Vehicle fault diagnostics using a sensor fusion approach

Vehicle electronics and mechanical systems continue to become more complex and interdependent. Automotive electronic control units (ECUs) used in vehicle sub-systems execute high performance algorithms requiring robust fault detection and diagnostics. Our technical approach begins by presenting an overview of several techniques used for processing information from multiple input sources (sensors in particular) in the context of ECU fault detection and diagnostic schemes. We assert that inter-relationships between groups of sensors can be exploited through sensor fusion for signal integrity estimation, and present two vehicle application examples (chassis and powertrain). We propose a virtual fusion/estimation technique which provides basic signal redundancy and fault tolerance. Dynamic vehicle sensor information was used to develop an uncompounded fusion-processing algorithm, expressed through a Matlab/Simulink model. The results of these modeling experiments are presented and subsequent conclusions are drawn regarding the implementation of a sensor fusion-based or sensor fusion-enhanced ECU algorithm as part of a more comprehensive diagnostic, control and service methodology.