The field of physics-informed neural networks (PINNs) is rapidly advancing, with a focus on developing innovative methods to integrate physical laws and constraints into machine learning models. Recent developments have led to the creation of PINN architectures that can enforce hard nonlinear equality and inequality constraints, enabling more accurate and reliable predictions in complex systems. Another key direction is the application of PINNs to specific domains, such as thermomechanical monitoring of nuclear reactor components, semiconductor film deposition, and additive manufacturing. These approaches have shown promising results, including improved accuracy, reduced computational time, and enhanced robustness. Notably, the use of physics-informed surrogate models has emerged as a powerful tool for scalable simulation and prediction in various fields. Some notable papers in this area include: The paper on KKT-Hardnet, which proposes a PINN architecture that enforces both linear and nonlinear equality and inequality constraints up to machine precision. The work on Reinforced Graph-Based Physics-Informed Neural Networks Enhanced with Dynamic Weights, which combines physics-based supervision with advanced spatio-temporal learning for accurate estimation of Remaining Useful Life and State of Health.