【作者】 孙来军； 沈永良；

【Author】 Sun Laijun Shen Yongliang (HLJ Province Key Lab of Senior-Education for Electronic Engineering,Heilongjiang University,Harbin 150080,China)

【机构】 黑龙江大学电子工程黑龙江省高校重点实验室；

【摘要】 传统故障诊断大多采用单信号单处理方法,而单信号的测量、诊断存在很多不确定性,影响诊断的准确率。本文在介绍传统融合方法的基础上,建立了多 SVM 输出分类信息的多级融合模型。该模型采用信号内和信号外两级融合,信号内融合是将信号的不同提取参量经不同 SVM 分类后的输出,经模糊化处理后,利用 D-S 证据理论进行融合;在决策级融合中, 利用模糊信息融合技术实现多信号的决策融合。在高压断路器故障诊断中的应用证实,该模型在取得较好故障分类效果的基础上减少了对经验知识的过度依赖,避免了单一多级融合中出现的过融合。更多还原

【Abstract】 Single-signal and signal-method diagnosis is applied in traditional fault diagnosis,but the meas- urement and diagnosis with that has so much indetermination that diagnosis accuracy is affected greatly. Based on the introduction of traditional data-fusion method,the paper presents a novel classification for HV circuit breakers which can solve this problem,and a multi-SVM multi-level information fusion model is established.In this model,double-level fusion is adopted.For parameters from the same signal but with different method,D-S evidence theory is used to fuse output of different SVM classification according to every kind of parameters;In the fusion of decision,multi-signal decision fusion is realized using fuzzy in- formation fusion technology.At the last,the effective application in the fault diagnosis of HV circuit breakers approves that,this new fusion model can reduce excessive dependence on experience knowledge, and avoid over-fusion which usually appears in the ancient multi-level single fusion.更多还原