神经网络在冷轧301S不锈钢极薄带材表面硬度预测中的应用

Applications of Neutral Network in Surface Hardness Prediction for Cold Rolling 301S Stainless Steel Thin Strip

基于现场生产冷轧极薄不锈钢带材表面硬度极难控制的问题，针对301S不锈钢的冷轧生产工艺进行了研究，分析了不锈钢冷轧生产过程中影响表面硬度的相关工艺参数，得出材料的抗拉强度、轧制速度、轧制油温度和压下率是影响轧后材料表面硬度的关键因素。利用BP神经网络建立了预测表面硬度的非线性映射模型，并根据此模型得出了预测数据的趋势图谱。研究结果表明，压下率的变化对冷轧不锈钢表面硬度的调节能力最强，而其他参数对硬度的影响为10HV左右。经检验，模型的预测值和实测值的相对误差为2．63％～2．76％，预测结果准确率高，可以用于产品质量的现场在线控制。

In order to solve the problem of 301S stainless steel strip surface hardness control, the production process of cold rolling 301S stainless steel strip was investigated, and the key process parameters affecting the surface hardness of cold rolling stainless steel strip were obtained, which were tensile strength, rolling speed, temperature of rolling oil and reduction rate. A non linear reflection model was built for predicting the surface hardness of the stainless steel strips by using BP neural network, and the collection of illustrative plates of predicting data were achieved by this model. The results show that the variation of reduction rate affects surface hardness values significantly, and the influence of other parameters on the hardness value is about 10 HV. The relative error between predicted and tested results of the surface hardness is between -2.63 % and 2.76 %. The neural network model can be used in the field of online control of product quality.