摘要
针对机械故障诊断知识的近似最优属性约简不惟一要求,从得到多个满足分类精度的属性约简集合的目标出发,提出了一种基于特征选择和变精度粗集的属性约简方法。该算法将特征选择过程,从由完全属性集开始的属性删除方式转变为从核属性出发以增加关键属性为主的多目标方式;同时引入决策属性支持度,保证了约简结果对于论域对象分类的准确率。通过机械故障状态数据的实例应用,表明该方法可获得旋转机械各类典型故障的关键属性,得到了给定准确度下的多个约简集合。
Aiming at the multi-objective demands of optimal attributes reduction for rotating machinery fault diagnosis knowledge,and starting with obtaining diverse attribute reduction sets of meeting given classification precision,an attribute reduction approach based on attributes selection and variable precision rough set model is put forward.The procedure of the attributes selection in the method is to transform the mode of calculation procedure of attributes reduction from decreasing attributes to increasing key attributes.Moreover,the results of attributes reduction were unique no longer.In the course of the calculation of attribute subsets,decision attribute support degree is introduced to ensure accuracy of reduction results to classification of universe elements.The attribute reduction method was applied to machinery fault data.The results show that the key attributes of all typical faults and the attribute reduction sets for meeting the given accuracy can be obtained.
出处
《机械科学与技术》
CSCD
北大核心
2010年第10期1412-1416,共5页
Mechanical Science and Technology for Aerospace Engineering
基金
国家自然科学基金项目(50875118)
甘肃省教育厅硕导基金项目(0903-11)资助
关键词
变精度粗糙集
特征选择
属性约简
机械故障诊断
variable precision rough set
attribute selection
attribute reduction
machinery fault diagnosis
