Complete-model diesel-engine diagnosis using gain schedule-mu analysis and non-linear estimator

Three different fault-diagnosis system for a turbocharged diesel engine with variable-geometry turbocharger control, were proposed in this paper. Numerous and diversified actuator and/or sensors faults such as air-leakage in the admission collector, compressor malfunctioning, intake-valves fault, intercooler fault, deterioration in the turbine-compressor coupling, defect in the variable geometry of the turbine are identified and analyzed. Furthermore, a complete non-linear engine model with four state variables was adopted. The first proposed strategy consists in developing a Fault Detection and Isolation algorithm (FDI) based on the theory of Gain Schedule Control operated on a Takagi-Sugeno model of the diesel. The second proposed strategy based on the theory of μ analysis control, carried out on a linearization model LTI of the diesel. The third strategy adopted and implemented in the current work follows similar general method published by Demetrion and Polycarpou [15]. The basic idea is to use an adaptive training of an on line observer to estimate the unknown faults parameters.