摘要
针对滚动轴承的故障诊断问题,提出了一种采用案例推理(CBR)的诊断方法.为了解决检索相似案例时案例属性多、人工确定关键属性及其权重困难的问题,提出了一种Filter/Wrapper复合型特征选择算法,用邻域粗糙集算法粗选属性,用遗传算法进一步精选属性和优化权重,并有效地解决了邻域粗糙集算法中需要人工确定邻域大小的问题.以滚动轴承运行时的振动信号为基本信息,建立了滚动轴承案例库,从案例库中检索与问题案例相似的历史案例,并根据这些历史案例来判定问题案例的故障类别.试验结果表明,故障诊断的正确率达到100%,故障位置诊断的正确率达到93.3%,且算法具有较好的稳定性.
The case-based reasoning approach is introduced into rolling bearing fault diagnosis.To solve the complexity of feature selection and weights optimization,a Filter/Wrapper integrated feature selection algorithm is proposed.Neighborhood rough set algorithm is applied to select essential features from the feature candidate set,then genetic algorithm is applied to refine the essential feature subset.This method solves the problem of determining the size of neighborhood manually in neighborhood rough set algorithm.Genetic algorithm is also used in feature weights optimization.With the runtime vibration signal of rolling bearing as the basic information,a rolling bearing fault case database is constructed.The historical cases similar to the problem case are recalled and chosen to decide the fault type.The database experiment shows the higher efficiency and accuracy for essential attributes and weights in fault diagnosis.
出处
《西安交通大学学报》
EI
CAS
CSCD
北大核心
2011年第11期79-84,共6页
Journal of Xi'an Jiaotong University
基金
国家自然科学基金资助项目(51035007)
关键词
案例推理
滚动轴承
故障诊断
case-based reasoning
rolling bearing
fault diagnosis
