Robust fault detection based on multimodel and interval approach. Application to a throttle valve

A multimodel approach has been set up in this paper. The purpose is to replace a nonlinear system by a linear one combined with interval analysis approach in order to take into account uncertainties in the model. Linear systems can be easily handled with interval computation. However, nonlinear systems need complex computations as set-computation based on sub-paving algorithms for example. To avoid a large amount of computation, the approach proposed herein is to decompose a nonlinear system in several linear models. Uncertain systems can take advantages of this approach called multimodel approach. In this paper a robust fault detection method based on multimodel will be presented. The detection will be made by an Interval Set-Valued Observer. The system is nonlinear and represents a throttle valve. It will be shown how the proposed method can replace advantageously an identification step by taking into account uncertainties on parameters.