Design of wheelset vibration diagnosis based on UKF

Vibration parameter of the train wheelset is a key technology of the estimated fault prediction and diagnosis of train. According to the nonlinear problem for the state signal of wheelset vibration diagnostic system, unscented transformation is introduced into Kaman filter. The UKF method could be applied to addresses the estimation inaccuracy of linear Kalman filter and the filtering divergence problem caused by error covariance matrix non-positive in general unscented Kalman filter. The results obtained from trail vertical irregularity simulation show that, UKF features rapid convergence, higher stability and accuracy, making it become an efficient estimation of diagnosis method for Vibration Diagnostic of train´s wheelset.