基于加权粗糙集的代价敏感故障诊断方法

在故障诊断领域，粗糙集已成为一种有效的不一致信息处理工具，然而当故障诊断存在明显的诊断代价差异时，经典粗糙集方法由于无法考虑先验知识而不能取得满意的效果。通过引入样本对象的主观加权，该文提出加权粗糙集的学习方法，设计了加权属性约简和加权规则提取算法，为粗糙集学习提供一种引入先验知识的途径。基于提出的加权粗糙集学习方法，开展了代价敏感故障诊断的研究，并进行了汽轮机振动的代价敏感故障诊断实验。实验表明，基于加权粗糙集方法的代价敏感故障诊断能优先选取高代价故障的关键征兆，并且使提取的规则集中高代价故障的规则具有更高的规则支持度和可信度，当诊断存在不一致的情况下，该方法倾向于将故障诊断为高代价故障，从而降低诊断代价。

Cost-sensitive Fault Diagnosis Based on Weighted Rough Sets

In the realm of fault diagnosis,rough sets have been a powerful tool to deal with the inconsistent information.However,when the misdiagnosis costs of different faults are unequal,classical rough sets can not acquire a satisfying result due to the absence of a mechanism considering the apriori knowledge.Through the introduction of subjective weights on data,this paper proposed a weighted rough set learning method to consider the apriori knowledge in rough sets,where an algorithm of weighted attribute reduction and an algorithm of weighted rule extraction were designed respectively.Based on weighted rough sets,the method of cost-sensitive fault diagnosis was provided and experiments on the cost-sensitive fault diagnosis of vibration faults of steam turbine was carried out.The results show that in the weighted rough set based cost-sensitive fault diagnosis,the key symptoms of high-cost faults are selected preferentially,and the bigger factors of support and confidence are obtained for the rules of high-cost faults in the generated rules.When the outputs of fault diagnosis are inconsistent,the weighted rough set based cost-sensitive fault diagnosis is inclined to output the high-cost fault and decreases the overall costs of fault diagnosis.