新型低合金钢的锻造工艺神经网络优化  

Neural Network Optimization of Forging Process of New Low Alloy Steel

作　　者：黄伟凡[1] HUANG Weifan(School of Mathematics and Computer Science,Yichun University,Yichun 336000,China)

机构地区：[1]宜春学院数学与计算机科学学院,江西宜春336000

出　　处：《热加工工艺》2021年第9期90-93,共4页Hot Working Technology

基　　金：国家自然科学基金资助项目(61662083)。

摘　　要：以始锻温度、终锻温度、模具预热温度和锻造比作为输入层节点,并以耐磨损性能(磨损体积)和耐腐蚀性能(腐蚀电位)作为输出项节点,构建了4×32×8×2四层拓扑结构的神经网络优化模型,并对该模型进行了预测和验证。结果表明,神经网络优化模型的磨损体积和腐蚀电位相对训练误差分别为2.63%~4.80%和2.59%~3.75%,平均相对训练误差分别为3.67%和3.20%;磨损体积和腐蚀电位的相对预测误差分别为2.50%~3.70%和2.70%~3.51%,平均预测误差值分别为3.06%和3.14%。模型预测能力强、精度高。Taking the initial forging temperature, final forging temperature, die preheating temperature and forging ratio as input layer nodes, and the wear resistance(wear volume) and corrosion resistance(corrosion potential) as output nodes, a neural network optimization model of 4×32×8×2 four-layer topology was constructed. The model was predicted and verified.The results show that the relative training errors of the wear volume and corrosion potential of the neural network optimization model are 2.63%-4.80% and 2.59%-3.75%, respectively, and the average relative training errors are 3.67% and 3.20%,respectively. The relative prediction errors of wear volume and corrosion potential are 2.50%-3.70% and 2.70%-3.51%,respectively, and the average prediction errors are 3.06% and 3.14%, respectively. And the model’s prediction capability is strong and the precision is high.

关 键 词：神经网络 低合金钢 锻造工艺 耐磨损性能 耐腐蚀性能 

分 类 号：TG316[金属学及工艺—金属压力加工]