CFD design exploration speeds up with latest geometric deep learning

PRODUCT INFORMATION

---

MANUFACTURED BY Siemens

Siemens has announced an expansion of Siemens Simcenter advanced engineering simulation and test portfolio with Siemens Simcenter PhysicsAI software, which allows engineers to evaluate design concepts and variations more quickly.

For Control Engineering subscribers, engineering simulation tools using AI are important because they can help reduce model-evaluation time while supporting validation against high-fidelity CFD results.

Simcenter PhysicsAI uses geometric deep learning technology from Siemens’ Simcenter STAR-CCM+ software to support faster AI-based design exploration. The software enables engineers to create AI reduced-order models (ROMs) from computational fluid dynamics (CFD) simulation data and conduct design exploration more quickly, with Siemens stating that some workflows can run about 1,000 times faster while using fewer computational resources than traditional methods.

Simcenter PhysicsAI allows engineers to train and validate AI surrogate models using simulation results, run predictions on new geometries within their simulation environment and use high-fidelity CFD simulation as a validation reference. This approach can reduce turnaround time, support the evaluation of more design options and lower computing requirements during early-stage analysis.

Key capabilities

Simcenter PhysicsAI supports the creation and use of AI ROMs for:

• Design space exploration: Evaluate performance across multiple design variants using predictive models trained on geometric data.
• Reuse of simulation data: Train AI models using historical and newly generated simulation results, including data from previous Design of Experiments (DOE) studies, to reduce the need for additional CFD runs.
• Shorter iteration cycles: Use AI inference for early-stage design screening and include AI ROMs in optimization studies to assess design variants more quickly than with solver-based workflows alone.
• Model validation: Assess prediction accuracy using error metrics and validation tools to compare AI model results with simulation references.
• GPU -based processing: Use GPU acceleration to improve processing speed, with Siemens stating that predictions can run up to 100 times faster on GPU than on CPU.

Simcenter PhysicsAI is now available as an add-on to Simcenter STAR-CCM+.

Siemens

www.siemens.com

Edited by Puja Mitra, WTWH Media, for Control Engineering

RELATED PRODUCTS