基于多特征信息融合的模拟电路故障预测

摘要

In order to solve the insufficient of fault prediction by using single information and improve prediction accuracy, a method of fusing data to analog circuit fault prediction is proposed, extracting multiple points' various faults characteristics of analog circuits, and carrying on the time series analysis, using ARMA model research on its forecast process, and turning the forecast results into analog circuit faults occur probability. Finally, the gotten data are weighted fusion. Experimental results show that the proposed method overcomes insufficient by using single information, and it's applicable to the hard faults and soft faults that device parameters offset is small, the fault prediction accuracy is high.%为解决利用单一信息进行故障预测的不足以及提高模拟电路故障预测的准确度,提出了一种将信息融合应用到模拟电路故障预测中的方法.提取模拟电路多个测点的多种故障特征量,对其进行时间序列分析,采用ARMA模型研究其预测过程,将得到的预测结果转换为模拟电路故障发生的概率,最后将得到的多个数据进行加权融合,实现了基于多特征信息融合的模拟电路故障预测.模拟实验结果表明:所提方法克服了利用单一信息预测方法的不足,对模拟电路硬故障与元件参数偏移较小的软故障均适用,故障预测准确率高.