基于BP神经网络的700 MPa级高强钢抗拉强度不合原因分析

Cause analysis of disqualification of tensile strength of 700 MPa grade high strengthsteel based on BP neural network

为了确定涟钢2250 mm产线700 MPa级高强钢抗拉强度不合原因,对全流程关键工艺进行了相关性分析,并利用大量历史生产数据建立了化学成分、热轧工艺参数与成品抗拉强度关系的神经网络预测模型。模型的均方根误差(RMSE)、平均绝对相对误差(AARE)分别为32 MPa、1.7%,验证数据的相关系数R为0.95。神经网络预测模型应用结果表明:铸坯入炉温度每提高100℃,成品抗拉强度降低4.5 MPa;铸坯出炉温度每提高10℃,成品抗拉强度降低12.3 MPa;粗轧温度对成品抗拉强度的影响呈抛物线趋势,其每提高10℃,成品抗拉强度仅降低0.5 MPa;粗轧第3道次速度每增加1.0 m/s,成品抗拉强度降低35.4 MPa,与相关性分析的54.3 MPa接近;卷取温度每提高10℃,成品抗拉强度降低5.2 MPa;层冷水温每上升1.0℃,成品抗拉强度降低1.6 MPa。基于相关性分析和神经网络模型,粗轧速度的提升是700 MPa级高强钢抗拉强度不合的主要原因,大幅度提高热装比、增加入炉温度是次要原因。对于粗轧再结晶区轧制而言,粗轧速度越快,再结晶形变因子越小,奥氏体的动态再结晶、静态再结晶越不充分,从而容易得到混晶组织,进而显著降低了热轧成品的抗拉强度。

To determine the reason of disqualification of tensile strength of 700 MPa grade high strength steel in 2250 mm production line of Lianyuan Iron and Steel Co.,Ltd.,a correlation analysis was conducted on the key processes throughout the entire process,and a neural network prediction model for composition,hot rolling process parameters and tensile strength of finish product was established using a large amount of historical production data.The root mean square error(RMSE)and average absolute relative error(AARE)of the prediction model is 32 MPa and 1.7%,respectively.And the correlation coefficient R of the validation data is 0.95.The results of the neural network prediction model indicate that for everyincreasing 100℃of slab entering furnace temperature,the tensile strength of finish product decreases by 4.5 MPa.Forevery increaseing 10℃of slab discharging temperature,the tensile strength of finish product decreases by 12.3 MPa.The effect of roughing rolling temperature on tensile strength of finish product shows a parabolic trend,with increasing 10℃only resulting in a decrease of 0.5 MPa in tensile strength of finish product.For every increaseing 1.0 m/s of the third pass speed of roughing rolling,the tensile strength of finish product decreases by 35.4 MPa,which is close to the correlation analysis of 54.3 MPa.For every increaseing 10℃of coiling temperature,the tensile strength of finish product will decrease by 5.2 MPa.For every increaseing 1.0℃of the temperature of the laminar cooling water,the tensile strength of finish product will decrease by 1.6 MPa.Based on correlation analysis and neural network models,the increase of roughing rolling speed is the main reason for the property problems of 700 MPa grade high strength steel,with a significant increase in hot charging ratio and charging temperature being secondary reasons.For recrystallization zone rolling of roughing rolling,the faster the roughing rolling speed,the smaller the recrystallization deformation factor,and the less sufficient the dynamic and static recrystallization of austenite,which makes it easier to obtain mixed crystal structure and significantly reduces the tensile strength of the hot rolled product.