基于BP神经网络斜轧穿孔轧制力的预测

Rolling Force Prediction of Rotary Piercing Based on BP Neural Network

借助Matlab工具箱中BP神经网络,对斜轧穿孔区轧制力进行预测。通过分析影响轧制力预报精度的因素及网络性能,确定了网络结构、有关参数和网络训练算法(优化BP算法),实现了轧制力的精确快速预报,预报相对误差1.67%～6.33%,平均3.735%,满足了工程计算的精度要求。

The rolling force in rotary piercing area was studied by BP neural network in Matlab toolbox. Through analyzing the factors of influencing the prediction precision of rolling force and network property, the network structure, relevant parameters and network training algorithm （optimized BP algorithm） were confirmed and the rolling forces were predicated precisely and rapidly. The relative error of prediction value is between 1.67% and 6.33%. The average value is 3.735%, satisfying the requirement of engineering precision.