基于增量贝叶斯学习模型的在线电路故障诊断

ONLINE CIRCUIT FAULT DIAGNOSIS BASED ON INCREMENTAL BAYESIAN LEARNING MODEL

现有的贝叶斯分类器用于在线电路故障诊断,能保证诊断精度,但随着样本数增加,学习过程耗时相应增长,将不能实时更新诊断模型。针对这一现状,提出一种基于贝叶斯增量学习的在线电路故障诊断方法。利用核函数主成分分析法对特征数据进行降维,以及灵敏度分析确定敏感元器件,并将增量式贝叶斯学习算法应用于双二阶RC有源滤波器进行故障诊断。通过对增量式贝叶斯学习算法和传统的批量式贝叶斯学习算法进行对比,证明了在精度方面与批量式贝叶斯学习算法保持近似的基础上,增量式贝叶斯学习算法大大缩减了模型更新时间。

The existing Bayesian classifier is used for online circuit fault diagnosis and can ensure the accuracy of diagnosis. However,as the number of samples increases,the learning process takes time to increase accordingly,and the diagnostic model cannot be updated in real time. Aiming at this situation,an on-line circuit fault diagnosis method based on Bayes incremental learning was proposed. The principal component analysis of the kernel function was used to reduce the dimension of the feature data, and the sensitivity analysis determined the sensitive components. The incremental Bayesian learning algorithm was applied to double second-order RC active filter for fault diagnosis. By comparing the incremental Bayesian learning algorithm with the traditional batch Bayesian learning algorithm,it was proved that the incremental Bayesian learning algorithm greatly reduced the model update time when the precision was consistent with the batch Bayesian learning algorithm.