Integrating MFEM for commercial IGA simulation acceleration

Automotive, defense, and other companies seek to bring safer products to market, faster, through leveraging engineering simulation software. However, mainstream software requires extraordinarily complex procedures to prepare computer-aided design data for simulation. Even today’s best methods for high-profile problems like simulating automotive crashes require millions of dollars of manual labor to prepare each simulation. After the analysis model is laboriously constructed, it often fails to perform robustly and instead produces low-fidelity or erroneous results, while also consuming vast amount of high-performance computing resources. This limits the opportunity for simulation to actually help create optimal designs. The company is leveraging its significant domain expertise to introduce a fundamentally new approach to computer-aided design and finite element analysis. Its patent-pending approach to isogeometric analysis allows design data to be transformed into a simulation far faster than traditional approaches and introduces a spectrum of ways for engineers to prepare simulations. With this technology, a simulation can be prepared with minimal input or some or all of the simulation mesh can be precisely controlled to allow for optimal computation. The method shifts the burden of providing the required accuracy onto computers. The proposed work will integrate a graphical-processing-unit-optimized library funded by the Advanced Scientific Computing Research Program with the company’s software to dramatically reduce computation time. The team has previously founded, grown, and exited a company in this space and is well-prepared to bring the resulting technology to commercial success. In Phase I, the company will complete a proof-of-concept integration. This will include improving the library support for isogeometric spline bases, developing optimal quadrature schemes to efficiently compute solutions using graphical processing units, and quantifying performance advantages. These efforts will position the company for a full commercial integration in Phase II. The developers of the library at a national laboratory will contribute as subcontractors to ensure success. The team will work closely with three global customers who seek to leverage this technology as they scale their own isogeometric analysis testing. The proposed methodology will remove error-prone parts of the simulation workflow and allow engineers of all experience levels to run analyses faster. This project promises significant bene- fits across industries, including improved harnessing of advanced manufacturing techniques and optimized design through easy-to-use simulation.