基于DPC-SVDD的制造过程异常诊断  

Anomaly diagnosis of manufacturing process based on DPC-SVDD

作　　者：沈维蕾[1] 杨雪春 吴善春 SHEN Weilei;YANG Xuechun;WU Shanchun(School of Mechanical Engineering,Hefei University of Technology,Hefei 230009,China)

机构地区：[1]合肥工业大学机械工程学院,安徽合肥230009

出　　处：《合肥工业大学学报（自然科学版）》2022年第4期433-439,共7页Journal of Hefei University of Technology：Natural Science

基　　金：国家自然科学基金资助项目(51975003)。

摘　　要：文章针对生产过程中质量数据分布类型未知引起的传统质量控制图异常检测精度低的问题,提出结合支持向量数据描述(support vector data description,SVDD)和密度峰值聚类(density peaks clustering,DPC)的制造过程异常检测方法。采用DPC算法对质量特征数据进行聚类分析,将聚类结果作为模型输入训练得到各类超球体中心和决策边界;以此建立基于内核距离的DPC控制图,实现对生产过程质量波动的实时监控;最后将该控制图应用到再制造曲轴生产过程监控中。结果表明,该文提出的DPC控制图可以有效监测再制造曲轴生产过程质量异常波动,验证了该检测方法的可行性和有效性。Aiming at the problem of low accuracy of traditional quality control chart anomaly detection caused by unknown quality data distribution type in production process,this paper proposes an anomaly detection method for manufacturing process that combines support vector data description(SVDD)and density peaks clustering(DPC).Firstly,the DPC algorithm is used for clustering analysis of quality characteristic data,and then the clustering results are used as model input to train various hypersphere centers and decision boundaries,so as to establish DPC control chart based on kernel distance and realize real-time monitoring of quality fluctuation in the production process.Finally,the control chart is applied to the production process monitoring of remanufactured crankshaft.The results show that the DPC control chart can effectively monitor the abnormal quality fluctuation in the production process of remanufactured crankshaft,thus verifying the feasibility and effectiveness of the method.

关 键 词：支持向量数据描述(SVDD)算法 密度峰值聚类(DPC)算法 异常检测 密度峰值聚类(DPC)控制图 

分 类 号：TH164[机械工程—机械制造及自动化]