摘要
以4×24×12×1的四层拓扑结构,以始锻温度、终锻温度、锻压速度、模具预热温度作为输入层函数,以耐磨损性能作为输出层函数,构建了优化汽车同步器齿环锻压工艺的神经网络模型,并对此模型进行了预测验证以及产线应用验证。结果表明,模型输出的室温和高温磨损体积平均相对预测误差分别为3.3%、3.2%,模型预测能力佳、精度高。与目前产线现用工艺相比,采用神经网络模型优化的工艺锻压的汽车同步器齿环室温磨损体积减小5.3%、高温磨损体积减小38.9%,齿环耐磨损性能提高。
In this paper,the neural network model for optimizing the forging process of automotive synchronizer gear ring is constructed based on a four-layer topology of 4×24×12×1,with the starting forging temperature,finishing forging temperature,forging starting forging temperature,die preheating temperature as input layer functions and wear resistance as output layer functions.Forecast validation and application verification of production line were done.The results show that the average relative validation errors of the model for wear volume at room and high temperature are 3.3%and 3.2%,respectively.The model has good prediction ability and high accuracy.Compared with the process parameters on current production line,the wear volume at room temperature with the forging process optimized by neural network model reduces by 5.3%,and the wear volume at high temperature reduces by 38.9%.The wear resistance of the gear ring is improved with the forging process optimized by neural network model.
作者
吴思俊
尚凯
WU Sijun;SHANG Kai(Zhejiang Industry Polytechnic College,Shaoxing 312000,CHN;Echnical Center,Dongfeng Motor Group Co.,Ltd.,Wuhan 430058,CHN)
出处
《制造技术与机床》
北大核心
2019年第11期160-163,共4页
Manufacturing Technology & Machine Tool
基金
国家自然科学基金项目(51605380)
中国博士后科学基金项目(2016M602842)
陕西省博士后科研项目(2016BSHYDZZ08)
陕西省自然科学基础研究计划项目(2017JQ5105)
关键词
汽车同步器齿环
神经网络优化
锻压工艺
耐磨损性能
automobile synchronizer gear ring
neural network optimization
forging process
wear resistance
