基于对抗神经网络的力学试验机故障诊断系统设计  

Design of a fault diagnosis system for mechanical testing machines based on generative adversarial networks

作　　者：国雪 李治淼[1] 崔天奇 曹梦雨[1] 杜相辉 李鸿婧 GUO Xue;LI Zhi-miao;CUI Tian-qi;CAO Meng-yu;DU Xiang-hui;LI Hong-jing(College of Mechanical Science and Engineering,Northeast Petroleum University,Daqing 163318,Heilongjiang Province,China;School of Mechanical and Electronic Control Engineering,Beijing Jiaotong University,Beijing 100044,China;Financial Assets Department of the Fourth Oil Production Plant of Daqing Oilfield Co.,Ltd.,Daqing 163319,Heilongjiang Province,China)

机构地区：[1]东北石油大学机械科学与工程学院,黑龙江大庆163318 [2]北京交通大学机械与电子控制工程学院,北京100044 [3]大庆油田有限责任公司第四采油厂财务资产部,黑龙江大庆163319

出　　处：《信息技术》2024年第11期69-76,共8页Information Technology

基　　金：东北石油大学引导创新基金(2020YDL-23)。

摘　　要：针对高校力学试验机随机故障频发问题,设计了力学试验机故障诊断系统。针对此类试验机使用间歇长、离散性强等特征,提出了基于离散性数据识别故障的设计思路。建立了机器学习知识库,基于对抗神经网络(GAN)理论设计了力学试验机故障诊断算法并建立了故障诊断系统,评价了故障诊断系统的性能指标。结果表明,所建立故障诊断系统在实验中的最低精确率、准确率和召回率分别达到96.12%、96.51%和95.15%,最高误识率仅为3.96%,性能满足使用要求。In response to the frequent occurrence of random faults in mechanical testing machines in universities,a fault diagnosis system for mechanical testing machines is designed.A design thought for identifying faults based on discrete data is proposed to address the characteristics of long intermittency and strong discreteness in the use of such testing machines.A machine learning knowledge base is established,and a mechanical testing machine fault diagnosis algorithm is designed based on Generative Adversarial Networks(GAN)theory.A fault diagnosis system is established,and the performance indicators of the fault diagnosis system are evaluated.The results show that the minimum accuracy,accuracy,and recall of the established fault diagnosis system in the experiment reach 96.12%,96.51%,and 95.15%,respectively,with a maximum error recognition rate of only 3.96%,indicating the established fault diagnosis system meets the usage requirements.

关 键 词：力学试验机 故障诊断 状态维修 机器学习 对抗神经网络 

分 类 号：TH871[机械工程—仪器科学与技术]