Massively Parallel Adaptive Computational Fluid and Solid Dynamics for Engineering Applications

Many large scale engineering problems can be addressed by accurate simulation of the coupled continuum mechanics of fluids and solids. Accurate and efficient simulation of these phenomena requires the application of advanced numerical methods to ensure that high resolution accuracy is only applied to the critical areas of interest, which may vary over time, otherwise resources are wasted in high resolution simulation of relatively invariant materials. The resulting computational intensity variation creates dynamic load imbalances that are difficult to manage. The Quinoa project aims to address these challenges by implementing state of the art physics algorithms alongside the best known numerical methods for the simulation of fluids coupled with solid dynamics. The resulting dynamic load imbalance will be addressed by the underlying Charm++ based adaptive runtime system. Los Alamos National Laboratory will collaborate with Charmworks Inc. and Dr. Hong Lou PhD., to bring the Quinoa project up to production strength for the accurate, efficient, and highly scalable simulation of coupled fluid and solid dynamics. The Phase I work will focus on the development of an efficient distributed parallel overset unstructured grid infrastructure. Charmworks will perform the bulk of the work in implementing mesh to mesh data transfer, improving Quinoa’s ease of use, increasing the overall robustness, and improving interoperability with other elements of typical engineering workflow use cases. Existing commercial codes in this space are generally acknowledged to be inefficient when run on large computing systems, and to have exorbitantly expensive licensing models. This project will result in open source design software that will improve the ability of engineering and manufacturing sectors to create better products, in less time, at a lower cost.