Mitigating the Impact: Advancing Extreme Event Simulations with Machine Learning-Enhanced Meshfree Methods

We aim to develop software facilitating accurate and efficient simulations of extreme events, providing analysts with the tools they need to make better-informed decisions faster. In a world shaped by climate change, high-fidelity physical simulations have the potential to revolutionize disaster preparedness and infrastructure resilience. However, the computational expense and complexity of handling physical intricacies have posed significant barriers to the broader adoption of simulation. To overcome these challenges, we will integrate machine learning for data-driven material modeling with modern meshfree computational methods. Our initial focus will be on landslide simulations. In Phase I, our goal is to create proof-of-concept, modern, meshfree software by implementing Thermodynamically Consistent Recurrent Neural Network (TCRNN) models. This will address the challenges of modeling complex materials, marking a pivotal advancement for the next generation of extreme event simulation tools. By effectively tackling key challenges in extreme event simulation, our software facilitates the application of cutting-edge technology in government agencies and industries, expediting its use in addressing climate-related challenges.