Intelligent monitoring of electric vehicle

This paper deals with a residual generation for actuators fault detection and isolation (FDI) of electric vehicle with decentralized control. The FDI approach is model based diagnosis, where the modeling step is done using bond graph theory, in order to model separately the considered vehicle components with their nonlinearity, then to show explicitly the interaction between the different sub-models. The system nonlinearities are considered as unknown inputs which are estimated by a nonlinear observers based on the super-twisting algorithm. Due to the unmeasured flow variables of the bond graph mechanical part, the same algorithm is used for mechanical states reconstruction. Then, a parity space approach is implemented for fault detection and isolation of the actuators, in order to prevent the critical accidental situations and to propose some reconfigurable control solutions. Experimental tests are done respectively on the real vehicle in off-line and online modes.