基于BP神经网络连接端子温挤压成型优化  

OPTIMIZATION OF CONNECTION TERMINAL TEMPERATURE EXTRUSION MOLDING BASED ON BP NEURAL NETWORK

作　　者：梅益[1] 肖展开 罗宁康 唐方艳 薛茂远 Mei Yi;Xiao Zhankai;Luo Ningkang;Tang Fangyan;Xue Maoyuan(College of Mechanical Engineering,Guizhou University,Guiyang 550025,Guizhou,China)

机构地区：[1]贵州大学机械工程学院,贵州贵阳550025

出　　处：《计算机应用与软件》2023年第5期97-102,159,共7页Computer Applications and Software

基　　金：贵州省科技计划项目(黔科合支撑[2019]2019);贵州省科技支撑计划项目(黔科合支撑[2018]2175)。

摘　　要：以某汽车连接端子作为研究对象,结合金属挤压成型流变力学基本理论,基于三维模拟软件DEFORM-3D进行模拟分析,获得大量神经网络的样本信号。建立多输入-单输出的BP网络拓扑结构,以模具温度、坯料温度和挤压时间作为输入变量,以连接端子在挤压工作带处具有最均衡的表面平整度为目标。通过正交实验分析成型设计最优方案,具有函数逼近功能的神经网络加以运用,优化设计连接端子的挤压模具参数。通过预测值和模拟值的对比验证最优解的可行性,并发现训练后BP神经网络模型的相对误差较小,具有很好的预测能力。通过DEFORM仿真验证最优解。该方法和运算思路对同类型的研究具有较大的借鉴价值。Taking an automobile connection terminal as the research object and combining the basic theory of metal extrusion molding rheology,the simulation analysis was carried out based on 3D simulation software DEFORM-3D,and a large number of neural network sample signals were obtained.The topological structure of BP network with multiple inputs and single outputs was established,taking die temperature,blank temperature and extrusion time as input variables,aimed at the most balanced surface flatness of connecting terminals at the extrusion strip.The optimum design scheme was designed through orthogonal experiment,and the neural network with function approximation function was used to optimize the die parameters of connecting terminals.Through the comparison of the predicted value and the simulated value,the feasibility of the optimal solution was verified,and it was found that the BP neural network model after training had small relative error and good predictive ability.The optimal solution was verified by DEFORM simulation.The analytical methods and operational ideas adopted in this paper have great reference value for the same type of research.

关 键 词：BP神经网络 连接端子 DEFORM-3D 温挤压 有限元 

分 类 号：TP205[自动化与计算机技术—检测技术与自动化装置]