Fault Detection for Nonlinear Systems with Periodic Input

Abstract

Fault diagnosis is important to improve the reliability of engineering systems. Most existing fault diagnosis techniques consider only systems with non-periodic input, though periodic signals are also common in control systems. It is, therefore, necessary to develop fault diagnosis techniques for systems with periodic input. A popular approach to detect abrupt and incipient faults for systems with random noise is the asymptotic local approach based on a statistic which is a function of the residuals obtained from the process model obtained under normal operating conditions. However, the results obtained from this technique are unsatisfactory when the input of the system is periodic. A new criterion is proposed in this paper that allows the asymptotic local approach to be extended to systems with periodic inputs. The application of this technique to nonlinear system is discussed and illustrated by a simulation example.