基于BP神经网络的电力角钢成分优化及生产实践

Composition optimization and production practice of large angle steel for iron tower based on BP neural network

基于BP神经网络,创建了角钢合金成分与力学性能关系预测模型,所建立的模型绝对误差平均值仅为3~4 MPa,具有较高的可靠性。利用该性能预测模型研究了残余元素Cr含量及VN12合金加入量对强度的量化贡献,并进行工业试制250 mm×250 mm×35 mm电力角钢,结果表明:当Cr质量分数大于350×10^(-6)时,可以采用每吨钢添加0.5 kg的VN12合金、0.3 kg的VFe合金,使钢中V质量分数为0.05%,其屈服强度和抗拉强度均可满足性能要求。基于BP神经网络创建的电力角钢预测模型对成分进行优化设计具有可靠的指导性。

Based on the BP neural network,a prediction model of the relationship between composition and mechanical properties of angle steel alloy was established.The average absolute error of the model was only 3-4 MPa,which had high reliability.The quantitative contribution of residual element Cr content and VN12 alloy addition to the strength was studied by using the performance prediction model,and the industrial trial production of 250 mm×250 mm×35 mm angle steel was carried out.The results show that when the mass fraction of Cr is greater than 350×10^(-6),0.5 kg of VN12 alloy and 0.3 kg of VFe alloy per ton of steel can be added to ensure the V mass fraction of 0.05%in steel,and meet the performance requirement of the strength.The prediction model of large angle steel for iron tower based on BP neural network has reliable guidance for composition optimization design.