The application of fuzzy logic to the diagnosis of automotive systems

Fault diagnosis of a physical plant is crucial for its healthy performance, as it ultimately could prevent catastrophic failure, help comply with environmental regulations, and enhance customer satisfaction. There exist several methods to detect and isolate incipient faults that might cause a plant's performance to deviate from the nominal. Fault detection and diagnostic methods may utilize subjective or objective knowledge as the basis for the diagnostic scheme.; A methodology for the integration of subjective (heuristic) and objective (analytical) knowledge for fault diagnosis and decision-making is proposed in this dissertation. The structure, challenges, and benefits of such integration are explored. The observed symptoms are measured using a wide range of sensors and typically these measurements are sampled over different time scales and knowledge from various sources is utilized. The fusion of such sensor measurements, the combination of different time scale data, and the integration of the knowledge sources is a serious challenge that must be overcome by the integration scheme. In this methodology two parallel paths are followed, making use of the system's input-output data. Along the first path a model-based scheme is applied, where primary residuals are generated using sliding mode nonlinear observers. These residuals are evaluated and processed using fuzzy logic for fault detection and isolation. Along the second path a knowledge-based scheme is applied using a combination of fuzzy estimation techniques and fuzzy heuristic rules for fault detection and isolation. The two paths merge together and integration of both schemes is completed in the inference mechanism. Finally a fuzzy decision-making process yields the diagnosis. The philosophy and structure of the diagnostic scheme is explained.; The experimental verification of this methodology is then carried out on a Ford 4.6 L V8 engine controlled by an electric dynamometer housed in the Ohio State University Mechanical Engineering Department and Center for Automotive Research. A hierarchical diagnostic method that explores the functional structure of the automotive engine subsystems and components is applied. The application and results in the context of automotive engine exhaust emission controls system diagnostics are demonstrated. Fault detection and isolation of induced faults was completed successfully.