stator vanes, handling and customer bleeds, sensor biases, and drift. Classical GPA has been confined to off-board analysis of averaged snapshot data when the engine is in steady state operation. This discrete data point approach, while reasonably accurate and repeatable, introduces a time latency to confidently detect and identify a faulty condition. Depending on the type and severity of the underlying fault, time to identify can be the differentiating factor in avoiding an unanticipated engine removal, flight delay or cancellation, in-flight engine shutdown, or catastrophic event. In this paper, we explore the use of streaming full flight data, which includes both transient and steady state operation. This type of data stream, when properly processed by gas path analysis and information fusion algorithms, allows faster anomaly detection, credible fault persistency checks, and timely fault identification within the current flight.

Sensor fault detection and classification is a key challenge for machine monitoring and diagnostics. To this purpose, a comprehensive approach for Detection, Classification and Integrated Diagnostics of Gas Turbine Sensors (named DCIDS), previously developed by the authors, is improved in this paper to detect and classify different fault classes. For a single sensor or redundant/correlated sensors, the improved diagnostic tool, called I-DCIDS, can identify seven classes of fault, i.e. out of range, stuck signal, dithering, standard deviation, trend coherence, spike and bias. Fault detection is performed by means of basic mathematical laws that require some user-defined input parameters, i.e. acceptability thresholds and windows of observation. This paper presents in detail the I-DCIDS methodology for sensor fault detection and classification. Moreover, this paper reports some examples of application of the methodology to simulated data to highlight its capability to detect sensor faults which can be commonly encountered in field applications.

Digitally enhanced services for wind power could reduce Operations and Maintenance (O&M) costs as well as the Levelised Cost Of Energy (LCOE). Therefore, there is a continuous need for advanced monitoring techniques which can exploit the opportunities of Internet of Things (IoT) and Big Data Analytics, revolutionizing the future of the energy sector. The heart of wind turbines is a rather complex epicyclic gearbox. Among others, extremely critical gearbox components which are often responsible for machinery stops are the rolling element bearings. The vibration signatures of bearings are rather weak compared to other components, such as gears, and as a result an extended number of signal processing techniques and tools have been proposed during the last decades, focusing towards accurate, early, and on time bearing fault detection with limited false alarms and missed detections. Envelope Analysis is one of the most important methodologies, where an envelope of the vibration signal is estimated usually after filtering around a frequency band excited by impacts due to the bearing faults. Different tools, such as Kurtogram, have been proposed in order to accurately select the optimum filter parameters (center frequency and bandwidth). Cyclic Spectral Correlation and Cyclic Spectral Coherence, based on Cyclostationary Analysis, have been proved as very powerful tools for condition monitoring. The monitoring techniques seem to have reached a mature level in case a machinery operates under steady speed and load. On the other hand, in case the operating conditions change, it is still unclear whether the change of the monitoring indicators is due to the change of the health of the machinery or due to the change of the operating parameters. Recently, the authors have proposed a new tool called IESFOgram, which is based on Cyclic Spectral Coherence and can automatically select the filtering band. Furthermore, the Cyclic Spectral Coherence is integrated along the selected frequency band leading to an Improved Envelope Spectrum. In this paper, the performance of the tool is evaluated and further extended on cases operating under different speeds and different loads. The effectiveness of the methodology is tested and validated on the National Renewable Energy Laboratory (NREL) wind turbine gearbox vibration condition monitoring benchmarking data set which includes various faults with different levels of diagnostic complexity as well as various speed and load operating conditions.

Sensor fault detection is a crucial aspect for raw data cleaning in gas turbine industry. To this purpose, a comprehensive approach for Improved Detection, Classification and Integrated Diagnostics of Gas Turbine Sensors (named I-DCIDS) was developed by the authors to detect and classify several classes of fault. For single-sensors or redundant/correlated sensors, the I-DCIDS methodology can identify seven classes of fault, i.e. Out of Range, Stuck Signal, Dithering, Standard Deviation, Trend Coherence, Spike and Bias. Since the considered faults are detected by means of a methodology which relies on basic mathematical laws and user-defined parameters, sensitivity analyses are carried out in this paper on I-DCIDS parameters to derive some rules of thumbs about their optimal setting. The sensitivity analyses are carried out on four heterogeneous and challenging datasets with redundant sensors installed on Siemens gas turbines.

