基于MSC. Marc有限元软件平台,针对SUS316马鞍形管-管焊接头的焊接残余应力预测,开发了2种能同时兼顾精度与效率的计算方法。第1种方法建立了与实际接头尺寸一致的全模型,采用移动热源与瞬间热源混合使用的方法,即采用移动热源模拟打底与盖面焊道的热输入,采用瞬间热源模拟填充焊道热输入；第2种方法利用接头几何形状的对称性,建立了1/4局部模型,并采用瞬间热源模拟全部焊道的热输入。由于SUS316加工硬化效应显著,在材料模型中采用各向同性硬化准则来考虑加工硬化,同时采用阶跃式退火模型来模拟材料的退火软化。比较计算结果和实验结果可知,不论是典型位置的焊接热循环还是接头的残余应力分布,数值模拟结果与实验结果均吻合较好。采用全模型既可以得到整个接头的残余应力分布也可以获得始终端位置应力分布特征。局部模型也能准确预测稳定区域应力的大小和分布,并可以大幅节省计算时间和存储空间。

A thick-walled SUS316 saddle tube-pipe welded joint is used in nuclear power equipment. A very long computing time and huge memory space are needed to simulate welding residual stress when the thermo-elastic-plastic finite element method is used because of the complex shapes, large sizes, and many weld passes of this joint. To solve the computational problem, two efficient and accurate computational approaches were proposed based on MSC. Marc finite element software platform. In the first computational approach, the finite element model of the SUS316 saddle tube-pipe welded joint was established with the same dimensions as the actual joint. Two heat sources were used to balance the computing time and calculation precision. The moving heat-source model was used to simulate the heat input for the backing and cover passes. In contrast, the instantaneous heat-source model was employed to consider the heat input for the other passes. Considering the geometric symmetry, a quarter model was developed in the second computational approach, and the instantaneous heat-source model was used to model the heat input for all passes. In the material model, both work hardening isotropic rule and annealing effect were considered because SUS316 is sensitive to work hardening. The simulation results of the thermal cycle during the welding process and residual stress distribution in and near the fusion zone were compared using the measured data. The results of thermal cycles and the residual stress distributions obtained using two computational approaches matched the experimental measurements. When the first computational approach was used, not only the residual stress distribution in the whole welded joints could be obtained, but also the features of residual stress distribution near the weld start-end location were able to capture. The second computational approach could predict the magnitude and distribution of residual stress in the stable range of the joint and could save computing time and huge memory space. Thus, the second computational approach is useful for practical engineering applications.