摘要
针对现有聚氨酯发泡机混合头混合均匀度差的问题,在某公司现有混合头结构的基础上,对该混合头进行结构参数优化。首先,设计正交试验,对该混合头进行流场分析,将得到的仿真试验数据进行极差分析,得到该混合头的结构参数对混合指数的影响顺序。其次,采用遗传算法(GA)优化BP神经网路后,得到训练好的GA-BP预测模型,用于预测该混合头的混合指数,拟合出混合头的5个结构参数与混合指数之间的映射关系,通过对比分析BP与GA-BP网络预测模型可知,BP网络模型的回归值R为0.5646,而GA⁃BP网络模型的R为0.9903,GA-BP网络预测系统与直接使用BP网络预测系统相比,有较高的预测准确性及稳定性。最后,采用GA对GA-BP网络预测模型进行全局极值寻优,得到较优的结构参数。GA-BP-GA优化后的混合指数与初始方案的混合指数相比,降低了52%,在一定程度上降低了该混合头混合指数,提高了该混合头的混合效果,通过对比Matlab中GA-BP-GA预测的混合效果0.1653 mm与Fluent仿真所得值0.165 mm,发现两者相对误差为0.18%,验证了该优化方法的可行性。
In order to solve the problem of poor mixing uniformity of the existing polyurethane foaming machine mixing head,the structural parameters of the mixing head were optimized based on the structure of the existing mixing head of a certain company.First,an orthogonal test was designed to conduct flow field analysis on the mixing head.The simulation test data would be obtained for range analysis to obtain the order in which the structural parameters of the mixing head affect the mixing index.Secondly,after using the genetic algorithm(GA)to optimize the BP neural network,the trained GA-BP prediction model was obtained,which was used to predict the mixing index of the mixing head and fitted the mapping relationship between the five structural parameters of the mixing head and the mixing index,through comparative analysis of the BP and GA-BP network prediction models,it can be seen that the R of the BP network model is 0.5646,while the R of the GA⁃BP network model is 0.9903.Compared with the direct use of the BP network prediction system,the GA-BP network prediction system has Higher prediction accuracy and stability.Finally,GA was used to perform global extreme value optimization on the GA-BP network prediction model to obtain better structural parameters.Compared with the mixing index of the initial scheme,the optimized mixing index of GA-BP-GA is reduced by 52%,which reduces the mixing index of the mixing head to a certain extent and improves the mixing effect of the mixing head.By comparing the GA in Matlab,the mixing effect predicted by GA-BP-GA is 0.1653 mm and the value obtained by Fluent simulation is 0.165 mm.The relative error between the two is 0.18%,which verifies the feasibility of the optimization method.
作者
江明会
梅益
罗彦英
余书发
胡大兵
Jiang Minghui;Mei Yi;Luo Yanying;Yu Shufa;Hu Dabing(College of Mechanical Engineering,Guizhou University,Guiyang 550025,China;Guizhou Huayun Automobile Trim Manufacturing Co.,Ltd.,Guiyang 550025,China;Zunyi Jingxing Aerospace Electric Co.,Ltd.,Zunyi 563125,China)
出处
《工程塑料应用》
CAS
CSCD
北大核心
2023年第12期92-99,共8页
Engineering Plastics Application
基金
贵阳科技计划项目(筑科合同[2022]5-38)。
关键词
混合头
流体分析
正交试验
mixing head
fluid analysis
orthogonal test
