摘要
由于BP神经网络全局寻优存在较大误差,本文基于遗传算法(GA)对BP神经网络进行优化,得出YG8硬质合金耐磨性预测模型。取试验条件中的深冷温度、降温速率、深冷时间、回火温度、回火次数等5项关键工艺参数作为GA-BP模型输入,取磨损量作为模型输出。结果表明,GA-BP预测模型更具有灵活性,预测YG8硬质合金耐磨性正确率达到99.54%,且预测精度较传统的BP神经网络提升了4.07%。
Due to the large error in global optimization of BP neural network,BP neural network was optimized based on genetic algorithm(GA),and the prediction model of wear resistance of YG8 cemented carbide was obtained.Five key process parameters such as cryogenic temperature,cooling rate,cryogenic time,tempering temperature and tempering times in the experimental conditions were taken as input of the GA-BP model,and the wear loss was taken as output of the model.The results show that the GA-BP prediction model is more flexible,the accuracy of predicting the wear resistance of YG8 cemented carbide is 99.54%,and the prediction accuracy is 4.07%higher than the traditional BP neural network.
作者
李帆
闫献国
陈峙
郭宏
姚永超
董良
陈玉华
Li Fan;Yan Xianguo;Chen Zhi;Guo Hong;Yao Yongchao;Dong Liang;Chen Yuhua(Department of Mechanical Engineering,Taiyuan University of Science and Technology,Taiyuan Shanxi 030024,China;Taiyuan Research Institute Co.,Ltd.,China Coal Technology Engineering Group,Taiyuan Shanxi 030006,China;Rugao Nonstandard bearing Co.,Ltd.,Rugao Jiangsu 226563,China)
出处
《金属热处理》
CAS
CSCD
北大核心
2019年第12期244-248,共5页
Heat Treatment of Metals
基金
国家自然科学基金(51675363)
天地科技股份有限公司科技创新创业资金专项项目(2018-TD-MS046)
