基于混合人工神经网络的冷连轧水平力预测

Prediction on horizontal force in cold continuous rolling based onhybrid artificial neural network

针对冷连轧机组辊系中工作辊沿带钢运动方向的水平力难以利用传统的数学模型进行计算的问题,提出了利用混合人工神经网络模型对其进行预测。分析了工作辊在轧制过程中的受力情况,并根据监测的状态参数,从中挑选了轧制力、弯辊力矩、张力、带钢厚度、弯辊力等几类对工作辊受到的沿带钢运动方向的水平力有影响的参数作为输入变量。提出了两种新型的粒子群优化算法,并对人工神经网络的初始化权值与阈值进行优化。通过对预测结果进行分析发现,提出的改进混合人工神经网络相比较改进前能够提高模型的预测精度,且拟合精度均达到90%以上,可用于指导实际生产。

For the problem that it is difficult to use the traditional mathematical model to calculate the horizontal force of working roll along the direction of strip steel movement in the cold continuous rolling mill roll system,a hybrid artificial neural network model was proposed to predict it,and the stress situation of working roll during the rolling process was analyzed.Then,based on the monitored state parameters,several types of parameters that had an impact on the horizontal force acting on the working roll along the direction of strip steel movement,such as rolling force,bending moment,tension,strip steel thickness and bending force,were selected as input variables,and two new particle swarm optimization algorithms were proposed to optimize the initialization weights and thresholds of artificial neural networks.The prediction analysis results show that the proposed improved hybrid artificial neural network can improve the prediction accuracy of the model compared with that before improvement,and the fitting accuracy is more than 90%,which can be used to guide the actual production.