基于均匀设计法的T91钢管焊接接头工艺参数优化  

Process Parameters Optimization of T91 Steel Pipe Welded Joint Based on Uniform Design Method

作　　者：尹香琴 孔凡玉 缪泽宇 刘辉 

机构地区：[1]中国计量大学质量与安全工程学院浙江省家具检测技术研究重点实验室

出　　处：《热加工工艺》2018年第1期245-248,252,共5页Hot Working Technology

基　　金：国家自然科学基金项目(51104134);浙江省自然科学基金项目(LY14E040001)

摘　　要：以某锅炉厂T91钢管环形焊接接头为研究对象,采用均匀试验设计,以焊接电流、焊接电压、焊接速度、预热温度、焊后热处理温度、保温时间作为可调整的工艺参数,使用Abaqus有限元分析软件对焊后热处理的接头残余应力进行模拟。通过RBF神经网络和粒子群算法对焊接参数进行优化,并采用优化后的焊接参数进行了试验。结果表明,通过均匀试验设计的方法得到RBF神经网络拟合用的训练样本是可行的,大大降低了数据计算量。在优化的焊接工艺参数下预测的焊后最大残余应力与实际模拟值相近。采用优化的焊接参数进行试验得到的接头金相检验合格。Taking T91steel pipe annular welded joint of a boiler plant as the research object, the uniform experimental design was adopted with welding current, welding voltage, welding speed, preheating temperature, post-weld heat treatment temperature and holding time as adjustable process parameters, and the residual stresses of the welded joints after post-weld heat treatment were simulated by using Abaqus finite element analysis software. The welding parameters were optimized by RBF neural network and particle swarm optimization algorithm, and the experiment was carried out by using the optimized welding parameters. The results show that it is feasible to get the training samples of RBF neural network by using the uniform experimental design method, which greatly reduces the amount of data calculation. Under the optimized welding parameters, the predicted maximum residual stress after welding is close to the actual simulation value. The metallographic examination of the welded joint obtained by using the optimized welding parameters is qualified.

关 键 词：T91钢管 均匀试验 数值模拟 神经网络 粒子群算法 

分 类 号：TG404[金属学及工艺—焊接] TG444.74