基于改进SSA-GA-BP神经网络的热连轧轧制力预测

Rolling force prediction of hot continuous rolling based on improved SSA-GA-BP neural network

针对热连轧过程传统轧制力模型不能准确反映实际轧制力的问题,提出了群智能优化算法优化机器学习的热连轧轧制力预测模型。以某热连轧生产线数据为基础,利用随机森林算法进行输入特征的重要性排序,选取重要度较大的参数作为BP神经网络的输入参数,并采用Pauta法则对原始轧制数据进行预处理,基于GA-BP模型建立了精度良好的轧制力预测模型,并采用改进的麻雀搜索算法对GA-BP模型进行了二次优化。将建立的改进SSA-GA-BP模型与传统轧制力模型、BP神经网络模型以及未二次优化的GA-BP模型进行对比。结果表明,改进的SSA-GA-BP模型较其他模型具有较高的预测精度和良好的泛化能力。

Aiming at the problem that the conventional rolling force model for hot continuous rolling does not accurately reflect the actual rolling force,a swarm intelligence optimization algorithm was proposed to optimize the machine learning hot continuous rolling force prediction model.Based on the data of a hot continuous rolling production line,the random forest algorithm was used to rank the importance of the input features,and the parameters with higher importance were selected as the input parameters of BP neural network,and the original rolling data was pre-processed using Pauta rule to establish a rolling force prediction model with good accuracy based on the GA-BP model.The GA-BP model was secondary optimized by improved sparrow search algorithm.The improved SSA-GA-BP model was compared with the conventional rolling force model,the BP neural network model and the GA-BP model without secondary optimization.The results show that the improved SSA-GA-BP model has higher prediction accuracy and good generalization ability than the other models.