Siemens has launched 'Simcenter PhysicsAI,' which accelerates the CFD (Computational Fluid Dynamics)-based design review process using AI. Engineers can now leverage existing simulation data to evaluate thousands of design variations in a short time and rapidly identify candidate designs in the early stages of product development.

Siemens announced on the 2nd that it is expanding its Simcenter engineering simulation and test solution portfolio through Simcenter PhysicsAI. This software is provided as an add-on to Simcenter STAR-CCM+, a CFD analysis software.

CFD is a core analysis technology used to analyze fluid flow, heat transfer, and aerodynamic performance. However, high-fidelity analysis requires long computation times and consumes significant computing resources, which can place a burden on development schedules in fields involving frequent iterative reviews, such as automotive aerodynamic design or electronic equipment cooling design.

Simcenter PhysicsAI generates AI reduced-order models (ROM) by learning from existing CFD results and DOE (Design of Experiments) data. ROM is a predictive model that simplifies complex analysis results and is used to rapidly estimate the performance of new design geometries. Siemens has applied geometric deep learning technology to this, which learns 3D geometric data such as CAD geometry or meshes.

According to the company, Simcenter PhysicsAI supports design exploration up to 1,000 times faster than existing workflows. GPU-based predictions are performed up to 100 times faster than CPUs, and built-in verification features allow comparison of AI prediction results with high-fidelity CFD analysis results.

Simcenter PhysicsAI is closer to an auxiliary tool for rapidly screening initial design candidates than a replacement for physics-based analysis. As the manufacturing industry demands both shortened development times and reduced computing costs, AI-based design exploration that reuses existing simulation data is expected to become a major driver of change in engineering workflows.