基于BP神经网络的Q235钢力学性能预测模型

Prediction Model on Mechanical Properties of Q235 Steel Based on BP Neural Network

随着计算机处理能力的不断提高,BP神经网络在材料领域发挥巨大作用。其最大的优势是可以回避探寻物理机制的繁琐过程,直接建立预测材料性能的模型。基于某钢铁公司实际生产的Q235钢工艺参数和力学性能数据,建立基于BP神经网络的Q235钢力学性能预测模型,提出新的隐含层选择方法,并进行泛化能力测试。结果表明,BP网络隐含层节点数与预测精度存在密切关系,隐藏层节点数为12时,抗拉强度、屈服强度、延伸率预测的精度达到最高,MAPE最小仅为8%;在测试模型泛化能力时发现,模型遇到其他统一来源的数据时仍然保持高的预测精度,证明所构建的BP神经网络预测模型具有一定的先进性,且新的隐含层节点数选取方法确实具有一定的实用性。同时运用MATLAB程序、Excel表格对Q235钢的变型抗力进行预测,也具有一定的先进性。

With the continuous improvement of computer processing power,BP neural network plays a huge role in the field of materials.Its biggest advantage is that it can avoid the cumbersome process of exploring physical mechanisms and directly establish a model for predicting material properties.Based on the Q235 steel process parameters and mechanical property data actually produced by Jinxi Iron and Steel,a Q235 steel mechanical property prediction model was established based on BP neural network.A new hidden layer selection method was proposed,and generalization ability testing was conducted.The results show that the number of hidden layer nodes of the BP network is closely related to the prediction accuracy.When the number of hidden layer nodes is 12,the accuracy of predicting tensile strength,yield strength,and elongation is the highest,and the minimum MAPE is only 8%.When the model is generalized,it is found that the model still maintains high prediction accuracy when it encounters data from other unified sources,which proves that the built BP neural network prediction model has a certain degree of advancement,and shows that the new hidden layer node number selection method indeed Has a certain practicality.At the same time,the use of MATLAB program and Excel table to predict the deformation resistance of Q235 steel is also advanced.