基于ELM-MPLS的质量相关故障检测方法研究  

Research on the Quality-related Fault Detection Method Based on ELM-MPLS

作　　者：张蕊 李强 杨治艳 杨茜[5] 赵美枝 ZHANG Rui;LI Qiang;YANG Zhiyan;YANG Xi;ZHAO Meizhi(CEPREI,Guangzhou 511370,China;Lanzhou Wanli Aviation Electromechanical Co.,Ltd.,Lanzhou 730070,China;AECC Guiyang Engine Design and Research Institute,Guiyang 550081,China;The Ministry of Industry and Information Technology Key Laboratory of Quality and Reliability Engineering Technology of Civil Aircraft and Aero-Engine,Guangzhou 511300,China;Beijing Institute of Precision Mechatronics and Controls,Beijing 100076,China;School of Missile Engineering,Rocket Force University of Engineering,Xi’an 710025,China;School of Physics and Electronic Science,Shanxi Datong University,Datong 037009,China)

机构地区：[1]工业和信息化部电子第五研究所,广东广州511370 [2]兰州万里航空机电有限责任公司,甘肃兰州7300701 [3]中国航发贵阳发动机设计研究院,贵州贵阳550081 [4]民用飞机及航空发动机质量与可靠性工程技术工业和信息化部重点实验室技术团队,广州511300 [5]北京精密机电控制设备研究所,北京100076 [6]火箭军工程大学导弹工程学院,陕西西安710025 [7]山西大同大学物理与电子科学学院,山西大同037009

出　　处：《电子产品可靠性与环境试验》2023年第3期85-90,共6页Electronic Product Reliability and Environmental Testing

摘　　要：在工业生产过程中,很多产品的质量难以使用传感器来直接测量。因此,出现了偏最小二乘和潜在结构全投影等过程监测方法,但它们的质量相关故障检测效果并不理想。为此,提出了一种基于极限学习机(ELM)与改进的偏最小二乘(MPLS)的质量相关故障检测方法。该方法首先采用ELM进行质量预测,然后根据预测结果进行过程监测,最后采用田纳西伊斯士曼(TE)过程验证其质量相关故障检测的有效性。实验结果表明,ELM-MPLS不仅可以实现质量预测还提高了质量相关故障检测率,具有良好的质量相关故障检测效果。In the process of industrial production,the quality of many products is difficult to be measured directly by sensors.Therefore,process monitoring methods such as partial least squares and total projection to potential structure are developed,but their quality-related fault detection effects are not ideal.To this end,a quality-related fault detection method based on extreme learning machine(ELM)and modified partial least squares(MPLS)is proposed.In this method,ELM is used for quality prediction firstly.Then,process monitoring is carried out according to the prediction results.Finally,TE process is used to verify the effectiveness of quality-related fault detection.The experimental results show that ELM-MPLS can not only achieve quality prediction but also improve the detection rate of quality-related faults,and has a good quality-related fault detection effect.

关 键 词：极限学习机 改进的偏最小二乘 质量预测 过程监测 故障检测 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]