基于神经网络与NSGA- Ⅱ算法的钢轨闪光焊工艺参数优化  

Optimization of Rail Flash Butt Welding Process Parameters Based on Neural Network and NSGA-ⅡAlgorithm

作　　者：刘新 王晓 杨博 郝美琪 吕其兵[1,2] LIU Xin;WANG Xiao;YANG Bo;HAO Meiqi;LYU Qibing(School of Materials Science and Engineering,Southwest Jiaotong University,Chengdu 610031,China;Chengdu Aigre Technology Co.,Ltd.,Chengdu 610097,China)

机构地区：[1]西南交通大学材料科学与工程学院,四川成都610031 [2]成都艾格科技有限责任公司,四川成都610097

出　　处：《铁道学报》2024年第12期104-112,共9页Journal of the China Railway Society

基　　金：四川省科技成果转移转化示范项目(2023ZHCG0038)。

摘　　要：高压阶段是钢轨闪光焊接过程中快速加热并建立起较宽温度场的关键阶段,其工艺参数与焊接加热效率之间是复杂的非线性关系,传统工艺调试方法难以获得最优的高效加热工艺。通过正交试验和极差分析研究高压阶段工艺参数对加热效率的影响规律;分别利用BP神经网络、遗传算法优化的BP神经网络(GA-BP)、粒子群算法优化的BP神经网络(PSO-BP)及RBF神经网络建立热输入与热效率的预测模型,预测结果表明RBF模型的预测精度最高;结合NSGA-Ⅱ算法的多目标寻优能力,以RBF模型作为工艺优化过程中的函数关系,对四个工艺参数进行优化,验证试验结果表明,采用优化后的工艺参数进行焊接时,有效热输入增加22.778%,900℃以上的高温区温度场宽度增加27.48%,有利于钢轨闪光焊的快速高效加热和获得更好的温度场分布,证明本文工艺参数优化方法的有效性。The high voltage stage is a key stage for rapid heating and establishing a wide temperature field in the rail flash butt welding process.Given complex nonlinear relationship between the process parameters and the welding heating efficiency,it is difficult to obtain the optimal and efficient heating process parameters by traditional process optimization methods.In this paper,the influence of process parameters on heating efficiency in high voltage stage was studied by orthogonal test and range analysis.BP neural network,BP neural network optimized by genetic algorithm(GA-BP),BP neural network optimized by particle swarm optimization(PSO-BP)and RBF neural network were respectively used to establish the prediction models of thermal input and thermal efficiency,with the prediction results showing the highest prediction accuracy of the RBF model.Based on the multi-objective optimization ability of NSGA-IIalgorithm,four process parameters were optimized with the RBF model as the functional relationship in the process optimization.The verification test results show that the valid heat input increases by 22.778%,and the temperature field width in the high temperature zone above 900℃increases by 27.48%using the optimized process parameters for welding,which is conducive to rapid and efficient heating of rail flash-butt welding and obtaining better temperature field distribution,proving the effectiveness of the process parameter optimization method in this paper.

关 键 词：钢轨闪光焊 正交试验 神经网络 NSGA-Ⅱ算法 工艺参数优化 

分 类 号：TG441[金属学及工艺—焊接]