Physics-Informed neural network based inversion and prediction of natural chloride diffusion in uncracked and cracked concrete systems

Key Points

• Chloride diffusion prediction can be accurately modeled using physics-informed neural networks, improving concrete durability.
• A significant correlation was observed in chloride levels over time, specifically in uncracked versus cracked concrete systems.
• Observational analysis utilizing neural networks and inversion techniques offers new insights into material behavior under conditions.
• Findings highlight the potential for advanced modeling to guide structural assessments and construction practices.