Incipient fault prediction for nonlinear stochastic distribution systems
Chang Zhang, Lina Yao
Abstract
Abstract In order to avoid the occurrence of major accidents in the industrial production process, incipient fault prediction that can predict fault in advance has attracted more and more attentions. In this article, a generalized correntropy filtering‐based incipient fault prediction strategy is proposed, which is suitable for nonlinear stochastic distribution systems with simultaneous actuator and sensor faults and non‐Gaussian noise. Three filters are designed to predict the incipient fault of nonlinear stochastic distribution systems: the first one is used to predict the time when the incipient fault occurs, and the other two are used to estimate the evolution rate of sensor fault and actuator fault, respectively. When the residual signal exceeds the predetermined threshold, the fault is detected and the time of occurrence can be also estimated. Once the fault is detected, the designed filter can be used to estimate the evolution rate of the unknown fault. After estimating the time of fault appearance and its evolution rate, the change process of incipient fault can be predicted. Finally, an example of a chemical reaction process is given to illustrate the validity of the prediction scheme.