Industrial machinery is developing in the direction of large-scale, automation, and high precision, which brings novel troubles to mechanical equipment management and maintenance. Intelligent diagnosis of mechanical running state based on vibration signals is becoming increasingly important, and it is still a great challenge at pattern recognition. As one of the indispensable components in mechanical equipment, planetary gearboxes are widely used in wind power, aerospace, and heavy industry. However, the problem of automatically maximizing the accuracy of planetary gearbox under different working conditions has not been solved. Therefore, an intelligent diagnosis method for planetary wheel bearing based on constrained independent component analysis (CICA) and stacked sparse autoencoder (SSAE) is presented in this research. Firstly, the fault signal with obvious time-domain characteristics is extracted by constrained independent component analysis (CICA), and the fault signals and noise is separated. Then, calculating the correlation kurtosis value of the time domain signals at different iteration periods as the eigenvalue to obtain the training samples and the test samples. The parameters of the network layer, the number of hidden nodes and learning rate are determined to build the model of SSAE. In the end, the training samples are input into the model for training and the whole network is fine-tuned. The advantages and disadvantages of the model are verified by the test samples. The intelligent classification and diagnosis of the mechanical running state are completed. Experiments analysis with real datasets of planetary wheel bearing show that the proposed method can achieve higher accuracy and robustness for fault classification compared with other data-driven methods. The application of this method in other major machinery industry also has bright prospects.

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.

Gas turbine autotuning systems are surfacing as a popular OEM and third party solution to minimize manual tuning for emissions compliance, combustion dynamics, and performance optimization. While these systems provide many valuable benefits, they also introduce new challenges. This paper presents a review of autotuning systems, including the general approaches, benefits, challenges, and best practices for combustion dynamics monitoring. The primary benefit of autotuning is that it provides a mostly automated solution to address operability/emissions challenges which arise due to changes in ambient conditions, fuel composition, and machine aging. The operability and emissions challenges include compliant NOx and CO, acceptable combustion dynamics levels, acceptable operability (i.e., turndown and lean blowout avoidance), and in some cases optimized performance. The overarching potential benefit of autotuning systems is their ability to address all of these issues simultaneously and continuously. The paper first presents a brief outline of the general tuning considerations for autotuning systems, with some examples from real plant data to illustrate the tuning sensitivities. It next reviews the major challenges that autotuning can introduce to combustion dynamics monitoring, such as combustor fault pattern recognition and greater consequences of instrumentation faults. Finally, the paper recommends best practices for monitoring combustion dynamics in systems with autotuning. These recommendations include what to do (and what not to do) to continue health monitoring with advanced pattern recognition software, and how to recognize the signatures of combustion dynamics instrumentation faults. This paper is directed at gas turbine operators. It presents familiar plant data to help this audience understand the core working principles of an autotuning system. This understanding is an important basis for determining when a combustion dynamics event is attributable to the operations of the autotuning system, an instrumentation fault, or combustion system hardware degradation. With this understanding established, this paper also presents a list of capabilities and best practices that should be incorporated into combustion dynamics monitoring strategies for units that use autotuning.

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.

The condition monitoring and fault diagnosis of rolling element bearing is a very important research content in the field of gas turbine health management. In this paper, a hybrid fault diagnosis approach combining S-transform with artificial neural network (ANN) is developed to achieve the accurate feature extraction and effective fault diagnosis of rolling element bearing health status. Considering the nonlinear and non-stationary vibration characteristics of rolling element bearing under stable loading and rotational speeds, S-transform and singular value decomposition (SVD) theory are firstly used to process the vibration signal and extract its time-frequency information features. Then, radical basis function (RBF) neural network classification model is designed to carry out the state pattern recognition and fault diagnosis. As a practical application, the experimental data of rolling element bearing including four health status are analyzed to evaluate the performance of the proposed approach. The results demonstrate that the present hybrid fault diagnosis approach is very effective to extract the fault features and diagnose the fault pattern of rolling element bearing under different rotor speed, which may be a potential technology to enhance the condition monitoring of rotating equipment. Besides, the advantages of the developed approach are also confirmed by the comparisons with the other two approaches, i.e. the Wigner-Ville (WV) distribution and RBF neural network based method as well as the S-transform and Elman neural network based one.

