基于K-S和加权证据理论的复合故障诊断方法

Concurrent Fault Diagnosis Method Based on K-S and Weighted Evidence Theory

针对复合故障难以识别的问题,提出双样本组合Kolmogorov-Smilnov检测对两条量纲一的指标曲线进行匹配,并通过加权证据理论方法对故障信息进行融合的新方法。实验结果表明:在双样本K-S检验中,峭度指标累积分布函数曲线匹配的准确率高于其他指标;通过D-S证据理论融合故障信息后,准确率虽无明显提高,但可以有效减小复合故障的范围,因此通过再分类可以提高准确率。

In view of the recognition difficulty of concurrent fault diagnosis,a new method is put forward which combines the two dimension one index curves using dual sample Kolmogorov-Smilnov（ K-S） test,then the fault information is fused using the weighted evidence theory. The experimental results show that accurate rate of matching of the kurtosis index accumulation distribution functional curve is relatively higher than others by dual sample of K-S test. Though the accuracy rate was not obviously improved after Dempster-Shafer（ D-S） evidence theory fuses fault information,it can effectively limit the scope of concurrent fault. Therefore the accuracy can be improved by reclassification.