基于PIO-RBF神经网络斜轧穿孔机调整参数预测  

Prediction of adjustment parameters of cross-rolling piercing millbased on PIO-RBF neural network

作　　者：孙继芸 王清华[1] 王贞艳[1] 胡建华[2] 徐洪岩[3] SUN Ji-yun;WANG Qing-hua;WANG Zhen-yan;HU Jian-hua;XU Hong-yan(School of Electronic Information,Taiyuan University of Science and Technology,Taiyuan 030024,China;School of Materials Science and Engineering,Taiyuan University of Science and Technology,Taiyuan 030024,China;Technology Center,Taiyuan Heavy Industry Co.,Ltd.,Taiyuan 030024,China)

机构地区：[1]太原科技大学电子信息工程学院,山西太原030024 [2]太原科技大学材料科学与工程学院,山西太原030024 [3]太原重工股份有限公司技术中心,山西太原030024

出　　处：《塑性工程学报》2023年第3期197-203,共7页Journal of Plasticity Engineering

基　　金：山西省科技重大专项(20191102009)。

摘　　要：针对无缝钢管二辊斜轧穿孔生产工艺中轧机调整参数对钢管质量影响较大,但其设定值精度不高的问题,提出了基于鸽群算法改进RBF神经网络斜轧穿孔机调整参数预测模型。首先,综合分析了传统的二辊斜轧穿孔调整参数数学模型并确定了主要特征参数,其次,建立了两辊斜轧穿孔时轧机参数(轧辊间距、导板间距和顶头前伸量)的RBF神经网络预测模型,并提出鸽群算法优化RBF神经网络的中心、方差(宽度)和隐层与输出层之间的连接权值。针对某厂采集的304L管的生产数据,对提出的预测模型进行了训练和验证。通过与基于聚类分析的RBF神经网络模型对比,将经PIO-RBF神经网络模型预测得到的轧机调整参数(轧辊间距、导板间距和顶头前伸量)数据与实际数据比较,其相对误差均可控制在9%以内。结果表明,由PIO-RBF神经网络建立的预测模型对轧辊间距、导板间距及顶头前伸量具有较高的预测精度且适用性强。Aiming at the problem that the adjustment parameters of rolling mill have great influence on the quality of steel tube in the production process of two-roll cross-rolling piercing of seamless steel tube,but the precision of its set value is not high.The prediction model of adjustment parameters of cross-rolling piercing mill based on improved RBF neural network by pigeon colony alogrithm was proposed.Firstly,the traditional mathematical model of two-roll cross-rolling piercing adjustment parameters was comprehensively analyzed,and the main characteristic parameters were established.Secondly,the RBF neural network prediction model of rolling mill parameters(roll spacing,guide plate spacing and plug forward stretch)was established,and pigeon colony algorithm was proposed to optimize the center,variance(width)and the connection weight between hidden layer and output layer of RBF neural network.According to the production data of 304L tube collected by a factory,the proposed prediction model was trained and verified.Compared with the RBF neural network model based on cluster analysis,the relative error between the data predicted by the PIO-RBF neural network model and the actual data of rolling mill parameters(roll spacing,guide plate spacing and plug forward stretch)can be controlled within 9%.The results show that the prediction model established by PIO-RBF neural network has high prediction accuracy and strong applicability for roll spacing,guide plate spacing and plug forward extension.

关 键 词：斜轧穿孔 调整参数预测 鸽群算法 RBF神经网络 

分 类 号：TG355[金属学及工艺—金属压力加工]