基于GRNN神经网络的SAE8620H汽车齿轮钢淬透性预测  

The Hardenability Prediction for SAE8620H Automobile Gear Steel Based on GRNN Neural Network

作　　者：徐兴田 孙小雨 李长宏 XU Xing-tian;SUN Xiao-yu;LI Chang-hong(Nanjing Iron&Steel Co.,Ltd.,Nanjing,Jiangsu 210035,China;College of Metallurgical Engineering,Anhui University of Technology,Maanshan,Anhui 243032,China)

机构地区：[1]南京钢铁股份有限公司,江苏南京210035 [2]安徽工业大学冶金工程学院,安徽马鞍山243032

出　　处：《云南冶金》2020年第5期108-111,116,共5页Yunnan Metallurgy

基　　金：国家自然科学基金-钢铁联合研究基金重点项目(U1860201)。

摘　　要：通过实验得到相关钢种的化学成份和淬透性的数据,再以钢的化学成分作为神经网络的输入、淬透性值作为神经网络的输出,建立关于钢淬透性的GRNN神经网络预测模型。经检验建立出来的神经网络线性回归的相关系数为0.9928,相对平均误差为3.5%,显示出较高精度,对于钢淬透性预测有一定的实际意义。The chemical component and the hardenability data of the relative steel grades is got through the experiments,and then the chemical component is acted as the input of the neural network,the hardenability value is acted as the output of the neural network,therefore the GRNN neural network prediction model of the steel hardenability is established.The correlation coefficient of the linear regression of the established neural network is 0.9928,the relative average error is 3.5%,it shows high precision,which has a certain practical significance for the predication of steel hardenability.

关 键 词：GRNN神经网络 淬透性 SAE8620H钢 

分 类 号：TG162.73[金属学及工艺—热处理]