In the fields of advanced composite materials, cement-based materials, and geotechnical engineering, multi-scale mechanics, performance prediction, structural safety, and parameter inversion have become the core areas of scientific research and engineering implementation. Traditional methods heavily rely on empirical formulas, manual trial-and-error, and conventional numerical simulations. In critical aspects such as small-sample generalization, multi-field coupling solution, nonlinear constitutive modeling, and cross-scale correlations between micro- and macro-levels, these methods generally face significant challenges related to high computational costs, lengthy timelines, insufficient precision, poor interpretability, and difficulty in reproducing results from top-tier journals. These limitations severely constrain scientific innovation and engineering efficiency. The deep integration of artificial intelligence and computational mechanics, as well as materials science, is fundamentally reshaping the research paradigms in the field of materials and structures. Cutting-edge technologies such as machine learning, physical information neural networks (PINNs), deep energy methods, explainable AI (SHAP), and generative AI are deeply integrated with multi-scale modeling, finite element simulation, RVE automation, and geotechnical PDE solving. These technologies form a new generation of research systems driven by data and physical constraints, capable of efficiently uncovering hidden patterns in composition-structure-performance relationships, enabling strength prediction, life assessment, and intelligent optimization. They can also embed control equations and constitutive relationships into neural networks, significantly enhancing small-sample precision, inversion capabilities, and extrapolative generalizability, thereby fully bridging the entire chain of research technology from microscopic mechanisms to macroscopic structures.
