Improving Virtually Guided Certification for Product Design with Implicit FEA Solver in LS-DYNA

Seid Koric, University of Illinois at Urbana-Champaign

Usage Details

This project aims to advance current state-of-the-art computationally-intensive models in the implicit finite element software LS-DYNA and to remove the scaling barriers in its multifrontal direct sparse solver. This proposal will build on the recent success with the explicit solver in LS-DYNA on Blue Waters. Our effort to enhance the parallel scaling of LS-DYNA will be done in the context of enabling virtual certification of gas turbine engines. The implicit solver generally performs more efficiently for high-fidelity thermal-mechanical analysis of jet engine processes than its explicit counterpart. Rolls Royce will provide large whole-engine FEA models for testing that are representative of gas turbine engine problems used in virtually guided certification for risk mitigation, to support physical engine testing, and for post-test validation of modeling and simulation techniques. The PI, in his recent academic work on Blue Waters with a similar multifrontal direct solver, has demonstrated an unprecedented level of scalability on over 65,000 cores, clearly demonstrating the feasibility of the proposed work in this project.



Given the popularity of implicit finite element methods, the results of this effort will further open the door for the high-fidelity multiphysics modeling of complex structures and assemblies in real scale in both industry and academia, and will be leveraged for future UIUC-led interdisciplinary proposals to DMDII and other funding agencies. Furthermore, lessons learned for increasing code performance at scale will be shared with both industrial and academic communities.