Development of Artificial Neural Network (ANN) models for Critical Heat Flux predictions

Safety is a fundamental aspect of the nuclear industry, particularly in the aftermath of the Chernobyl and Fukushima accidents. One of the most critical phenomena affecting reactor safety is the Critical Heat Flux (CHF), which marks the transition to a deteriorated heat transfer regime in nuclear reactor cores, potentially leading to severe damage. Accurate CHF prediction is therefore crucial for enhancing nuclear safety and reactor performance. Since 2006, CHF prediction has largely relied on the Look-Up Table (LUT) method, a well-established empirical approach. However, with advancements in computational techniques, machine learning (ML) has emerged as a promising tool to improve prediction accuracy. In response, the Task Force on Artificial Intelligence and Machine Learning for Scientific Computing in Nuclear Engineering, under the supervision of the Expert Group on Reactor Systems Multi-Physics (EGMUP), has been actively developing ML-based models for CHF prediction. Within this framework, this thesis work first provides a theoretical analysis of CHF, examining its governing physical mechanisms and the most influential parameters during the operation of a Pressurized Water Reactor (PWR). A comprehensive dataset (from USNRC) is analyzed to determine the most relevant input variables through a correlation analysis. Subsequently, bootstrapped Artificial Neural Network (ANN) models are developed (and evaluated against reference empirical methods, e.g., LUT) to assess their potential in improving CHF prediction accuracy and precision. The results highlight the advantages of ANN-based approaches, demonstrating their potential to enhance safety margins and optimize reactor operation. Given the increasing complexity of nuclear systems and the growing demand for safe and efficient energy production, the development of advanced predictive models will play a strategic role in the future of nuclear engineering, enabling more reliable safety assessments and improved reactor design.