Data‐driven fault‐tolerant control for SISO nonlinear system with unknown sensor fault

Abstract

SummaryThis paper studies the adaptive fault‐tolerant control problem for a general nonlinear discrete‐time SISO system with unknown system model and sensor fault. First, utilizing the input‐output (I/O) data, an equivalent full‐form dynamic linearization (FFDL) data model is to be constructed by introducing a pseudo‐gradient vector. Then, to estimate the system's actual output from the sensor measurements corrupted by unknown faults, a nonlinear autoregressive with external input neural network (NARXNN) is employed and well‐trained, by which the compensation of the fault signal can hence be derived indirectly. Based on the optimality criterion, an adaptive fault‐tolerant control (FTC) strategy is therefore proposed, which promises the convergence of tracking error and the boundedness of system signals. The effectiveness of the proposed FTC algorithm is illustrated by simulation results.