Synthetic Multiphysics Data for AI Training - Archived

Physics AI models are only as good as the data used to train them, yet high-quality physics-modeling-based datasets are often scarce, expensive, or difficult to obtain experimentally. In this webinar, we will discuss how the COMSOL Multiphysics^® software can be used to generate synthetic data for training physics AI models by simulating single-physics and coupled physics systems for a range of design parameters. There will be simulation examples from structural mechanics, CFD, electromagnetics, acoustics, chemical, and electrochemical engineering.

The presentation will highlight efficient parametric sweeps using design of experiments methods, data extraction, and cluster sweeps, as well as examples of trained physics AI models that can be developed for use in simulation apps, optimization, uncertainty quantification, and digital twins.