Disturbance Attenuation Observer in Time-Delay Nonlinear Systems with Application to Automotive Engine Fault Detection

A Time Delay Observer (TDO) is proposed in order to attenuate the effects of system and output disturbances on estimation error dynamics. This is an important issue in robust fault detection because the effect of disturbances on the error dynamics could lead to false alarms which must be avoided as much as possible. The principle here is that the output disturbance is attenuated by its immediate past value, therefore, even if the difference between the current disturbance and its last sampling value is amplified by observer gain K, its effect on estimator’s error dynamics would still be small. In the meanwhile system disturbance is attenuated by the additional term multiplied by system disturbance gain matrix. Moreover, the implementation of this TDO approach requires less on-line computation when compared to Proportional Integral (PI) observer. The TDO fault detection approach is then applied to detect faults in an automotive engine. Simulations shows that this TDO fault detection strategy can work very well.