A Bayesian least squares support vector machines based framework for fault diagnosis and failure prognosis

Prognostics;PHM;CBM;Diagnostics;Classification;Anomaly detection;Regression;Support vector machines

Khawaja, Taimoor Saleem ; Electrical and Computer Engineering

University:Georgia Institute of Technology

Department:Electrical and Computer Engineering

关键词: Prognostics;    PHM;    CBM;    Diagnostics;    Classification;    Anomaly detection;    Regression;    Support vector machines;   

Others  :  https://smartech.gatech.edu/bitstream/1853/34758/1/Khawaja_Taimoor_S_201008_phd.pdf

美国|英语

来源: SMARTech Repository

【 摘 要 】

A high-belief low-overhead Prognostics and Health Management (PHM) systemis desired for online real-time monitoring of complex non-linear systems operatingin a complex (possibly non-Gaussian) noise environment. This thesis presents aBayesian Least Squares Support Vector Machine (LS-SVM) based framework for faultdiagnosis and failure prognosis in nonlinear, non-Gaussian systems. The methodologyassumes the availability of real-time process measurements, definition of a setof fault indicators, and the existence of empirical knowledge (or historical data) tocharacterize both nominal and abnormal operating conditions.An efficient yet powerful Least Squares Support Vector Machine (LS-SVM) algorithm,set within a Bayesian Inference framework, not only allows for the development ofreal-time algorithms for diagnosis and prognosis but also provides a solid theoreticalframework to address key concepts related to classication for diagnosis and regressionmodeling for prognosis. SVM machines are founded on the principle of StructuralRisk Minimization (SRM) which tends to nd a good trade-o between low empiricalrisk and small capacity. The key features in SVM are the use of non-linear kernels,the absence of local minima, the sparseness of the solution and the capacity controlobtained by optimizing the margin. The Bayesian Inference framework linked withLS-SVMs allows a probabilistic interpretation of the results for diagnosis and prognosis.Additional levels of inference provide the much coveted features of adaptabilityand tunability of the modeling parameters.The two main modules considered in this research are fault diagnosis and failureprognosis. With the goal of designing an efficient and reliable fault diagnosis scheme, a novel Anomaly Detector is suggested based on the LS-SVM machines. The proposedscheme uses only baseline data to construct a 1-class LS-SVM machine which,when presented with online data, is able to distinguish between normal behavior andany abnormal or novel data during real-time operation. The results of the schemeare interpreted as a posterior probability of health (1 - probability of fault). Asshown through two case studies in Chapter 3, the scheme is well suited for diagnosingimminent faults in dynamical non-linear systems.Finally, the failure prognosis scheme is based on an incremental weighted BayesianLS-SVR machine. It is particularly suited for online deployment given the incrementalnature of the algorithm and the quick optimization problem solved in the LS-SVRalgorithm. By way of kernelization and a Gaussian Mixture Modeling (GMM)scheme, the algorithm can estimate (possibly) non-Gaussian posterior distributionsfor complex non-linear systems. An efficient regression scheme associated with themore rigorous core algorithm allows for long-term predictions, fault growth estimationwith confidence bounds and remaining useful life (RUL) estimation after a faultis detected.The leading contributions of this thesis are (a) the development of a novel BayesianAnomaly Detector for efficient and reliable Fault Detection and Identification (FDI)based on Least Squares Support Vector Machines , (b) the development of a data-drivenreal-time architecture for long-term Failure Prognosis using Least Squares SupportVector Machines,(c) Uncertainty representation and management using BayesianInference for posterior distribution estimation and hyper-parameter tuning, and finally (d) the statistical characterization of the performance of diagnosis and prognosisalgorithms in order to relate the efficiency and reliability of the proposed schemes.