基于GRU神经网络和遗传算法的飞机装配站位物料配置方案优化  

Material Configuration Optimization Method Based on GRU Neural Network and Genetic Algorithm in Aircraft Assembly Stations

作　　者：张琦 蒋昌健 韩嘉威 刘金炜 ZHANG Qi;JIANG Changjian;HAN Jiawei;LIU Jinwei(AVIC XAC Commercial Aircraft Co.,Ltd.,Xi’an 710089,China;Northwestern Polytechnical University,Xi’an 710072,China)

机构地区：[1]中航西飞民用飞机有限责任公司,西安710089 [2]西北工业大学,西安710072

出　　处：《航空制造技术》2024年第17期78-82,92,共6页Aeronautical Manufacturing Technology

基　　金：国家重点研发计划(2019YFB1707501)。

摘　　要：装配站位是飞机装配线基础管理单元,由于飞机装配作业过程繁杂并存在大量随机扰动,其管理者需要根据不断变化的工况频繁对在装架次的物料配置方案进行优化。为此,提出一种基于门控循环单元(Gated recurrent unit,GRU)神经网络和遗传算法的优化方法。为了克服离散事件仿真在评估效率方面的局限性,以仿真历史数据为学习样本,采用GRU神经网络构建物料配置方案评估仿真代理模型,模型以物料配置方案为输入,以在装架次预计完工时间和关键物料平均滞留时间为输出。将仿真代理模型作为目标函数评估模型与遗传算法相结合,实现物料配置方案全局优化。仿真验证结果表明,基于GRU神经网络的仿真代理模型能够准确、高效地评估物料配置方案,输出的优化方案能够有效缩短在装架次的预计完工时间和关键物料平均滞留时间。The assembly station is the basic management unit of the aircraft assembly line.Due to the complicated process of aircraft assembly and a large number of random disturbances,its managers need to frequently optimize the material configuration of the aircraft being processed.To this end,an optimization method based on the gated recurrent unit(GRU)neural network and genetic algorithm was proposed.In order to overcome the limitation of discrete event simulation in terms of efficiency,a simulation agent model of material configuration evaluation based on GRU neural network was constructed by taking the simulation historical data as the learning sample.The model took the material configuration as the input,and took the estimated completion time and the average residence time of key materials as the output.The simulation agent model was combined with the genetic algorithm as the objective function evaluation model to realize the global optimization of the material configuration.The simulation verification results show that the simulation agent model based on GRU neural network can accurately and efficiently evaluate the material configuration,and the output optimization configuration can effectively shorten the estimated completion time and average residence time.

关 键 词：装配站位 物料配置方案 离散事件仿真 GRU神经网络 遗传算法 

分 类 号：V262.4[航空宇航科学与技术—航空宇航制造工程]