基于神经网络的机械油泵轴锻造工艺优化  

Forging process optimization on mechanical oil pump shafts based on neural network

作　　者：莫洪武[1] 塔金星[2] Mo Hongwu;Ta Jinxing(Guangxi Agricultural Vocational College,Nanning 530007,China;Institute of Machinery,Northeast Forestry University,Harbin 150040,China)

机构地区：[1]广西农业职业技术学院,广西南宁530007 [2]东北林业大学机械学院,黑龙江哈尔滨150040

出　　处：《锻压技术》2018年第9期21-25,共5页Forging & Stamping Technology

基　　金：广西高校科研项目(KY2016YB686)

摘　　要：为了优化机械油泵轴锻造工艺,提升机械油泵轴的综合性能,基于5×25×1三层拓扑结构,以坯料加热温度、始锻温度、终锻温度、模具预热温度和锻造变形量为5个输入参数,以磨损性能为输出参数,以tansig函数为传递函数,构建了机械油泵轴锻造工艺神经网络优化模型,并进行了神经网络模型的训练、预测与验证。结果表明该模型平均相对训练误差为3.2%,相对预测误差低于5%,具有较高预测精度和较强预测能力。与生产线现用工艺相比,采用模型优化工艺锻造的SKH-51高速钢机械油泵轴的磨损体积减小了51.9%,磨损性能得到明显提高。In order to optimize the forging process and improve the comprehensive performance of mechanical oil pump shaft,based on the5 × 25 × 1 three-level topological structure,the neural network optimization model for forging process of mechanical oil pump shaft was designed and constructed by taking the heating temperature of billet,initial forging temperature,final forging temperature,mold preheating temperature and forging deformation amount as five input parameters,wear performance as output parameter,and tansig function as transfer function,and the training,prediction and validation of the neural network model were carried out. The results show that the average relative training error of the model is 3. 2%,the relative prediction error is lower than 5%,and the model has high prediction accuracy and strong prediction ability. Compared with the present production line,the wear volume of the high speed steel SKH-51 mechanical oil pump shaft forged by model optimization is reduced by 51. 9%,and the wear performance is obviously improved.

关 键 词：神经网络 锻造工艺优化 机械油泵轴 训练误差 预测误差 

分 类 号：TU599[建筑科学—建筑技术科学]