The Study of Large Equipments Condition Maintenance Policy Based on Kalman Filtering

Abstract

Aimed at the problem that because common proportional hazards model (PHM) cannot fuse new failure data of long-life complex equipment, which features a small-sample, the reliability estimation accuracy will decline, a new condition-based maintenance strategy based on dynamic PHM was proposed. Kalman filtering theory was adopted to fuse in-time new failure data and expand sample size. Extended Kalman filtering method was used to solve the nonlinearity of the observation equation of PHM and then its regression coefficient was online updated, according to which the residual life was estimated and the optimal maintenance decision was made. Finally, the condition monitoring data and historic operation data of a certain kind of wind power gearbox were used to validate this method. The result indicates that this method has good dynamic estimation ability under the condition of small sample with a 20.6% increase in the accuracy of regression coefficient estimation and a 8.7% decrease in optimal preventive maintenance interval estimation error.