基于证据理论与粗糙集集成推理策略的内燃机故障诊断

Fault Diagnosis Based on Reasoning Integration of Rough Sets and Evidence Theory

证据理论是处理不确定性问题的有效工具,但其依赖专家知识提供证据,并要求各证据体相互独立,致使实际应用困难。针对此问题,提出了证据理论与粗糙集(Rough Sets)集成的方案:基于系统聚类的方法对特征数据离散化;基于变精度粗糙集条件熵的方法对决策表约简;约简后各条件属性作为证据并计算基本可信度分配;利用证据理论组合规则对各证据进行合成及决策。柴油机的实际诊断结果验证了将证据理论与粗糙集相结合进行故障诊断的良好效果。

Evidence theory is an effective tool in dealing with uncertainty questions. It relies on the expert knowledge provide evidences, needing the evidences to be independent, and this makes it difficult in application. To solve the problem, a hybrid system of rough sets and evidence theory is proposed. Firstly, the continuous attributes in decision table are discretized with systemic clustering algorithm. Secondly, simplifications are made based on VPRS conditional entropy. Thus, the basic probability assignment for all evidences can be calculated. Thirdly, Dempster＇s rule of combination is used, and a decision-making is given. Diagnosis in a diesel engine gives a better result by combining evidence theory with rough sets.