基于加权的多无量纲指标信息融合故障诊断方法

Multi-dimensionless Parameters Fusion Method Based on Improved D-S Evidence Theory

针对传统D-S证据理论的组合规则存在高冲突问题,提出了一种改进加权的D-S证据理论,结合时域中的5种常用无量纲指标,对旋转机械进行了故障诊断。首先,由于不同的无量纲指标对故障的敏感程度不同,可将各无量纲指标自身固有的诊断可靠度作为证据的先验权重;其次,计算各证据间的证据距离,构建证据距离矩阵以及相似矩阵,获得各证据的后验权重。通过分析多级离心风机故障诊断实验平台的实验结果,发现采用复合权重的多无量纲指标信息融合方法的诊断结果可信度明显提高,并且高冲突证据的融合问题也能得到有效解决,证明了该方法的可行性和有效性。

Aiming at the invalidation problem of Dempster - Shafer evidence combination mle with high conflict,a fault diagnosis method of five common dimensionless parameters information fusion which is based on the D - S algorithm improved by the weighted concept is prop-osed .Firstly, for multi - dimensionless parameters vary in terms of fault sensitivity, prior weights are obtained by the basis of inherent reli-ability of multi - dimensionless parameters diagnosis information, and a similarity matrix between evidences is constructed by calculating the distance of evidences as posterior weights. Finally, the experimental results for multistep centrifugal machine demonstrate that, compared with other methods, the new algorithm can significantly improve credibility of fault diagnosis and reduce influence of conflict evidence effec-tively ,which indicates the feasibility and validity of this method.