In recent years, deep learning has attracted widespread attention in empowering various fields. Physics-informed machine learning deeply integrates data with physical equations and solves partial differential equations using artificial intelligence algorithms, which has become the focus of computational mechanics research. Compared with traditional numerical methods, Physics-informed machine learning shows significant advantages in solving inverse problems, handling high-dimensional nonlinear systems, and addressing multiphysics coupling challenges. This research provides a comprehensive overview of recent advancements in physics-informed machine learning and highlights our research group's contributions across multiple directions. Specifically, we present our original work on: 1) Physics-informed neural networks for inversion analysis in the Three Gorges Reservoir area, 2) Physics-informed kernel function neural networks, 3) Curriculum-transfer learning-based physics-informed networks, 4) Transfer learning-based physics-informed Kolmogorov-Arnold networks, and 5) DeepONet based on Physics-informed kernel functions. These methodological innovations incorporate kernel function-based prior constraints, domain decomposition techniques, and novel network architectures, collectively achieving significant improvements in both accuracy and efficiency for solving complex forward and inverse mechanical problems.