Abstract:

To avoid the bad effects on differential calculation and to possess a faster convergent speed relative to simple-proportional-type algorithm in the process of the iterative learning, a new fault diagnosis algorithm is proposed by applying the scheme of proportional difference type learning to fault diagnosis. This algorithm uses residual errors and the differential signal of the adjacent two residual errors to correct the virtual fault signals, which enables the virtual fault to appromixiate the fault occurred actually in the system, thereby attaining the end of system fault diagnosis. The convergence of the fault estimator is proven by the contraction mapping approach. This algorithm can not only detect different-type fault of the system effectively, but also estimate the fault signal accurately. Finally, simulation results verify the validity of the algorithm.