面向大型激光装置的智能装配调度  

Intelligent assembly scheduling for large laser devices

作　　者：熊召[1] 尹灵钰 裴国庆[1] 王成程[1] 周海[1] Xiong Zhao;Yin Lingyu;Pei Guoqing;Wang Chengcheng;Zhou Hai(Laser Fusion Research Center,CAEP,Mianyang 621900,China)

机构地区：[1]中国工程物理研究院激光聚变研究中心,四川绵阳621900

出　　处：《强激光与粒子束》2023年第9期77-84,共8页High Power Laser and Particle Beams

基　　金：四川省科技计划项目(2022ZYD0114)。

摘　　要：针对大型激光装置精密装校过程中的智能装配调度问题,提出一种基于人工神经网络的调度优先规则获取方法。该方法离线阶段通过遗传算法对典型算例进行优化求解,从优化解中抽取任务比较轨迹及特征数据,采用人工神经网络学习生成任务优先模型;在线阶段基于该模型构建闭环调度决策模式,实现动态不确定生产环境下的快速响应与精准决策。数据实验和实际应用案例验证了该方法的有效性,随着光机模块数量增加,ANN调度算法的优势更加明显,ANN调度算法和GA算法二者优化结果小于6%时,前者的计算效率是后者的400倍以上。Aiming at the assembly scheduling problem of optical and mechanical modules for large laserdevices,a scheduling priority rule acquisition method based on artificial neural networks(ANNs)is proposed.In theoffline phase,this method optimizes the scheduling data through genetic algorithms,extracts task comparisontrajectories and feature data from the optimization solution,and uses ANNs to learn the task priority comparisonmodel.In the online phase,a closed-loop decision scheduling mode is constructed based on this model to achieve rapidresponse and accurate decision-making in dynamic uncertain production environments.Data experiments and practicalapplication cases verify the effectiveness of this method.With the increase of the number of optical-mechanicalmodules,the advantages of ANN scheduling algorithm become more obvious.When the optimization results of ANNscheduling algorithm and GA algorithm are less than 6%,the computational efficiency of the former is more than 400times that of the latter.

关 键 词：人工神经网络 调度规则 智能装配调度 

分 类 号：TG156[金属学及工艺—热处理]