Multiscale investigation of mechanical behavior in the heat-affected zone of high-strength steel

During the welding process, thermal cycles lead to local changes of microstructure in the base materials of high-strength steel (HSS), accompanied by great changes in mechanical properties. It is reported that softening occurred in heat affected zone (HAZ) can significantly lead to the strength deterioration of high-strength steel welded joints. However, macro tests including joint tensile test or hardness test can not reveal the load-bearing mechanism of the welded joints. In this paper, a microstructure-based representative volume element (RVE) model is performed and applied to characterize the mechanical performance of HAZ in high-strength steel Q690 welding joints. Firstly, some simulated HAZ-based tensile tests are performed to experimentally identify the stress-strain curves in the HAZ. Secondly, a well-trained deep neural network model is selected for segmenting metallographic structure from scanning electron microscope (SEM) images. Then, in-situ tensile tests based on SEM were performed with digital image correlation (DIC) to determine the strain distribution of different phases and phase boundary parts of HAZ during the tensile process. Meanwhile, RVE model representing specific areas of HAZ are constructed using corresponding segmented images. The results show the microstructure-based model provides a suitable method to explain softening or hardening mechanisms in HAZ of HSS welded joints. • A metallographic dataset of HAZ microstructures in high-strength steel was established. • A U-Net-based segmentation model was developed for automated microstructure phase recognition. • A 2D RVE model was constructed from metallographic images with real multiphase microstructures. • Local strain fields of HAZ were measured by SEM-DIC during in-situ tensile testing. • The RVE model reproduced experimental responses, enabling analysis of HAZ softening and hardening.