摘要
过程数据的相关性是设计和应用控制图进行质量监测时需要解决的关键且极具挑战性的问题。本文利用Transformer的自注意力机制能有效捕捉数据全局信息的优点挖掘数据的相关性,提出基于Transformer的残差EWMA (TREWMA)控制图,实现了对自相关过程的过程均值监测。实验通过仿真和真实数据演示了TREWMA控制图的实施;其结果表明,TREWMA控制图在监测自相关过程的均值漂移时,展现出比现有控制图更好的性能。Correlation in process data is the key and challenging problem to be solved when designing and applying a control chart to carry out a Quality monitoring. Bying utilizing the self-attention mechanism of Transformer to learn the correlation and overall information of data, propose a Transformer-based residual EWMA (TREWMA) control chart, and achieve process-mean-monitoring for autocorrelation process. Experiments demonstrate the implementation of the TREWMA control chart through simulation and real data;the results show that the TREWMA control chart exhibits better performance than existing control charts in monitoring the mean shift of autocorrelation processes.
出处
《应用数学进展》
2024年第12期5330-5337,共8页
Advances in Applied Mathematics
基金
陕西省自然科学基础研究计划资助项目(2024JC-ZDXM-23)
长安大学中央高校基本科研业务费专项资金资助项目(310812163504)。
