Time-Shift Sliding Mode Predictor for Fault Reconstruction of Time-Delay Systems Using Least Squares Identification

This paper presents a sliding mode predictor observer for state estimation and fault reconstruction of linear time-invariant systems with output delays based on a novel Time-Shift Approach. Plants with arbitrary time-varying output delays are considered. The actuator fault is modeled as a weighted sum of known functions and unknown coefficients. The fault at current time is identified using continuous time Recursive Least Squares (RLS) method. Then, an open loop predictor based on variation of constants formula advances the delayed estimate of the state. Ideal sliding mode can be theoretically achieved even in the presence of delays. Numerical simulations are presented to show the effectiveness of the method.