基于BP神经网络的镁合金轮毂旋转挤压工艺  

Rotary extrusion process for magnesium alloy wheel hub based on BP neural network

作　　者：王剑锋[1] Wang Jianfeng(Department of Railway Locomotive and Car,Baotou railway Vocational and Technical College,Baotou 014060,China)

机构地区：[1]包头铁道职业技术学院铁道机车车辆系,内蒙古包头014060

出　　处：《锻压技术》2020年第6期111-115,共5页Forging & Stamping Technology

基　　金：内蒙古自治区高等学校科学研究项目(NJZC350)。

摘　　要：以镁合金轮毂挤压成形工艺为研究对象,采用有限元分析软件DEFORM-3D对其成形工艺进行模拟。为使制件成形效果达到最佳,选取凸模冲压速度、凹模旋转速度、成形温度作为输入层,以成形后制件的损伤值和应变标准差为输出层,通过构建关于工艺参数的BP神级网络对输入、输出层参数关系进行拟合。并采用遗传算法,基于构建的神经网络寻求最优解。得出最优参数组合为:凸模冲压速度为5.0 mm·s-1、凹模旋转速度为30 r·min-1、成形温度为380℃,采用最优参数进行试模,制件成形效果良好,与预测结果基本一致,验证了有限元模拟和优化的正确性,为实际生产提供指导。For the extrusion process of magnesium alloy wheel hub,the forming process was simulated by finite element analysis software DEFORM-3D.In order to achieve the best forming effect of part,the stamping speed of punch,the rotational speed of die and the forming temperature were selected as the input layer,the damage value and the strain standard deviation of formed parts were chosen as the output layer,and the relationship between input layer and output layer parameters was fit by constructing the BP neural network about process parameters.Furthermore,the optimal solution was found by genetic algorithm based on the constructed neural network.And the optimum parameter combination is the stamping speed of punch of 5.0 mm•s'1,the rotational speed of die of 30 r•min^-1 and the forming temperature of 380℃.Finally,the actual forming process was conducted by optimal parameters.And the forming effect of part is good,which is basically consistent with the predicted results to verify the correctness of the finite element simulation and optimization and provides guidance for actual production.

关 键 词：镁合金 轮毂 BP神经网络 挤压成形 DEFORM-3D 

分 类 号：TG316.11[金属学及工艺—金属压力加工]