论文部分内容阅读
以4200 mm轧机轧制71块钢板的实测数据为基础,利用Matlab神经网络工具箱,分别建立了轧制变形区的应力状态系数与轧前厚度、轧后厚度及轧辊直径对应关系的Elman神经网络预测模型和RBF神经网络预测模型。结果表明,所建立的两种网络模型均建立了金属应力状态系数输入和输出关系,RBF神经网络模型比Elman网络模型数据稳定,性能更优,实现了与实测结果的高度拟合。并得出不同轧辊直径对神经网络模型精度的影响规律,对轧制工艺规程的制定提出了合理建议。
Based on the measured data of 71 plates rolled by 4200 mm rolling mill, an Elman neural network with corresponding relationship between the stress state coefficient and the thickness before rolling, the thickness after rolling and the roll diameter was established respectively by using Matlab neural network toolbox. Prediction Model and RBF Neural Network Prediction Model. The results show that the two established network models both establish the relationship between the input and the output of metal stress state coefficients. The RBF neural network model is more stable and has better performance than the Elman network model data, achieving a high degree of fitting with the measured results. The influence of different roller diameters on the accuracy of neural network model was obtained, and reasonable suggestions for the formulation of rolling process rules were put forward.