High-fidelity numerical simulations are a powerful tool for modeling and analyzing complex physical systems. These simulations use advanced mathematical models and computational algorithms to generate highly detailed and accurate representations of phenomena such as fluid flow, heat transfer, and molecular dynamics. However, the increasing complexity and scale of these simulations have led to a dramatic increase in the amount of data generated, making it difficult to store, transmit, and analyze the results efficiently. By reducing the size of the data required to represent the simulation results, data compression can significantly improve the efficiency of data writes, data storage, data transmission, and data analysis. In nuclear engineering, high-fidelity numerical simulations are used to model and analyze the behavior of nuclear reactors and other nuclear systems. These simulations are critical for designing and optimizing nuclear power plants, evaluating the safety of nuclear facilities, and developing new nuclear technologies. In these applications, the ability to compress simulation data can have a significant impact on the overall design process, reducing costs and improving performance.
MLDC will be a robust framework for in-situ, scalable, and data-driven compression in high-fidelity numerical simulations. The defining feature of this framework will be a compression algorithm that presents simulation data in a format that occupies less memory, reduces read and write times, and enables fast knowledge extraction. The feature set will be realized using Data-informed Local Subspaces (DLS) data compression, a novel data compression methodology that uses modern machine learning-based feature extraction and the generalized finite element method (GFEM) to compress high-fidelity data into accurate reduced representations that can be queried in real-time.
The MLDC framework will provide scientists and engineers with a robust mechanism for using the considerable knowledge gained from high-fidelity simulations in real-world systems engineering tasks. This will accelerate the pipeline for transitioning ideas from simulation, through design, and into production, all while helping ensure the DOE and other agencies are capable of fully realizing the potential of exascale computing resources and high-fidelity simulation software.