摘要
为解决实际工程环境中因轴承故障数据缺失引起的数据不平衡,进而影响模型诊断的问题,提出了不平衡数据处理技术在轴承故障诊断中的应用,即使用少量数据,训练出一个能够诊断多种轴承故障的模型。针对不同种类故障数据的边界混淆及类内不平衡,首先对少数类样本进行高斯混合模型(GMM)聚类,根据簇密度分布函数使用GMMSMOTE进行权重采样,然后针对边界混淆问题使用Tomek’s Link数据清洗技术进行边界混淆样本数据清洗,形成类内类间平衡的数据集,结合超参数优化的支持向量机(SVM)模型进行诊断分析。实验结果表明,该方法效果良好,在轴承的故障诊断中有较好的实际工程意义和推广性。
In order to solve the problem that the data imbalance caused by the missing of bearing fault data in actual engineering environment affects the model diagnosis,the application of unbalanced data processing technology in bearing fault diagnosis is proposed,that is,using a small amount of data,training a model that can diagnose multiple bearing faults.In order to solve the boundary confusion and intraclass imbalance of different kinds of fault data,Gaussian mixture model(GMM)was used to cluster a few samples,and GMMSMOTE was used for weight sampling according to the cluster density distribution function.Then,Tomek’s Link data cleaning technology was used to clean the boundary confusion sample data.The data set of intraclass equilibrium was formed,and the diagnosis analysis was carried out with the support vector machine(SVM)model optimized by hyperparameter.The experimental results show that this method is effective and has good practical engineering significance and generalization in bearing fault diagnosis.
作者
王振亚
刘韬
王廷轩
杨永灿
WANG Zhenya;LIU Tao;WANG Tingxuan;YANG Yongcan(Faculty of Mechanical and Electrical Engineering,Kunming University of Science and Technology,Kunming 650500,China)
出处
《机械与电子》
2021年第6期29-34,共6页
Machinery & Electronics
基金
国家自然科学基金资助项目(52065030)
中央引导地方科技发展资金资助项目(202007AC110001)。
