GTC Digital Demo: Accelerating Scientific & Engineering Simulation workflows with AI

A new demo introduces the recently announced NVIDIA SimNet Toolkit, the first multi-physics (CFD and Heat Transfer) analysis using physics-informed neural networks.

Simulations form an integral part of product design to reduce significant iterations in physical prototyping and testing to improve quality, cost and time-to-market. However, this process is very time consuming and can take weeks to months since, in a typical simulation workflow, several iterations are involved if the results are not satisfactory for a given design. Typically, there is never enough time or compute power to examine all the design variations.

SimNet is an end-to-end AI-driven simulation framework based on a novel Physics Informed Neural Network (PINN) architecture. This demonstration of SimNet is solving a multi-Physics problem to perform automatic design space exploration, a thousand times faster than traditional simulation, with the accuracy of numerical solvers.

Such unprecedented throughput enables optimized design selection, which we show using SimNet. Completing these design tasks take seconds not hours, and complex design optimization can be completed in days instead of months.

Click on the video below to watch the demo.