Turbine-generator shaft systems used in power generation applications are exposed to degradation mechanisms that could result in high consequence failures if not discovered prior to damage accumulation. Grid-induced torsional vibration, growth of cracks in the shaft forging, and large blade vibration are some examples of degradation that remains unmonitored in most commercial plants today. In many cases, the sensing and subsequent trending of high-quality vibration data obtained directly from the shaft surface can be the basis for a decision to continue to operate versus inspect or repair. Detection of small changes in torsional and lateral vibration mode properties can be sensed at a single shaft location and trended using techniques such as Advanced Pattern Recognition to reveal the very early signs of rotor distress. Contemporary barriers to widespread application of wireless shaft vibration measurements for health monitoring were studied and addressed in the development of the Turbine Dynamics Monitoring System (TDMS). The resulting design evolved around the industry need for low sensor maintenance, high reliability, ease of installation, and high data quality to enable early detection of critical component changes. These improvements capitalized on advances in strain gage and accelerometer technology, micro-telemetry, radio-frequency power systems, and advanced adhesives for installation. The new system has been successfully applied in the field on large steam turbine-generators to detect grid-induced torsional vibration. The paper will describe background of turbine-generator torsional vibration as well as the technical features of this advanced telemetry application with examples of field data.

Centrifugal compressor is a piece of key equipment for factories. Among the components of centrifugal compressor, impeller is a pivotal part as it is used to transform kinetic energy to pressure energy. But it usually leads to blade crack or failure as irregular aerodynamic load effect on the blade. Therefore, early crack feature extraction and pattern recognition is important to prevent it from failure. Although time series analysis for monitored signal can be used on feature extraction, incipient weak feature extraction method should be investigated. In this research, pressure pulsation sensors arranged in close vicinity to crack area are used to monitor the blade crack and feature extraction. As there are different kinds of flow interference, the pressure pulsation signal for centrifugal compressor is full of nonlinear characteristics. Therefore, how to obtain the weak information from monitored signal is investigated. Although FFT and envelope analysis have been widely used for rotating equipment, they are not suitable for the determination of incipient crack of a blade as the signal modulation and noise interference. In this research, stochastic resonance is used for the pressure pulsation signal. The results show that it is an effective tool to blade incipient crack classification on centrifugal compressor.

Centrifugal compressor is a piece of key equipment for factories. Among the components of a centrifugal compressor, impeller is a pivotal part as it is used to transform kinetic energy to pressure energy. The blades are exposed to centrifugal forces, gas pressure, and the friction force which usually lead to cracks. Therefore, early crack feature extraction and pattern recognition are important to prevent it from failure. Although time series analysis for monitored signals can be used on feature extraction, it is not enough. So the incipient weak feature extraction method should be investigated. In this research, pressure pulsation sensors arranged close to crack area are used to monitor the blade crack signal and extract the feature information. As the different kinds of interference of flow, the pressure pulsation signals for a centrifugal compressor are full of nonlinear characteristics. Therefore, how to obtain the weak information from monitored signals effectively should be investigated. A method on blade crack classification is present by continuous wavelet transform (CWT) and envelope spectrum in this research. Simulation signal analysis and experimental investigation on blade crack classification are carried out to verify the effectiveness of this method. The results show that it is an effective tool for blade incipient crack classification for a centrifugal compressor.

A significant challenge in improving the regeneration process of jet engines is the reduction of engine down-time during inspection. As such, early defect detection without engine disassembly will speed up the regeneration process. Defects in the engines hot-gas path (HGP) influence the density distribution of the flow and lead to irregularities in the density distribution of the exhaust jet which can be detected with the optical Background-Oriented Schlieren (BOS) method in a tomographic set-up. The present paper proposes a combination of tomographic BOS measurements and supervised learning algorithms to develop a methodology for an automatic defect detection system. In a first step, the methodology is verified by analyzing the exhaust jet of a swirl burner array with a non-uniform fuel-supply of single burners with tomographic BOS measurements. The measurements are used to implement a Support Vector Machine (SVM) pattern recognition algorithm. It is shown that the reconstruction quality of tomographic BOS measurements is high enough to be combined with pattern recognition algorithms. The results strengthen the hypothesis, that it is possible to automatically detect defects in jet engines with tomographic BOS measurements and pattern recognition algorithms.

A method enabling the automated diagnosis of Gas Turbine Compressor blade faults, based on the principles of statistical pattern recognition is initially presented. The decision making is based on the derivation of spectral patterns from dynamic measurements data and then the calculation of discriminants with respect to reference spectral patterns of the faults while it takes into account their statistical properties. A method of optimizing the selection of discriminants using dynamic measurements data is also presented. A few scalar discriminants are derived, in such a way that the maximum available discrimination potential is exploited. In this way the success rate of automated decision making is further improved, while the need for intuitive discriminant selection is eliminated. The effectiveness of the proposed methods is demonstrated by application to data coming from an Industrial Gas Turbine while extension to other aspects of Fault Diagnosis is discussed.

