摘要
文章在分析比较几种诊断方法的基础上,根据滚动轴承的故障特点,建立了SVM"一对一"聚类结构并对滚动轴承故障进行诊断;该方法基于结构风险最小化,能较好地解决小样本学习问题,避免了人工神经网等智能方法在对滚动轴承状态进行诊断时所表现出来的过学习、泛化能力弱等缺点;利用SVM"一对一"聚类结构对滚动轴承故障类别进行投票,降低了单个支持向量机的误判概率;具体实验结果表明,该聚类结构对滚动轴承的故障类别具有很高的诊断精度,能够取得理想的聚类效果。
Based on the comparative analysis of several diagnosis faults of rolling bearings, the＂one against one＂ SVM clustering methods and the characteristics of the structure is constructed and the faults of rolling bearings are diagnosed. The presented method is based on structure risk minimization, so it can solve the small-batch learning better and avoid such disadvantages as over-training and weak normalization capability in the application of artificial neural networks and other artificial intelligence methods to prediction. This proposed method polls according to the faults, thus reducing the probability of making mistakes by a single SVM. Therefore, it is suitable for diagnosis of the faults of rolling bearings. The experiment shows that the model has higher diagnostic accuracy and can achieve ideal diagnostic effect.
出处
《合肥工业大学学报（自然科学版）》
CAS
CSCD
北大核心
2009年第1期4-8,共5页
Journal of Hefei University of Technology：Natural Science
基金
国家自然科学基金资助项目(70672096)
关键词
滚动轴承
故障诊断
支持向量机
“一对一”聚类结构
rolling bearing
fault diagnosis
support vector machine（SVM）
＂one against one＂ clustering structure
