基于BP神经网络预测热处理后35CrMoA钢的低温冲击性能  

Prediction of low-temperature impact property of 35CrMoA steel after heat treatment based on BP neural network

作　　者：金帅 王会强[1] 崔建英[2] 王艳山 闫学兰 李文翔 Jin Shuai;Wang Huiqiang;Cui Jianying;Wang Yanshan;Yan Xuelan;Li Wenxiang(College of Mechanical and Electrical Engineering,Hebei Agricultural University,Baoding Hebei 071001,China;Juli Sling Co.,Ltd.,Baoding Hebei 072550,China;Baoding Dongli Machinery Co.,Ltd.,Baoding Hebei 071100,China;Baoding Golden Sunshine Energy Equipment Technology Co.,Ltd.,Baoding Hebei 071000,China;Guantao Flying Machinery Equipment Manufacturing Co.,Ltd.,Handan Hebei 057750,China)

机构地区：[1]河北农业大学机电工程学院,河北保定071001 [2]巨力索具股份有限公司,河北保定072550 [3]保定市东利机械制造股份有限公司,河北保定071100 [4]保定金阳光能源装备科技有限公司,河北保定071000 [5]馆陶县飞翔机械装备制造有限公司,河北邯郸057750

出　　处：《金属热处理》2025年第3期125-131,共7页Heat Treatment of Metals

基　　金：河北省科技计划(20312201D);河北省重大成果转化项目(21287001Z)。

摘　　要：为了预测正火态35CrMoA钢试棒进行亚温淬火和高温回火后在-45℃的低温冲击性能,采用不同热处理温度下-45℃低温冲击性能实际状态参量作为学习样本,对3层BP人工神经网络(BPANN)进行训练和预测低温冲击性能。结果表明:BPANN能够对正火态35CrMoA钢试样在不同亚温淬火温度、回火温度下的-45℃低温冲击性能进行预测,误差范围区间为5%~9%;BPANN的预测值比实际值低,可通过增加训练样本提升预测精确度,预测数据变化趋势与实测变化趋势相同,能够实现具有参考意义的预测。本研究可通过预测减少实际生产中试验次数,节约试验成本,有助于35CrMoA钢在其他亚温淬火温度下的低温冲击性能预测的研究。In order to predict the low-temperature impact performance at-45℃of normalized 35CrMoA steel test bars after subcritical quenching and high-temperature tempering,the actual state parameters of low-temperature impact property at-45℃under different heat treatment temperatures were used as learning specimens to train and predict the low-temperature impact property using a three-layer back propagation artificial neural network(BPANN).The results show that the BPANN can predict the low-temperature impact property at-45℃of the 35CrMoA steel specimens subcritical quenched at different temperatures and tempered at different temperatures,with an error range of 5%to 9%.The predicted values of the BPANN are lower than the actual values,and the prediction accuracy can be improved by increasing the training specimens.The trend of the predicted data is the same as that of the measured data,and it can achieve a meaningful prediction.This study can reduce the number of tests in actual production through prediction,save test costs,and is helpful for the research on the low-temperature impact property prediction of the 35CrMoA steel at other subcritical quenching temperatures.

关 键 词：35CRMOA钢 BP神经网络 热稳定性 亚温淬火 预测 

分 类 号：TG156[金属学及工艺—热处理]