APPROACHES TO DIMENSION REDUCTION AND FAULT DIAGNOSIS OF THE HIGH-DIMENSIONAL DYNAMIC SYSTEM

摘要

In this paper a nonlinear PCNN model and an instantaneous stochastic gradient descent algorithm for dimension reduction of the high-dimensional dynamic system are described. A fault diagnosis method via an adaptive observer for the dimension-reduced system is proposed by using a linear residual signal, where an adaptive tuning rule is established that guarantees the monotonically decreasing of a selected Lyapunov function. Finally, The efficiency of the proposed approaches is illustrated through a simulation example and the encouraging results have been obtained.