Systematic Data-Driven Approach to Real-Time Fault Detection and Diagnosis in Automotive Engines

The competitive businesses' desire to provide "smart services" and the pace at which the modern automobiles are increasing in complexity, are motivating the development of automated intelligent vehicle health management systems. Current On-Board Diagnosis (OBD II) systems use simple rules and maps to perform diagnosis, and significant human intervention is needed to troubleshoot a problem. More research is needed on developing innovative, easy-to-use automated diagnostic approaches for incorporation into the OBD systems. In addition, developing intelligent remote diagnosis technology, building a bridge between on-board and off-board diagnosis are open areas of research in the automotive industry.Here, we propose a systematic data-driven process that utilizes knowledge from signal- processing and statistical domains to detect and diagnose faults in automotive engines. The proposed approach is applied to a Toyota Camry engine, and the experimental results are presented in detail. The experimental system consists of an engine running with manual transmission on a dynamometer test- stand. For our experiments, the data for five faults (three sensor faults and two physical faults) with different severity levels under various operating conditions (e.g., different throttle angles, engine speeds, etc) is collected from the engine, and the application of a data-driven diagnostic process is examined.