Fault diagnosis approach based on multiple model estimator with simplified CDKF

As the direct application of traditional Kalman filter to fault diagnosis of the nonlinear system usually results to low estimation accuracy, a new fault diagnosis approach was proposed. The method combines the multitude model adaptive estimation and the simplified central difference Kalman filter. Therefore, it achieves on-line fault detection rapidly, and makes the state estimation values converge to real values correctly benefits from the replacement of the Jacobian matrix calculation by central difference transformation. Moreover, the repeated process of solving measurement equations and variances is avoided. In the presents of various actuator faults, the simulation results indicate the effectiveness and rapidity of the proposed algorithm compared with the other filters.