基于端边云协同和MIRF_WPSO的流程工艺参数自适应实时优化模型

针对流程工业生产过程中因工序间相互耦合、工艺数据量庞大且处理时延高而导致的工艺参数优化实时性难以保证的问题,提出一种基于端边云协同和MIRF_WPSO的流程工艺参数自适应实时优化模型.首先,基于边缘计算技术搭建多源异构流程工艺参数端边云协同实时优化架构;其次,构建基于互信息随机森林MIRF和自适应惯性权重粒子群WPSO的工艺参数优化算法MIRF_WPSO,并将MIRF_WPSO算法部署在边缘端以实现工艺参数的实时优化,同时通过在云端部署自更新机制来实现边缘端算法模型的自感知更新,从而形成集算法训练-更新-调用的端边云高效协同自动化闭环网络;最后,搭建实验平台,实验结果表明,“端-边-云”协同模式可以有效缓解云端计算压力,能够实时、高效地对流程工艺参数进行自优化调控,将质量指标平均标偏从1.86%降到1.25%,优化速度提高11.4%,为流程工业生产过程智能化进一步发展提供新的思路.

Adaptive real-time optimization model of process parameters based on end-edge-cloud collaboration and MIRF_WPSO

In view of the problem that it is difficult to guarantee the real-time optimization of process parameters due to the mutual coupling between processes, the large amount of process data and high processing delay in the process industrial production process, an adaptive real-time optimization model of process parameters based on end-edge-cloud collaboration and MIRF_WPSO is proposed. Firstly, an end-edge-cloud collaborative real-time optimization architecture for process parameters of multi-source heterogeneous processes is built based on edge computing technology. Then, a process parameter optimization algorithm based on mutual information random forest and adaptive inertia weighted particle swarm(MIRF_WPSO) is constructed, which is deployed at the edge to realize real-time optimization of process parameters, while a self-aware update mechanism is deployed at the cloud to realize an efficient automated closed-loop network of algorithm training-updating-recall. Finally, an experimental platform is built, and the experimental results show that the “end-edge-cloud” collaborative mode effectively relieves the computational pressure on the cloud, and enables real-time and efficient self-optimized regulation of process parameters. The average standard deviation of quality index is reduced from 1.86% to 1.25%, and the optimization speed is increased by 11.4%, providing new ideas for the further development of intelligent production processes in process industries.