集合结构的实时故障诊断方法研究

为提高工业系统故障诊断的速度和准确性,提出一种由数据实时预处理、增量式故障特征提取、在线分类与聚类组成的实时故障诊断集合结构。基于该结构,提出了一系列实时故障诊断方法,开发了相应的实验研究平台。以青霉素实验过程为例,实验采用提升小波方法预处理、递归主元方法特征提取、递归最小二乘支持向量机分类的集合型实时故障诊断方法进行实验研究。实验结果表明,集合结构的实时故障诊断方法的性能优于传统的单一结构的故障诊断方法。

The Research of Real-time Ensemble Fault Diagnosis

To improve the speed and accuracy of the industrial system fault diagnosis,an ensemble framework of real-time fault diagnosis is proposed. Based on the structure,a series of real-time fault diagnosis methods are put forward. The corresponding experimental research platform is developed. Taking Penicillin experimental process as an example,the realtime fault diagnosis method with lifting wavelet,recursive PCA,and recursive least square support vector machine are conducted. Experimental results showthat the performance of real-time ensemble-based approaches is superior to that of traditional fault diagnosis methods with single structure.