The concept of performance monitoring for prevention of certain serious failures in gas turbines is described. The use of compressor mapping as the key to avoiding surge is developed, and an example is presented showing how the compressor in a steam-injected gas turbine can be much closer to surge in one of two nearly-identical operating points on a steam-injection control envelope than the compressor in the other. The technique of monitoring blade-path temperature spread in the exhaust of a gas turbine is then described, and examples of its value in preventing combustor burnout and turbine blade failures in high-frequency fatigue are given. Next, a concept of diagnosing internal deterioration by recognizing patterns of deviation of operating parameters from baseline data is described, and illustrated for a single-shaft generator-drive gas turbine. Finally, the use of a modern computer-controlled data acquisition system to perform the above monitoring functions in real time is demonstrated.

Integrated life, vibration and performance monitoring/diagnostics capable of detecting and classifying developing engine faults is critical to reducing engine operating and maintenance costs while optimizing the life of “hot section” engine components (Troudet and Merrill, 1990). Advanced fault pattern recognition and classification techniques utilizing complex finite-element and empirical models of structural and performance related engine areas can now be accessed in a real-time monitoring environment (Dietz et al., 1989). Integration and implementation of these proven technologies presents a great opportunity to significantly enhance current engine health diagnostic capabilities and safely extend engine component life (Ali and Crawford, 1988).

The feasibility of eliminating contact in a noncontacting flexibly mounted rotor (FMR) mechanical face seal is studied. The approach for contact elimination is based on a parametric study using FMR seal dynamics. Through clearance adjustment it is possible to reduce the maximum normalized relative misalignment between seal faces and, therefore, eliminate seal face contact Clearance is measured by proximity probes and varied through a pneumatic adjustment mechanism. Contact is determined phenomenologically from pattern recognition of probe signals and their power spectrum densities as well as angular misalignment orbit plots, all calculated and displayed in real-time. The contact elimination strategy is experimentally investigated for various values of stator misalignment and initial rotor misalignment Contrary to intuition but compliant with the parametric study, the experimental results show that for the seal under consideration contact can be eliminated through clearance reduction.

The goal of Gas Turbine Performance Diagnostics is to accurately detect, isolate and assess the changes in engine module performance, engine system malfunctions and instrumentation problems from knowledge of measured parameters taken along the engine’s gas path. Discernable shifts in engine speeds, temperatures, pressures, fuel flow, etc., provide the requisite information for determining the underlying shift in engine operation from a presumed nominal state. Historically, this type of analysis was performed through the use of a Kalman Filter or one of its derivatives to simultaneously estimate a plurality of engine faults. In the past decade, Artificial Neural Networks (ANN) have been employed as a pattern recognition device to accomplish the same task. Both methods have enjoyed a reasonable success.

Real-time, integrated health monitoring of gas turbine engines that can detect, classify, and predict developing engine faults is critical to reducing operating and maintenance costs while optimizing the life of critical engine components. Statistical-based anomaly detection algorithms, fault pattern recognition techniques and advanced probabilistic models for diagnosing structural, performance and vibration related faults and degradation can now be developed for real-time monitoring environments. Integration and implementation of these advanced technologies presents a great opportunity to significantly enhance current engine health monitoring capabilities and risk management practices. This paper describes some novel diagnostic and prognostic technologies for dedicated, real-time sensor analysis, performance anomaly detection and diagnosis, vibration fault detection, and component prognostics. The technologies have been developed for gas turbine engine health monitoring and prediction applications which includes an array of intelligent algorithms for assessing the total ‘health’ of an engine, both mechanically and thermodynamically. This includes the ability to account for uncertainties from engine transient conditions, random measurement fluctuations and modeling errors associated with model-based diagnostic and prognostic procedures. The implementation of probabilistic methods in the diagnostic and prognostic methodology is critical to accommodating for these types of uncertainties.

Efficiency of gas turbine monitoring systems primarily depends on the accuracy of employed algorithms, in particular, pattern recognition techniques to diagnose gas path faults. In investigations many techniques were applied to recognize gas path faults, but recommendations on selecting the best technique for real monitoring systems are still insufficient and often contradictory. In our previous works, three recognition techniques were compared under different conditions of gas turbine diagnosis. The comparative analysis has shown that all these techniques yield practically the same accuracy for each comparison case. The present contribution considers a new recognition technique, Probabilistic Neural Network (PNN), comparing it with the techniques previously examined. The results for all comparison cases show that the PNN is not practically inferior to the other techniques. With this inference, the recommendation is to choose the PNN for real monitoring systems because it has an important advantage of providing confidence estimation for every diagnostic decision made.

