Nonlinear observers applied to car suspension fault detection

This paper is a comparison of several observer-based residual generation methods applied to the fault detection of a nonlinear shock absorber. A physical nonlinear model is first developed for the system. The proposed diagnostic model is linear parameter-dependent and thus, the diagnostic procedure is stated as a parameter estimation problem. A few observer design methodologies are devoted to parameter estimation. They are discussed and compared. State observers may also be used with an augmented diagnostic model including parameters as new states. However we show that almost all these approaches fail because of unsatisfied observability assumptions.