The microstructure and residual mechanical properties of several groups of T92/Super304H dissimilar steel welded joints (hereinafter referred to as welded joints) in service for 70,000~85,000 h were analyzed. The results show that the early service history of the welded joint results in the polygonization of the martensite lath and the coarsening of the precipitated phase on the side of T92 steel. In the further creep process, the cavities nucleate along the precipitated phase interface and the triple junction grain boundary. Under the same load, the creep life of the joint decreases rapidly with the increase in service time and, finally, type IV cracking occurs. Type IV cracking needs to meet two conditions: 1. a large degree of precipitated phase coarsening and cavity nucleation in the fine grain heat-affected zone (FGHAZ) and 2. a much lower loading stress than the yield strength.

In order to improve the recognition accuracy and model interpretability of metal fracture scanning electron microscope (SEM) images, this research presents an improved dual-branch model (IDBM) based on multi-scale feature fusion. This model employs VGG19 and Inception V3 as parallel branches to separately extract local texture features and global semantic features. Furthermore, it integrates channel and spatial attention mechanisms to enhance the responsiveness of discriminative regions. By integrating dual-branch features using a fixed fusion ratio of 0.8:0.2, the model was trained and validated on an image dataset comprising 800 representative fracture surface images across four categories: cleavage, dimple, fatigue, and intergranular fracture. The results indicate that under small-sample data conditions, the IDBM achieves a Validation Accuracy (Val ACC) of 99.50%, a Recall rate of 99.51%, and an Area Under The Curve (AUC) value of 0.9998, significantly outperforming single models and other fusion strategies. Through integration with class activation mapping (CAM) and feature space visualization analysis, the model exhibits strong interpretability. Furthermore, scale adaptability tests reveal that IDBM maintains stable recognition performance across a magnification range of 100 to 10,000 times, and identifies the optimal observation magnification ranges for the four types of fractures.