The work presented in this paper focuses on robust fault detection for suspension system. Most of suspension system - classical, active or semi active ones - contains non linear parameters within their reference model such as dampers or springs. The approach taken relies in developing fault indicator for such system. The non-linear model is rewritten as an LPV one, where the scheduling parameters are linked with the non-linear parameters. Then fault indicators are synthesized for the discrete-time LPV model based on the extended parity-space approach. Validation of the approach is performed with CarSim.