Digital Exploration Made Possible By High Performance Computing on GPUs 

Abstract:

The quest for simulation-driven product and process development is hampered by the reality that even today, only a minority of engineers make use of the power of 3D physics-based simulation software. The truth of the matter is that the vast majority of engineering software tools are difficult to use, learn and remember, do not provide results with useful response time, and do not have the functionality for true concept and design exploration. To respond to these issues, the Ansys Discovery project took on the challenge of creating the first simulation-driven design tool combining instant physics simulation, high-fidelity simulation, and interactive geometry modeling in a single easy-to-use experience. By combining interactive modeling and multiple simulation capabilities in a first-of-its-kind product, Discovery allows critical design questions to be answered earlier in the design process. This upfront approach to simulation saves time and effort on prototyping by supporting the exploration of multiple design concepts in real-time with no need to wait for simulation results. In this presentation, we will look behind the scenes at the technology and methods that make Ansys Discovery possible. The software system is based on a native GPU implementation of geometry modeling, automated repair, and discretization into elements or cells, massively parallel GPU solvers and GPU-based postprocessing. A variety of methods have been implemented including finite volume for CFD, finite elements for solid mechanics and thermal, finite elements for electrostatics, and unique topology optimization methods. While the usability of the software is much enhanced through an innovative GPU-based geometry modeling and mesh generation process that makes use of methods like level sets for simulation geometry representation and mesh generation (abstracted from the original CAD geometry), the other key element is highly performant solver methods that take advantage of many-core, stream computing on GPUs. The end result is unprecedented speed, automation, and robustness permitting design engineers to get repeatable results with a near-instantaneous response time.

Biography:

Dr. Dipankar Choudhury, Ansys Fellow, leads advanced technology and exploratory R&D in several areas at Ansys Inc. including High-Performance Computing. In addition, he leads Ansys’s Academic Program which includes coordinating research and collaborative education partnership programs with Academia for engineering simulation. Prior to taking this role, Dipankar was responsible for Ansys Inc.'s product strategy and planning, and corporate product management functions. Dipankar obtained his Ph.D. in the area of Computational Fluid Dynamics and Heat Transfer from the University of Minnesota. After his graduation, he held engineering and product management positions at Creare Inc. and Fluent Inc. where he was part of the founding group and the Chief Technology Officer. He is a member of the ASME and the AIAA and has technical publications in journals, conference proceedings, and trade magazines.