Integrating MFEM for Commercial IGA Simulation Acceleration

C53-02b-271246Computer simulation allows for testing of parts and assemblies without having to manufacture and physically manipulate them, which can save significant time and money in the product development process. Automotive, defense, and other industries use simulation to bring safer products to market faster. Unfortunately, mainstream simulation software is incompatible with the data structures used for design models. This means the design models must be approximated via a meshing process into a different mathematical format before they can be simulated. Even today’s best methods for problems like simulating automotive crashes require millions of dollars of manual labor to prepare each simulation. And even after all this work, these approximated simulation models take up vast amounts of high-performance computing resources. While simulation has an important role in product development, these inefficiencies limit how much it can actually be used to perfect new designs. The company is introducing radically new, next-generation computer simulation technology that avoids many of these inefficiencies. Its patent-pending approach to isogeometric analysis allows finite element analysis directly on design models, without the approximation step. It allows users to manage the trade-offs between manual and automated pre-processing. An engineer can choose to spend time preparing an exact simulation model and reaping the benefits of faster computation, or can choose a fast, automated model preparation which places a higher computing burden on the computer. Most importantly, and unique to this new technology, is that the engineer can choose any level in between. They can invest varying amounts of manual model preparation time depending on their needs. This will allow unprecedented accuracy and flexibility in computer simulation of new products across many industries. The proposed work in this project will integrate a graphical-processing-unit-optimized library, funded by the Advanced Scientific Computing Research Program, with the company’s isogeometric analysis software to dramatically reduce computation time across the entire spectrum of simulations run with isogeometric analysis. In Phase I, the company collaborated with a national laboratory to add support for spline bases to the library, develop new quadrature schemes, and complete a proof-of-concept integration. The proof of concept was successful and demonstrated significant speed improvements. During Phase II, the company will expand the proof of concept and develop an efficient, fully commercializable, isogeometric analysis solver for general purpose use. Specifically, it will extend the capability of the library to process structured data, complete the integration of the library with its own commercial software, and optimize this software to take full advantage of the library. The ability to run computer simulations on design models without the inefficient meshing step will have benefits for many industries. With isogeometric analysis removing data approximations and delivering more accurate simulation results, these companies can reduce or even eliminate physical prototyping. Better products, more optimally designed, can get to market faster for consumers. Markets that are concerned about consumer safety, such as automotive and defense, will be able to more easily develop new designs because they will no longer have to physically build and crash their products to test them for safety. Designers who are concerned about developing energy-efficient products will be able to design, simulate, and optimize their designs for energy efficiency much more quickly and at less cost.