基于神经网络热挤压模具钢热处理工艺研究

Research on Heat Treatment Process of Hot Extrusion Die Steel Based on Neural Network

以钢材牌号、钢材长度、表面硬度和热疲劳级别为输入参数,以预热温度、预热时间、淬火温度、淬火时间、回火温度和回火时间为输出参数,构建4×12×3×6四层结构的热挤压模具钢热处理工艺神经网络模型,并进行了试验验证和现场应用确认。结果表明,神经网络模型的的热处理温度预测误差小于4℃,热处理时间预测误差小于2 min,预测精度较高。模型对生产现场的4Cr5MoSiV1热挤压模具钢预测出的热处理工艺,完全能满足企业设计要求,实用性强。

The neural network model of heat treatment process for hot extrusion die steel, which was with 4×12×3×6 four layers, was established with steel series, steel length, surface hardness and thermal fatigue level as input parameters, and with preheat temperature, preheat time, quenching temperature, quenching time, tempering temperature and tempering time as output parameters. Experimental verification and field application were also given out. The results show that the neural network model is with high precision; the prediction error of heat treatment temperature is less than 4 ℃, and its prediction error of heat treatment time is less than 2 min. And the predicted heat treatment process for hot extrusion die steel 4Cr5MoSiV1 based on the neural network model can fully meet the design requirement of company and it has strong practicability.