Nuclear reactor safety assessment relies on the ability to predict the structural response of critical components under extreme conditions. One such component is the pressurizer. In Pressurized Water Reactors (PWR), the pressurizer plays a vital role in maintaining and regulating the pressure of the primary system. During severe accident scenarios, such as a loss of power or loss of flow, the pressurizer wall is subjected to significant thermal transients. Heat propagates radially from the reactor coolant through the vessel wall, generating evolving temperature gradients. Understanding the temperature distribution within these multi-layered walls is essential for evaluating structural integrity and thermal fatigue. Traditional safety analysis tools, such as RELAP and MELCOR, provide detailed predictions of thermal behavior in reactor components. While accurate, these methods are computationally intensive and not easily scalable for large parametric studies. They are also slow for real-time decision-making or efficient uncertainty quantification. In this work, we explore surrogate modeling approaches, ranging from polynomial regression to deep neural networks. Building on this, we developed a physics-informed neural network (PINN) that incorporates the governing heat conduction equation directly into the learning process. By integrating data and physical laws, the PINN provides a robust and efficient framework for predicting temperature distributions. This approach offers a promising path toward real-time nuclear safety assessment and advanced reactor monitoring.
This cover image was created using Al-generated content from "Canva". The authors confirm that no human likenesses, copyrighted elements, or misleading representations are included in the image.
View this paper
Login
Register
Submit a Paper
Propose a Special lssue