Application of signal analysis and data-driven approaches to fault detection and diagnosis in automotive engines

The modern era of sophisticated automobiles is necessitating the development of generic and automated embedded fault diagnosis tools. Future vehicles are expected to contain more than one hundred complex Electronic Control Units (ECUs) and data acquisition systems to control and monitor large number of system variables in real-time. There exists an abundant amount of literature on fault detection and diagnosis (FDD). However, these techniques are developed in isolation. In order to solve the problem of FDD in complex systems, such as modern vehicles, a hybrid methodology combining different techniques is needed. Here, we apply an approach based on signal analysis that combines various signal processing and statistical learning techniques for real-time FDD in automotive engines. The data under several scenarios is collected from an engine model running in a real-time simulator and controlled by an ECU.