Severity-based diagnosis for vehicular electric systems with multiple, interacting fault modes

Abstract

Complex systems are comprised of multiple components that continuously interact in terms of how they degrade and fail. Diagnosing fault severity and causes of failures in these systems is often a non-trivial task. To address this challenge, we propose a data-driven, severity-based diagnosis framework for systems with multiple, interacting fault modes. We focus on the components of the automotive electric power generation and storage system, specifically, the Vehicle-Engine Start system comprised of the battery and the start-stop starter. Our framework leverages sensor data from several component-fault severity combinations. Using multiple feature extraction tools, we train separate classifiers using Regularized Multinomial Regression, and combine the performance of the classifiers using ensemble methods. We demonstrate the effectiveness of our approach by performing degradation-based diagnostic tests utilizing a real-world engine test-rig.