Model-based Diagnosis of an Automotive Electric Power Generation and Storage System

This paper presents mathematical models, design and experimental validation, and calibration of a model-based diagnostic algorithm for an electric-power generation and storage automotive system, including a battery and an alternator with a rectifier and a voltage regulator. Mathematical models of these subsystems are derived, based on the physics of processes involved as characterized by time-varying nonlinear ordinary differential equations. The diagnostic problem focuses on detection and isolation of a specific set of alternator faults, including belt slipping, rectifier fault, and voltage regulator fault. The proposed diagnostic approach is based on the generation of residuals obtained using system models and comparing predicted and measured value of selected variables, including alternator output current, field voltage, and battery voltage. An equivalent input-output alternator model, which is used in the diagnostic scheme, is also formulated and parameterized. The test bench used for calibration of thresholds of the diagnostic algorithm and overall validation process are discussed. The effectiveness of the fault diagnosis algorithm and threshold selection is experimentally demonstrated.