针对测试信息不足造成模拟电路故障诊断准确率较低的问题,为充分利用有限测试信息,提出一种模拟电路故障诊断信息融合新方法.首先,将采集的故障样本集变换到不同特征空间,然后利用所提出的马氏距离分布熵求取各特征空问的相对优势分类集,在此基础上,定义相对优势属性约简提取各特征空间的局部最佳可分性信息,最后,对基分类器结合所提出的自适应类模糊密度赋值方法进行模糊积分融合.国际标准电路故障诊断实例表明,所提方法能有效提高模拟电路的故障诊断率.
To solve the problem that test information in analog circuit diagnosis is always insufficient, a new algo rithm of information fusion is presented to improve the diagnosis accuracy in analog circuit. Firstly, fault samples are transformed in different feature spaces, and the relative dominance class set for each feature space is computed by distribution entropy of mahalanobis distance. Subsequently, the relative dominance attribute reduction is defined to extract local prime separability information. And finally, the trained basic classifiers are ensembled by fuzzy inte gral. The fault instance indicates that the presented method can inprove diagnosis accuracy effectively.