The vehicle lateral control system performs lane-keeping and lane-changing tasks on automated highways. The controller acquires the vehicle's lateral positions through two sets of magnetometers. Failures on either set of magnetometers may cause instability of the system and jeopardize safety of the passengers. Hence it is of crucial importance to design and implement a fault detection and identification (FDI) mechanism which is responsible for handling the magnetometers' malfunctions. In this paper, we propose an FDI scheme consisting of two switching observers and a probabilistic inference algorithm. The existence of failures is detected by evaluating the posterior probabilities of fault occurrences, while the sizes and types of the faults are estimated by the observers. Simulation results show that sensor failures can be detected and identified in a timely manner. Moreover the lateral control system maintains stability, and an acceptable level of performance is achieved in the presence of sensor failures.