基于PSO-BP的双辊热轧AZ91D镁合金板横向厚度预测  

Analysis of Thickness Transverse Distribution for AZ91D Hot Rolling Strip Based on PSO-BP

作　　者：杨淑贞[1] 董彬[2] Yang Shuzhen;Dong Bin(Equipment Engineering Department,Henan Technical College of Construction;School of Jiaotong,Huanghe Science and Technology College)

机构地区：[1]河南建筑职业技术学院设备工程系 [2]黄河科技学院交通学院

出　　处：《特种铸造及有色合金》2018年第11期1212-1214,共3页Special Casting & Nonferrous Alloys

摘　　要：将粒子群优化算法(PSO)与BP神经网络相结合,引入到双辊热轧AZ91D镁合金板带横向厚度分布的预测与控制中。从不同压下量和板宽两个角度与AZ91D镁合金板带横向厚度分布预测值和实测值进行比较,确定了该模型针对板宽的适用范围在100～500mm,热轧时压下量控制在30μm以内是较优的。该预测模型平均绝对误差为4.2μm,验证了将PSO-BP神经网络应用到双辊热轧板带横向厚度分布控制系统中是可行的。The particle swarm optimization（PSO）combined with BP neural network was used for the forecast and control of thickness transverse distribution of hot rolled strip.According to theoretical analysis,the basic conditions and parameters of forecasting model were established.And based on the reduction and plate width,the measured data and simulated one of thickness transverse distribution of AZ91D hot rolling strip were analyzed comparatively.The model presents the optimized effects with strip width in 100～500 mm and reduction of 30μm.The average error of thickness is less than 4.2μm.Finally,the neural network was used for thickness control system of the hot rolling strip.

关 键 词：神经网络 板厚 双辊热轧 

分 类 号：TG333.7[金属学及工艺—金属压力加工]