Siemens Adds AI Surrogate Modeling to STAR-CCM+ With Simcenter PhysicsAI

What happened: Siemens has released Simcenter PhysicsAI, an add-on to its Simcenter STAR-CCM+ simulation software. The tool applies geometric deep learning to speed AI-driven computational fluid dynamics design exploration. It lets engineers build AI reduced-order models from CFD data and run what-if studies far faster than conventional workflows.

Why it matters: Engineers can evaluate thousands of design variants using fewer computing resources. Siemens states the what-if exploration runs on the order of 1,000 times faster than traditional methods. Predictions are validated against high-fidelity CFD results, and GPU acceleration delivers up to 100 times faster predictions than CPU.

Industry context: The software trains predictive models using a transformer neural network architecture optimized for geometric data, drawing on historical results and prior Design of Experiments studies to reduce repeat CFD runs. Built-in error metrics quantify prediction accuracy. The product is part of the combined Siemens-Altair simulation ecosystem.

Our take: Shifting early-stage screening from solver runs to AI inference looks to solve the compute bottleneck in simulation, though validation against full CFD remains the stated reference point.