Fault Diagnosis and Estimation of Electric Scooter Based on Double-scale UKF

This paper presents a double-scale unscented Kalman filter (UKF) method for fault parameter estimation of an electric scooter, which is modelled by a bond graph (BG) technique. Based on the BG model, faults can be detected by means of deriving analytical redundancy relations (ARRs) and comparing coherence vector (CV) with fault signature matrix (FSM). In the process of parameter degradation, the system can be regarded as a hierarchical system whose different parameters' change speed varies significantly. Therefore, the double-scale UKF algorithm is proposed and applied to estimate the magnitude of the fault parameters in the fault candidates. Different from the UKF, the developed algorithm considers the system parameter variations on a slow-time scale and the system state variations on a fast-time scale, which can improve the computational efficiency compared to traditional methods. In order to validate the effectiveness of the proposed algorithm, simulation is carried out and the results prove the feasibility of the proposed method.