Using a Multiple-Model-Based Hybrid Kalman Filter, Sensor Fault Detection, Isolation, and Identification
Abstract
The multiple model (MM) method is used in this research to present an unique sensor failure detection, isolation, and identification (FDII) technique. The system is based on multiple hybrid Kalman filters (HKF), which integrate a number of piecewise linear (PWL) models with a nonlinear mathematical model of the system. By interpolating the PWL models using a Bayesian method, the proposed fault detection and isolation (FDI) scheme is able to identify and isolate sensor defects over the full operational regime of the system. Additionally, the suggested multiple HKF-based FDI methodology is enhanced to determine the size of a sensor defect by utilising a modified generalised likelihood ratio (GLR) method that depends on the system's healthy operational mode.