A novel incipient fault detection based on residual evaluation by using correlation dimension and inverse of largest Lyapunov exponent

One of the most significant challenges in fault detection and isolation methods is early detection of faults with very slow development rate. This paper proposes a novel robust residual evaluation to detect an incipient fault. Residual evaluation with largest Lyapunov exponent and correlation dimension is proposed to perform early fault detection. Since the residual in fault free case and independent identical distribution (i.i.d) Gaussian noise are typically similar, an analytical formula for largest Lyapunov exponent of i.i.d Gaussian noise with length of N is presented. The incipient faults of DAMADICS benchmark actuator are used to test the proposed approach; which the achieved simulation results confirm the method could detect incipient faults earlier.