Fault detection method for nonlinear systems based on probabilistic neural network filtering

A fault detection method for nonlinear systems, which is based on Probabilistic Neural Network Filtering (PNNF), is presented. PNNF limits the maximum estimation error of the asymptotic Bayes optimal result and describes the tracking process with an expression. On the basis of these properties of PNNF and the statistical characteristics of the noise of the system, a fault threshold can be better calculated, especially for the tracking process corresponding to a strong disturbance. According to the threshold, the fault can be detected by evaluating the residuals. Also, for some special cases when a fault is potential but the system is in steady state, which causes the information for fault detection may be insufficient and a group of disturbances are artificially input with definite amplitudes, so that the result of detection can be enhanced by comparing the real with the expected tracking processes of the filter. Examples are given to demonstrate the method of fault detection based on PNNF.

[1]  Donald F. Specht,et al.  Probabilistic neural networks and the polynomial Adaline as complementary techniques for classification , 1990, IEEE Trans. Neural Networks.

[2]  Michèle Basseville,et al.  Fault Detection and Isolation in Nonlinear Dynamic Systems: A Combined Input-Output and Local Approach , 1998, Autom..

[3]  L. Teng.  Theory and application of Kalman filtering , 1970 .

[4]  H. Sirisena,et al.  Computation of optimal output feedback gains for linear multivariable systems , 1974 .

[5]  Rolf Isermann,et al.  Process fault detection based on modeling and estimation methods - A survey , 1984, Autom..

[6]  Marios M. Polycarpou,et al.  Automated fault diagnosis in nonlinear multivariable systems using a learning methodology , 2000, IEEE Trans. Neural Networks Learn. Syst..

[7]  R. E. Kalman,et al.  New Results in Linear Filtering and Prediction Theory , 1961 .

[8]  Marios M. Polycarpou,et al.  Neural-network-based robust fault diagnosis in robotic systems , 1997, IEEE Trans. Neural Networks.

[9]  Dingli Yu,et al.  A bilinear fault detection observer , 1996, Autom..

[10]  P. Frank.  On-line fault detection in uncertain nonlinear systems using diagnostic observers: a survey , 1994 .

[11]  R. Marino.  Adaptive observers for single output nonlinear systems , 1990 .

[12]  Mark H. A. Davis.  The application of nonlinear filtering to fault detection in linear systems , 1975 .

[13]  D. F. Specht,et al.  Probabilistic neural networks for classification, mapping, or associative memory , 1988, IEEE 1988 International Conference on Neural Networks.

[14]  Alan S. Willsky,et al.  A survey of design methods for failure detection in dynamic systems , 1976, Autom..

[15]  Rolf Isermann,et al.  Process Fault Detection Based on Modeling and Estimation Methods , 1982 .

[16]  Paul M. Frank,et al.  Fault Diagnosis in Dynamic Systems via State Estimation - a Survey , 1987 .

[17]  John E. White,et al.  Detection Filter Design: Spectral Theory and Algorithms , 1986, 1986 American Control Conference.

[18]  Steven X. Ding,et al.  Frequency domain approach to optimally robust residual generation and evaluation for model-based fault diagnosis , 1994, Autom..

[19]  A. Willsky,et al.  Analytical redundancy and the design of robust failure detection systems , 1984 .

[20]  Donald F. Specht,et al.  Probabilistic neural networks , 1990, Neural Networks.