| Gears are widely used in machinery and gear fault is one of the main causes of machinery failure. Wear condition monitoring of gear pair is very necessary to detect early fault and reduce losses. Oil monitoring and vibration monitoring are the most common monitoring methods. A large number of monitoring data including general parameters of wear particles, fractal parameters of wear particles and vibration parameters will be generated in the monitoring process. Further study on how to accurately determine the wear condition of gear pair from a large number of monitoring data remains to be made. In this article, application of data mining technology in gear pair wear condition monitoring was studied, content of this article is as follows:A novel method of identifying scaling region of wear particles was developed based on simulated annealing K-Means clustering algorithm to overcome the local optimum problem in identification of scaling region, which can significantly improve the calculation accuracy of the fractal dimension of wear particles.The wear test of spur gear pair was conducted. The oil parameters, vibration parameters and wear rate of gear pair in the entire lifecycle of gear were obtained and then filtered and reduced by correlation analysis and principal component analysis. A new gear pair monitoring parameters set was proposed.Since the traditional wear condition partition method based on the wear rate was not accurate, a new wear condition partition method was proposed. The entire lifecycle of gear pair was divided into six wear conditions according to the new parameters set by the new method, the corresponding monitoring parameters set to different wear condition were obtained. The new partition method includes and refines the traditional method.Association rules between the monitoring parameters distribution and gear pair wear condition were extracted based on association rules algorithm, test data acquired from gear pair wear test were applied to association rules and results show that the wear conditions identification accuracy rate of proposed mining model based on monitoring parameters reaches90%, which can effectively identify gear pair wear condition. |
