A neural model trained on physics simulations can predict structural behavior in milliseconds. The expensive solver remains in the loop for final candidates.
How it works
First, a set of expensive simulations is generated. Then the network learns to approximate those results quickly.
Weaknesses
This is an approximation, not the final certifiable truth.
Sub-technologies (4)
▶
Classical Surrogates
3 applications
Classical surrogate models in engineering simulation loop
Engineering & Simulation · Engineering
Early design-stage surrogate screening
Engineering & Simulation · Design
Concept-level feasibility surrogate screening
Engineering & Simulation · Concept
▶
ML Surrogates
2 applications
Approximate test-response prediction before full validation