An adaptive SIF and KF estimation strategy for fault detection based on the NIS metric

State estimation strategies play an essential role in the effective operation of dynamic systems by extracting relevant information about the system’s state when faced with limited measurement capability, sensor noise, or uncertain dynamics. The Kalman filter (KF) is one of the most commonly used filters and provides an optimal estimate for linear state estimation problems. However, the KF lacks robustness as it does not perform well in the face of modelling uncertainties and disturbances. The sliding innovation filter (SIF) is a newly proposed filter that uses a switching gain and innovation term, and unlike the KF, it only results in a sub-optimal estimate. However, the SIF has been proven to be robust to modelling uncertainties, disturbances, and ill-conditioned problems. In this work, we propose an adaptive SIF and KF (SIF-KF) estimation algorithm that can detect faulty or uncertain conditions and switch between the KF and SIF gain in the absence or presence of such conditions, respectively. A fault detection mechanism based on the normalized innovation squares (NIS) metric is also presented, which is responsible for triggering the activation of the respective gain in the proposed SIF-KF strategy. Experimental simulations are carried out on a simple harmonic oscillator subject to a fault to demonstrate the proposed SIF-KF’s effectiveness over traditional approaches.