基于SVM与改进AHP-DS证据融合的故障诊断方法

Fault Diagnosis with SVM and Improved AHP-DS Evidence Fusion

针对传统DS(Dempster-Shafer)证据理论应用于高冲突证据融合中失效的问题,提出一种改进层次分析法(Analytic Hierarchy Process,AHP)结合DS证据理论的故障诊断方法。对多分类SVM的核函数参数和惩罚因子全局寻优,形成基于SVM的3个证据体进行故障模式识别;引入差异性、冲突性和不确定性3项证据冲突衡量标准,构造层次结构模型;对层次分析法进行改进,消除两两比较矩阵构造过程中主观性影响,获得证据体的可信度权重,进行加权证据融合。船舶空调系统故障诊断试验结果表明,该方法能够有效处理高冲突证据融合问题,提升故障的诊断精度。

The improved analytic hierarchy process and Dempster-Shafer evidence theory are jointly used to achieve fault diagnosis when evidences are in high conflict.The parameters of the Kernel function of the multi classification SVM and the penalty factor are globally optimized.The fault diagnosis is conducted based on three evidence bodies from SVM.The grading standards for differences,conflicts and uncertainties of the evidence bodies are defined and the hierarchical structure model is built.The AHP is improved to avoid impact of subjective judgment on the construction of comparison matrixes.The credibility weights for the evidence bodies are decided and fusing of evidence bodies are performed.The process is carried out to diagnose the fault of a ship air conditioning system.