Power plant owners require their plants’ high reliability, availability and also reduction of the cost in today’s power generation industry. In addition, the power generation industry is faced with a reduction of experienced operators and sophistication of power generation equipment. Remote monitoring service provided by original equipment manufacturers (OEMs) has become increasingly popular due to growing demand for both improvement of plant reliability and solution of experienced operator shortage. Through remote monitoring service, customers can benefit from swift and appropriate operational support based on OEM’s know-how. Before implementation of remote monitoring, the customer and OEM often required repeated interchanges of information about operation and instrumentation data. These interchanges took a lot of time. Data analysis and estimation of deterioration were time-consuming. Remote monitoring has enabled us, OEMs, not only to access to a plant’s real-time information but also to trace the historical operation data, and therefore the required time of data analysis and improvement has been reduced. Mitsubishi Heavy Industries, Ltd. also embarked on around-the-clock remote monitoring service for gas turbine plant over a decade ago and has increased its ability over time. At present, the application of remote monitoring systems have been extended not only into proactive maintenance by making use of diagnostic techniques carried out by expert engineers but also into building a pattern recognition system and an artificial intelligence system using expert’ knowledge. Conventional diagnostics is only determining whether the plant is being operated within the prescribed threshold levels. Pattern recognition is a state-of-the-art technique for diagnosing plant operating conditions. By comparing past and present conditions, small deterioration can be detected before it needs inspection or repair, while all the operating parameter is within their threshold levels. Mahalanobis-Taguchi method (MT method) is a technique for pattern recognition and has the advantage of diagnosing overall GT condition by combining many variables into one indicator called Mahalanobis distance. MHI has applied MT method to the monitoring of gas turbines and verified it to be efficient method of diagnostics. Now, in addition to the MT method, automatic abnormal data discrimination system has been developed based on an artificial intelligence technique. Among a lot of artificial intelligence techniques, Bayesian network mathematical model is used.

Efficiency of gas turbine condition monitoring systems depends on quality of diagnostic analysis at all its stages such as feature extraction (from raw input data), fault detection, fault identification, and prognosis. Fault identification algorithms based on the gas path analysis may be considered as an important and sophisticated component of these systems. These algorithms widely use pattern recognition techniques, mostly different artificial neural networks. In order to choose the best technique, the present paper compares two network types: a multilayer perceptron and a radial basis network. The first network is being commonly applied to recognize gas turbine faults. However, some studies note high recognition capabilities of the second network. For the purpose of the comparison, both networks were included into a special testing procedure that computes for each network the true positive rate that is the probability of a correct diagnosis. Networks were first tuned and then compared using this criterion. Same procedure input data were fed to both networks during the comparison. However, to draw firm conclusions on the networks’ applicability, comparative calculations were repeated with different variations of these data. In particular, two engines that differ in an application and gas path structure were chosen as a test case. By way of summing up comparison results, the conclusion is that the radial basis network is a little more accurate than the perceptron, however the former needs much more available computer memory and computation time.

A great deal of attention has been attracted in the analytical model-based engine diagnostics over the past years. Meanwhile, an increasing number of researchers and practitioners make an attempt to gain an intelligent diagnoser in a pattern recognition way. A question naturally emerges of how to combine the two techniques to improve the robustness of an on-board diagnostic system. In this context, this paper suggests an integrated approach that combines the unknown input observer (UIO) with the support vector machine (SVM) technique to aircraft engine fault diagnosis. Sensor faults and actuator faults are separately considered. To reduce the effect of engine disturbances on diagnostic performance, we first design a bank of UIOs, each of which is sensitive to all sensor and actuator faults but only one signal. Then, the magnitudes of a set of residuals between the UIO-based estimations and the engine measurements are fed into an SVM classifier to detect and isolate engine faults. Experimental results demonstrate an encouraging potential of the suggested method and that the UIO-oriented approach is superior or competitive to the Kalman-based algorithm.

In recent years, rapid growth in wind energy as a substantial source of electricity generation has created greater demands on wind turbine system reliability and availability. To reduce service costs and maximize return on investment, wind farm operators have begun to take a more proactive approach to turbine problems by relying on intelligent condition monitoring and automated failure detection systems. The challenge is how to effectively convert large amounts of data into actionable decisions to detect and isolate failures at an early stage. This paper describes a unique data analysis and modeling technique for online turbine health monitoring and automatic root cause assessment. It provides a means to capture failure signatures for specific root causes based on historical events as well as engineering knowledge. Both continuous and discrete turbine condition monitoring data are processed to provide a failure probability assessment. First, statistical trend analysis, feature extraction and classification methods are developed to analyze a continuous sensor data set. Secondly, a pattern recognition method is applied to calculate failure indicators from various discrete control system events, or fault messages. Then failure likelihoods derived from both the continuous and the discrete models are combined in a fusion model to increase predictive accuracy. A demonstration of the method on bearing failure modeling using SCADA data will be provided with promising results.