AI Physics End-to-End Workflow Demo

AI-Enabled Engineering Workflow Overview

Learn more about AI Physics Powered by NVIDIA on Rescale

Step 1 – Generate or Use Existing Simulation Data. Run Multiple CFD Simulations in Batch or Large Scale DOE, in this case with Siemens STAR-CCM+ to conduct a rear vehicle wing design study.

Step 2 – Train Physics-Informed Surrogate Model by Ingesting simulation training data and run neural net training workflow, in this case with NAVASTO NAVPACK run on Rescale.

Step 3 – Run AI Inference to Predict New Design Performance to Evaluate thousands of new geometries for feedback in minutes or single geometries in milliseconds. AI inference software used – NAVASTO NAVPACK. Then assess prediction results quantitatively or visually within a Workstation (e.g. Paraview).

Step 4 – Validate Designs to Determine Prediction Accuracy. In this case the same CFD solver, Siemens STAR-CCM+ is used to validate. Repeat steps 1-3 as needed for AI model tuning.

Results: 1000x acceleration of aerodynamics optimization design predictions compared to traditional simulation analysis