In an effort to better compare particular gas turbine diagnostic solutions and recommend the best solution, the software tool called Propulsion Diagnostic Method Evaluation Strategy (ProDiMES) has been developed. This benchmarking platform includes a simulator of the aircraft engine fleet with healthy and faulty engines. The platform presents a public approach, at which different investigators can verify and compare their algorithms for the diagnostic stages of feature extraction, fault detection, and fault identification. Using ProDiMES, some different diagnostic solutions have been compared so far. This study presents a new attempt to enhance a gas turbine diagnostic process. A data-driven algorithm that embraces the mentioned three diagnostic stages is verified on the basis of ProDiMES. At the feature extraction stage, this algorithm uses a polynomial model of an engine baseline to compute deviations of actual gas path measurements from the corresponding values of a healthy engine. At the fault detection and fault identification stages, a common classification for fault detection and fault identification is firstly constructed using deviation vectors (patterns). One of the three chosen pattern recognition techniques then performs both fault detection and fault identification as a common process. Numerous numerical experiments have been conducted to select the best configurations of the baseline model, a pertinent structure of the fault classification, and the best recognition technique. The experiments were accompanied by a computational precision analysis for each component of the proposed algorithm. The comparison of the final diagnostic ProDiMES metrics obtained under the selected optimal conditions with the metrics of other diagnostic solutions shows that the proposed algorithm is a promising tool for gas turbine monitoring systems.

Monitoring aircraft performance in a fleet is fundamental to ensure optimal operation and promptly detect anomalies that can increase fuel consumption or compromise flight safety. Accurate failure detection and life prediction methods also result in reduced maintenance costs. The major challenges in fleet monitoring are the great amount of collected data that need to be processed and the variability between engines of the fleet, which requires adaptive models. In this paper, a framework for monitoring, diagnostics, and health management of a fleet of aircrafts is proposed. The framework consists of a multi-level approach: starting from thresholds exceedance monitoring, problematic engines are isolated, on which a fault detection system is then applied. Different methods for fault isolation, identification, and quantification are presented and compared, and the related challenges and opportunities are discussed. This conceptual strategy is tested on fleet data generated through a performance model of a turbofan engine, considering engine-to-engine and flight-to-flight variations and uncertainties in sensor measurements. Limitations of physics-based methods and machine learning techniques are investigated and the needs for fleet diagnostics are highlighted.

This paper presents a two-layer multi-model gas path fault diagnosis method for gas turbines that includes a fault detection layer and a fault isolation layer. A health model and a gas path fault model based on a back propagation neural network are used for the real-time estimation of the output parameters of a gas turbine in the fault detection layer and the output parameter residual in the fault isolation layer, respectively. A fault detection algorithm is proposed based on fuzzy inference, and the fuzzy membership function of the output parameters residual is realized using data statistics. A similarity distance method is used to realize fault isolation, and a fault probability algorithm based on the Mahalanobis distance is presented. Finally, the proposed method is verified by a three-shaft gas turbine simulation platform, and the simulation test results show that the two-layer multi-model gas path fault diagnosis method can detect and isolate the gas path fault accurately with a low calculation cost and good extensibility.

This paper determines the influence of different defective components in the hot-gas path (HGP) of a civil aircraft engine on the density distribution in the exhaust. The intention is to automate the identification of defective components inside the HGP through an analysis of the density distribution in the exhaust jet. The defects include an increased radial gap of the blades in the high-pressure turbine (HPT), and a reduction of the film cooling air mass flow in the first stage of the HPT. In addition, several combinations of both defects are simulated. In the present paper the exhaust density distributions are generated numerically using CFD simulations of the HGP. The density distribution in the exhaust jet is reconstructed with synthetic Background-Oriented Schlieren (BOS) measurements and automatically analyzed. The methodology for the automated defect detection consists of two algorithms, a Support Vector Machine (SVM) algorithm to automatically classify each measurement into a corresponding defect or reference class and an outlier detection algorithm to detect variations from the reference state — without assignment. It is shown that BOS provides a sufficient reconstruction quality to automatically detect defective HGP components with a SVM algorithm. It is possible to automatically detect both defects, even when they occur at the same time. For this purpose, different features were calculated to isolate the influence of each defect on the density distribution. The outlier detection algorithm allows for an automated detection of variations in the density distribution compared to the reference state without any previous knowledge of the influence of the defects on the density distributions during the training procedure. With this algorithm it is possible to detect unknown or new defects which have not been observed or regarded yet. These results strengthen the hypothesis, that an automated detection of defects in jet engines prior to the disassembly is possible.

