This research aims to develop a data-integrated evaluation methodology for materials subjected to creep-fatigue loading, with a focus on understanding the degradation mechanisms of mechanical properties and microstructural evolution under varying loading levels and dwell times. A systematic experimental program will be conducted to generate a multi-modal dataset capturing both microscopic features and macroscopic indicators. To leverage this data, the project will establish a hybrid deep learning framework. The CNN component is responsible for automated feature extraction from high-resolution microstructural images. These features will be fused with standardized physical parameters to form a high-dimensional feature space. The SVR component will then be used to achieve robust damage classification and accurate prediction of remaining life. Through Bayesian optimization, the CNN–SVR model architecture will be refined to maximize predictive performance and interpretability. This fusion-based methodology aims to overcome the limitations of traditional models that rely solely on physical properties or static imaging, allowing for a more dynamic and accurate evaluation of material degradation behavior. The proposed work contributes to the fundamental understanding of time-dependent damage processes and supports the development of intelligent lifetime design methods for structural materials operating under harsh conditions. The program’s resources and interdisciplinary network will play a critical role in promoting innovation at the intersection of materials science, mechanics, and artificial intelligence, ultimately advancing high-performance structural reliability solutions for clean energy applications.