摘要:
采用神经网络算法技术,以钒含量、钛含量、淬火温度、淬火冷却方式、回火温度和回火冷却方式作为输入层参数,以耐磨损性能和冲击韧性为输出层参数,可以构建出6×24×12×2四层拓扑结构的钒钛改性高铬铸铁热处理工艺优化模型。模型输出的耐磨损性能平均相对预测误差为2.8%、冲击韧性平均相对预测误差为2.5%。模型不仅具有较佳的预测能力和较高的预测精度,而且在热处理生产线上具有很好的应用效果,使产线上的钒钛改性(0.8%钒+0.5%钛)高铬铸铁的平均晶粒尺寸减小32%、磨损体积减小50%、冲击韧性提高62%。
Abstract:
The neural network model of heat treatment process for high chromium cast iron modified by vanadium and titanium,which is the structure of 6×24×12×2,was established by the method of neural network algorithm,with vanadium content,titanium content,quenching temperature,tempering cooling mode,tempering temperature and tempering cooling mode as input parameters and with wear resistance and impact toughness as output parameters.The average relative prediction errors of wear resistance and impact toughness were 2.8% and 2.5%,respectively.The neural network model has good prediction ability and high precision.Meanwhile,it has a favorable application effect in heat treatment production line,which could make the impact toughness,average grain size and wear volume of high chromium cast iron modified by vanadium and titanium online,increase by 62%,decrease by 32% and 50%,respectively.
下载:
