A Robust time-varying Fault Detection and Isolation method for a SparkInjection Engine

摘要

In this paper we propose a Fault Detection andIsolation (FDI) filter design method for Spark Injection En-gines. Starting from a detailed nonlinear Mean-value enginemathematical representation, a LPV approximation based ona judicious convex 'interpolation of a family of linearizedmodels is obtained. An LPV-FDI filter based on the Luenbergerobserver theory is synthesized by ensuring guaranteed levelsof disturbance rejection and fault detection and isolation. Theresulting diagnostic filter is parameter-dependent and uses a setof scheduling engine parameters, assumed measurable on-line.The effectiveness of the LPV-FDI framework is illustrated bynumerical examples. The obtained LPV approximation is herevalidated and the diagnostic capabilities of the proposed FDIarchitecture proved.