Sensors provide a means of detecting the actual operating condition of the gas turbine at any point in time. As such, detecting when they exhibit anomalous behaviour is very important to prevent inaccurate performance predictions and costly gas turbine component failures. This paper presents an automated sensor fault detection approach based on the maximal overlap discrete wavelet transform for the detection of the presence of anomalies such as bias, spike, stuck signal, cross-talk and erratic fault in a sensor signal. A multi-sensor validation scheme based on using the median signal as a reference signal within a sensor group is shown to be a robust technique to differentiate sensor anomalies due to gas turbine transient operation from actual sensor faults. The approach presented in this paper lends itself not only suitable for gas turbines in service, but also for gas turbines undergoing acceptance testing. As the timescales involved during a gas turbine acceptance test are just in hours, the conventional approach of trending and setting alarm limits are not sensitive to sensor anomalies which occur within the set alarm limits. Wavelet-based approach combined with a multi-sensor validation scheme provides a viable alternative.

Wind industry experiences a tremendous growth during the last few decades. As of the end of 2016, the worldwide total installed electricity generation capacity from wind power amounted to 486,790 MW, presenting an increase of 12.5% compared to the previous year. Nowadays wind turbine manufacturers tend to adopt new business models proposing total health monitoring services and solutions, using regular inspections or even embedding sensors and health monitoring systems within each unit. Regularly planned or permanent monitoring ensures a continuous power generation and reduce maintenance costs, prompting specific actions when necessary. The core of wind turbine drivetrain is usually a complicated planetary gearbox. One of the main gearbox components which are commonly responsible for the machinery breakdowns are rolling element bearings. The failure signs of an early bearing damage are usually weak compared to other sources of excitation (e.g. gears). Focusing towards the accurate and early bearing fault detection, a plethora of signal processing methods have been proposed including spectral analysis, synchronous averaging and enveloping. Envelope analysis is based on the extraction of the envelope of the signal, after filtering around a frequency band excited by impacts due to the bearing faults. Kurtogram has been proposed and widely used as an automatic methodology for the selection of the filtering band, being on the other hand sensible in outliers. Recently an emerging interest has been focused on modelling rotating machinery signals as cyclostationary, which is a particular class of non-stationary stochastic processes. Cyclic Spectral Correlation and Cyclic Spectral Coherence have been presented as powerful tools for condition monitoring of rolling element bearings, exploiting their cyclostationary behaviour. In this work a new diagnostic tool is introduced based on the integration of the Cyclic Spectral Coherence along a frequency band that contains the diagnostic information. A special procedure is proposed in order to automatically select the filtering band, maximizing the corresponding fault indicators. The effectiveness of the methodology is validated using the National Renewable Energy Laboratory (NREL) wind turbine gearbox vibration condition monitoring benchmarking dataset which includes various faults with different levels of diagnostic complexity.

Micro gas turbines (mGT) are emerging power sources for distributed generation facilities considering their environment friendliness, high fuel flexibility and efficient cogeneration of heat and power (CHP). Numerical simulation and diagnostic tools are essential for cycle optimization of mGT and prediction of performance degradation. This work is focused on fault diagnostics of T100 mGT through the application of AE-T100 model, a simulation tool, which has been developed within a collaboration between the University of Genoa (Unige) and Ansaldo Energia. This model has been developed for steady-state simulation of mGT in off-design conditions. Leveraging on previous efforts for model development, tuning, first phase of validation through Ansaldo Enegia test rig (AE-T100) and diagnostic application of the model, the present work deals with further utilization of the diagnostic capability of the model for mGT cycles. For this purpose, it was used in real operating conditions through the AE-T100 test rig for the diagnostic aim of some unexpected machine behavior. The model results indicated high deviation from the actual field data in terms of fuel flow and efficiency, and so justified the diagnostic capability of the AE-T100 tool. Afterwards, based on the experimental observation of some bypass leakage from the burner in the same test rig, AE-T100 model was applied to model such leakage through the variation of AS ratio, leakage from the recuperator outlet to the ambient, combustor pressure drop and turbine section modification, and carried out the sensitivity analysis of these parameters. Sensitivity analysis has verified that the accurate impact of this leakage on overall mGT performance cannot be modeled with the help of AE-T100 tool in its current capacity. Therefore, some other investigations like analysis of compressor maps must be carried out to explain such a performance deviation in case of leakage. Afterwards, the analysis of compressor maps resulted from the operating conditions of the test conducted on the same machine after leakage repair, has highlighted the change in compressor operating point and thus, more efficient compressor functioning, which results in higher net power. Hence, this analysis has provided a more comprehensible explanation of this leakage impact on the mGT performance.

