Sliding Mode Observer Based Predictive Fault Diagnosis of a Steer-By-Wire System

This paper presents a nonlinear observer and prediction based analytical redundancy for a SteerBy-Wire (SBW) system. A Sliding Mode Observer was designed to estimate the vehicle steering angle by using the combined linear vehicle model, SBW system, and the yaw rate feedback. The estimated steering angle along with the current input was used to predict the steering angle at various prediction horizons via a long range prediction method. This analytical redundancy methodology was utilized to reduce the total number of redundant road-wheel angle (RWA) sensors, while maintaining a high level of reliability. The Fault Diagnosis algorithm was developed using a majority voting scheme, which was then used to detect faulty sensor(s) in order to maintain safe drivability. The proposed observer-prediction based fault detection algorithms as well as the linearized vehicle model were modelled in MATLAB-SIMULINK. Two different fault types were used to evaluate the effectiveness of the proposed algorithms: persistent and incipient faults. Simulation results show that the faulty sensor identification time decreases with the increase of prediction horizon illustrating advantages of the predictive analytical redundancy based algorithms against single point faults.