| This thesis presents the design and implementation of Fault Detection Filters in automotive subsystems. First, the existing framework of the Beard-Jones Detection Filter in linear time-invariant systems is extended to periodic systems because many vehicle subsystems exhibit periodic behaviors due to the rotation of the crankshaft in a multi cylinder internal combustion engine. Then a highly flexible and efficient VLSI architecture is proposed to implement detection filters in both linear time-invariant and periodic systems. Finally, the design of an onboard real-time fault detection system is presented which includes a floating-point coprocessor for error residual generation, a dedicated instruction set processor for fault detection and isolation, a set of data acquisition units for data collection, and other peripheral components to support the system operation. This design is very flexible in that it can be used, with some minor modifications, to accommodate not only the diagnostic observers but also other fault detection methodologies that are formulated using state space models, for example the parity equations and the Kalman filters. |
