摘要
12Cr13马氏体不锈钢淬火和回火工艺参数对其力学性能影响较大,测定其热处理工艺对力学性能的影响周期长且成本高。在神经网络与遗传算法基础上建立12Cr13马氏体不锈钢热处理工艺参数与力学性能的预测模型。模型输入单元为淬火温度、淬火保温时间、冷却方式及回火温度,输出单元为抗拉强度、屈服强度、延伸率及断面收缩率。采用Traincgf算法的神经网络收敛速度快,误差小。隐含层节点单元为6,动量因子为0.6,学习速率为0.2时,网络测试的均方误差值均最小。经过网络测试的抗拉强度、屈服强度、延伸率及断面收缩率的最大相对误差绝对值分别为3.24%,2.48%,9.45%和8.82%。12Cr13马氏体不锈钢的预测模型具有结构简单,拟合精度高的特点。可利用12Cr13马氏体不锈钢热处理工艺参数预测其力学性能,为工艺优化设计提供参考。
The quenching and tempering technological parameters of 12Cr13 martensitic stainless steel have great influence on its mechanical properties, the period is long and cost is high to determine the influence of its heat treatment process on mechanical properties. The predictive model of 12Cr13 martensitic stainless steel heat treatment technological parameter and mechanical properties was established on the basis of artificial neural network（ANN） and genetic algorithm（GA）. The model input unit has quenching temperature, holding time, cooling way and tempering temperature, output unit has tensile strength, yield strength, elongation rate and reduction of area. The neural network adopting Traincgf algorithm convergence speed is fast and error is small. When concealed layer node unit is 6, momentum factor is 0.6, learning rate is 0.2, the mean square error value of network measurement is all minimum. The maximum relative error absolute value of tensile strength, yield strength, elongation rate and reduction of area tested with network is 3.24% ,2.48% ,9.45% and 8.82% respectively. The predictive model of 12Cr13 martensitic stainless steel has the characteristic of simple structure and high fitting precision. The mechanical properties of 12Cr13 martensitie stainless steel can be predicted with its heat treatment technological parameters, it supplies references for designing process optimization.
出处
《金属制品》
2013年第3期39-43,共5页
Metal Products
