Investigation of Multivariate Profile Monitoring on Complex Thin-walled Components Batch Machined using a Sliding Time Window Cluster Method
Investigation of Multivariate Profile Monitoring on Complex Thin-walled Components Batch Machined using a Sliding Time Window Cluster Method
复制标题
使用滑动时间窗口聚类方法对复杂薄壁部件批量加工进行多变量轮廓监测的研究
DOI：
10.1080/0951192x.2020.1858500
复制
发表时间：
2021-02
影响因子：
4.1
通讯作者：
Xie Shengtuo
中科院分区：
文献类型：
--
作者：
Wang Pei;Zhang Yang;Feng Bo;Liu Dekun;Xie Shengtuo
摘要针对复杂薄壁零件批加工过程中质量监测精度低、波动检测灵敏度低的问题，提出了多变量轮廓度监测方法。采用扩展误差流方法建立了多工序加工波动的轮廓表示法，可以用误差源来描述多零件生产周期的波动和生产线的衰退。在此基础上，建立了基于一步超前预测误差和T平方(OSFE-T平方)的批量复杂薄壁零件加工过程监控统计量。然后，基于OSFE-T平方监控模型，采用基于滑动时间窗聚类的方法，建立了基于OSFE-T平方监控模型的批量零件不同操作的轮廓监控方法。通过蒙特卡罗模拟，分别对质量监测方法的正确分类分数、灵敏度、特异度、假阳性和假阴性进行了分析，证明了该方法具有较好的性能。最后给出了一个加工仿真实例，其中包括10道加工工序，验证了该加工误差控制方法的有效性。与其他替代方法相比，本文提出的方法在波动检测的灵敏度方面具有更大的优势。
ABSTRACT Aiming at solving the problems of low precision of quality monitoring and low sensitivity of fluctuation detection in the batch machining process of complex thin-walled parts, the multivariate profile monitoring approach was proposed. The extended error stream method was used to establish profile representation of the multi-operation machining processing fluctuation, which could describe the fluctuation of multi-part production cycle and the decline of production line represented by error sources. On this basis, the monitoring statistics of the batch complex thin-wall part machining process were built based on one-step ahead forecast error and Tsquare (OSFE-T square). Then, the profile monitoring method was developed based on the OSFE-T square monitoring model of batch parts on the different operation using the sliding time window cluster-based method. Through the Monte-Carlo simulation, the fraction correctly classified, sensitivity, specificity, false positive and false negative of the quality monitoring method were analyzed respectively, which proved that the proposed method had better performance. Finally, amachining simulation instance was presented including ten operations demonstrated the effectiveness of the machining error control method. Compared with other alternative methods, the proposed method in this paper has more advantages in the sensitivity of fluctuation detection.