For the unmeasured fault parameters estimation and prediction problem of nonlinear systems, the state and parameter joint estimation algorithm based on multiple fading factors strong tracking unscented Kalman filter(MSTUKF) is presented. Multiple fading factors of MSTUKF are introduced to improve the tracking ability for fault parameters with the unknown changing function. With the fault parameters estimation of MSTUKF, the constrained AR prediction model is updated by recursive least squares. Online estimation and prediction of fault parameters are realized by using the proposed method. Simulation results show that the fault parameters estimation ability of MSTUKF is better than that of UKF and single fading factor strong tracking UKF, and the predicting accuracy of constrained AR model is superior to the unconstrained AR model.