There exist many approaches for gas turbine engine condition monitoring and fault diagnosis. Among them, gas path analysis depends on the relations between deviations of performance parameters and deviations of measurements, such as pressure, temperature, at some positions in the flow path. A dynamic tracking filter combined with a nonlinear gas turbine model can be used to implement a fault detection system. The dynamic tracking filter is composed with multiple feedback loops in which the residuals between model output and measurement are driven to zero by adjusting the performance parameters. In many cases, the number of measurement parameters is less than that of performance parameters, which impose a limit on the application of tracking filter in practical situations. In the present study, the tracking filter is retrofitted to be driven by the error between state of the model and the estimated state reconstructed from the engine measurement. For the time being, only linear time-invariant (LTI) state observer is considered. A nonlinear simulation model of the heavy-duty gas turbine under study is used to derive the linear system model needed for the design of LTI state observer. Input vector and output vector are chosen according to the practical situations. The linear system model obtained around the nominal operating point is observable, so a state observer can be designed. With 6 state variables observable from the engine, 6 performance parameters can be tracked with the proper design of tracking filter. The structure of the tracking filter consists of a static decoupling matrix DM and in series with 6 PI controllers. Deviations of performance parameters are implanted into the gas turbine model by scaling the performance maps used; and then simulation results are taken as measurements needed for the tracking filter to run. Tracking results of performance parameters in different cases are given to show the tracking capability for isolated performance deviations and concurrent performance deviations.

A vast amount of operating data, control data and log data are generated in the monitoring process of a gas turbine. This massive amount of data includes not only spatial distance features, but also related features. Currently, the analysis of data-related features based on data-driven diagnosis methods is challenging. In response to this limitation, a gas turbine fault diagnosis method based on network community detection is proposed in this paper. First, a gas turbine topology network model is established based on complex network theory. In the network model, gas turbine sampled-data are defined as the nodes; the similarity degree between the sampled-data is defined as the edge of the network; and the reciprocal function is selected as the similarity criterion function. Second, to reflect the community characteristics of the network, the topology network is converted to a training network by a threshold criterion that is designed based on the variation of the average path length and the average clustering coefficient. This method can establish a training network without prior knowledge of the number of clusters. Third, a fault detection algorithm is proposed based on community modularity, and fault reasoning is presented by calculating the probability based on the change in the community modularity which demonstrates the occurrence possibility of multiple faults. The effectiveness of the algorithm is verified by a three-shaft gas turbine fault simulation data-set. The results indicate that the proposed algorithm can be used to identify known faults, unknown faults and the graphical representation of a network can reflect the spatial distance and relationship among sampled-data.

Defects in combustion chambers of aircraft engines might have an impact on the reliability of the downstream turbine and the machine’s performance. Detecting failures in the combustion chamber of an aircraft engine during operation may improve the resource management and the availability of the system. Aim of the ongoing research project is to find an approach to evaluate the state of the jet engine by analyzing the temperature and emissions field in the exhaust jet. This investigation is part of the collaborative research center SFB 871. The SFB 871 deals with the improvement of the regeneration process of complex capital goods such as aircraft engines. Maintenance, repair, and overhaul processes would be more efficient if the internal status of the engine would be known while still on the wing before it is disassembled. The feasibility of this approach is investigated for a pilot scaled model combustor, which provides optical access and allows the selection of “defined errors” in the combustor. It consists of an atmospheric tubular combustor with an array of eight premixed swirl burners with a maximum output of 160 kW. The operating conditions of one of the eight burners concerning power and air-fuel ratio, can be controlled. A power distribution between the burners is typical fault in an aircraft combustor and will be investigated in this study. It is observed that it is possible to determine small deviations by measuring density profiles applying a tomographic background-oriented schlieren (BOS) technique behind the combustor. Additionally, particle image velocimetry is used to measure differences in the velocity field of the exhaust gases. This study shows that a minimum power deviation of one burner in an array of a total of eight burners is detectable in the exhaust plane with the above mentioned measurement techniques.