Parameter estimation based fault diagnosis of uncertain nonlinear three tank system using HOSM differentiator observer

A synthesis of parameter estimation based fault diagnosis scheme for uncertain nonlinear systems using higher order sliding modes is presented. The uncertain parameters of nonlinear system are estimated using an uncertainty estimator or observer. The estimated parameters carry information about system and actuator faults. A fault signal representing the dynamics of the fault is reconstructed using the estimated parameters and comparing it with the nominal (fault free) values of the corresponding parameters. For the fault free case, the magnitude of the fault reconstructed signal `residual' remains close to zero and a fault declared when the magnitude of the reconstructed fault signal becomes non-zero and crosses some predefined threshold. To validate the technique, simulations have been carried out for an uncertain nonlinear MIMO Three Tank System. Proposed fault diagnosis technique requires no prior information about the uncertain/unknown system parameters and statistical knowledge of process.