基于支持向量异常检测算法的新故障检测

Novel Fault Detection Based on Support Vector Novelty Detection Algorithm

对于一个智能故障分类与诊断系统,需要有检测新出现的故障模式的能力。采用一种支持向量异常检测算法,即支持向量数据描述(SVDD),建立已知故障类训练样本的描述模型,并用于检测新的训练中未见的故障类样本。以实测的轴承多种故障类样本为例,结果表明:通过选取合适的算法参数,SVDD对设定的新故障类样本的检测率达88%—100%,同时对已知故障类样本的识别率达83%—94%。

The ability to detect a new fault class can be a useful feature for an intelligent fault classification and diagnosis system. In this paper, a support vector novelty detection algorithm, the support vector data description （SVDD）, was adopted to represent known fault class samples, and to detect new fault class samples. The experiments on real multi-class bearing faults data showed that the propesed approach can effectively detect prescribed ＇unknown＇ fault samples with detection rated 88%- 100%, and identify known faults samples with recognition rated 83%-94% via choosing appropriate SVDD algorithm pararneters.