Model based sensor fault detection for an active vehicle suspension

Abstract

At the beginning, a short introduction into the topic of active vehicle suspensions and the LOLIMOT neuronal network, which is used to identify accurate models of the active suspension system, is presented. Subsequently, physical mathematical models of the considered active vehicle suspension implemented in a test rig are derived. The structure of these nonlinear physical models is then used to define the inputs and outputs of the local linear neuronal network LOLIMOT. Hence, a semi-physical modelling approach results. This network is then trained with test rig data, in order to obtain accurate parity equations for model based fault detection. Various faults are simulated with test rig data and it is shown, how these faults are detected.