Functional Improvements and Technical Developments of a Community-driven and Physics-informed Numerical Library for Disruption Studies

The DisruptionPy library, historically developed by the Disruption Studies Group at MIT PSFC, is designed to tackle computational disruption research. The original MATLAB workflows have since been translated into Python in order to share knowledge and tools and lower any barrier towards effective AI/ML simulations. Under the framework of the DOE grant "Open and FAIR Fusion for Machine Learning Applications", DisruptionPy has been further developed and aligned to best software practices. First, we started a thorough review of the core physics methods of the library to ensure physical accuracy and reproducibility of the workflows. Then, we applied Continuous Integration and Continuous Deployment practices to systematically test each new commit to prevent unwanted modifications, and subsequently update public installations available at machine-specific cluster installations. Finally, we enforced code style throughout the codebase and implemented systematic linting automation to adhere to established coding standards. Such thorough review is paramount due to the imminent open sourcing of the DisruptionPy library, which will then be ready for code contributions from the broader Fusion community.