基于AFCM-SVM的滚动轴承退化状态评估与剩余寿命预测

Degradation State Assessment and Life Prediction of Rolling Bearings Based on AFCM-SVM

针对支持向量机模型状态数需要人为设定的不足,提出了一种基于自适应模糊C均值-支持向量机(AFCM-SVM)的滚动轴承退化状态评估与剩余寿命预测方法。该算法采用相对特征建立敏感特征数据集,利用聚类评价指标构造自适应函数,实现了模型聚类结果的自动更新,获得了轴承运行过程中的最佳状态数;基于AFCM-SVM模型与各个运行状态的一一对应关系,确定轴承在不同退化状态下的时间间隔,实现轴承的健康等级评估与寿命预测。根据美国NSFI/UCR智能维护中心提供的滚动轴承全寿命数据对所提算法进行了验证。结果表明,不受轴承个体差异的影响,AFCM-SVM能有效实现自动聚类,识别结果符合轴承退化演变规律;与分层狄利克雷(HDP)和K-means算法相比,AFCM-SVM具有更快的运算速度和更准确的辨识能力。

Aiming at the deficiency of the Support Vector Machine model whose states must be determined by users in advance,a method for assessing the degradation state and predicting the residual life of rolling bearings based on Adaptive Fuzzy C-means Clustering-Support Vector Machine(AFCM-SVM)was proposed.By using relative features to build sensitive feature data sets and clustering evaluation indexes to construct adaptive function,the clustering results was automatically updated and the optimum states number of rolling bearings during the operation was obtained.Based on the corresponding relation between the AFCM-SVM model and each running state,the time interval of rolling bearings under different degradation states was determined,the health grade assessment and life prediction of rolling bearings was realized.The proposed algorithm was verified based on the life-cycle data of rolling bearings provided by NSFI/UCR Intelligent Maintenance Center in the United States.The results show that AFCM-SVM can effectively realize automatic clustering without the influence of bearing individual differences,and the recognition results accord with the evolution law of bearing degradation.Compared with hierarchical Dirichlet Process(HDP)and K-means algorithms,AFCM-SVM has faster operation speed and more accurate identification ability.