Abstract: To solve the optimization problems of process manufacturing process parameters,such as the complexity of multi process coupling model construction,the difficulty of multi-objective analysis and the difficulty of ensuring real-time and accuracy,a cloud edge linkage optimization method for process manufacturing parameters was proposed by combining Gated Recurrent Unit—Attention(GRU-Attention)with whale algorithm.A cloud edge collaborative architecture for training computing was designed for multi process coupling production.Based on the efficient collaboration between the equipment edge node and the cloud platform,the cloud training of the prediction model and optimization model,the edge data collection,the model download and call calculation were completed.On this basis,the production process quality prediction model based on GRU-Attention multi-layer neural network was established.The output quality index was taken as fitness,and the whale algorithm was invoked to optimize the production process parameters globally,so as to obtain the optimal process parameter combination of different processes and realize real-time prediction and comprehensive optimization of processing quality of different processes in process manufacturing.An example of a process silk production line was used for experimental verification.The results showed that the proposed cloud edge linkage method based on deep learning could achieve comprehensive dynamic optimization of production quality and reduce the completion time of process parameter control tasks.

Key words: process manufacturing, GRU-Attention multi-layer neural network, cloud edge collaboration, linkage optimization

摘要： 为解决流程制造工艺参数优化面临的多工序耦合模型构建复杂、多目标冲突分析困难、实时和准确性难以保障等问题,提出一种融合GRU-Attention与鲸鱼算法的流程制造工艺参数云边联动优化方法。设计了适用于多工序耦合生产的训练计算云边协同架构,通过设备边缘节点与云平台的高效协同,完成了预测模型和优化模型的云端训练,边缘端数据收集、模型下载和调用计算。在此基础上,建立了基于GRU-Attention多层神经网络的生产工艺质量预测模型,将输出质量指标作为适应度,调用鲸鱼算法对生产工艺参数进行全局寻优,获得不同工序最优工艺参数组合,实现流程生产不同工序加工质量的实时预测和综合优化。最后,以某流程制丝生产线为例进行了实验验证,结果表明,所提基于深度学习的云边联动方法可实现生产质量的综合动态优化,同时可降低工艺参数调控任务的完成时间。

关键词: 流程制造, GRU-Attention多层神经网络, 云边协同, 联动优化