基于多元图可视化分析和人机交互的设备故障诊断方法研究

摘要

Aiming at the limitations of the data-oriented fault dignosis method, this paper presents a novel fault diagnosis technology which is based on the visualization analysis of empirical samples' fault patterns expressed by multivariate graphs and the human-computer interaction (HCI) according to the basic theories of multivariate graph expression. It realizes the combination of the data-oriented machine fault diagnosis and the object-oriented fault diagnosis by experts' participation in the fault dignosis process, thus overcoming the obstacles in single mechine learning. The fault diagnosis technology based on multivariate graphical visual analysis and HCI was tested by the experiments using the fault database of the machine learning repository, Irvine, University of California (UCI). The experimental results show the process of the visual analysis and HCI can improve the aecuracy of the data-oriented prosing fault diagnosis.%针对面向数据的故障诊断方法的局限性,根据多元图表示基本理论,提出了基于多元图表达的经验样本故障模式可视化分析和人机交互(HCI)的故障诊断技术,该技术通过专家参与机器自动识别诊断过程实现了面向对象领域的故障诊断方法和面向数据的故障诊断方法的有效结合,克服了单一机器学习的局限性.采用国际标准UCI数据库中的故障数据库进行了数据实验,实验结果显示,信息可视化人机交互过程有利于提高面向数据的故障诊断研究的分类正确率.