10.30 Alex_Fischer - SimScale.pdf

1、Ready for AI?Building surrogate models via Cloud-native AI simulationSpeakerAlexander FischerPM&Co-founder-SimScaleRequirements Q=Flow rate(m3/h)H=Head(m)n=Rotating speed(rpm)P=Motor power(kW)=Q*H/P=Efficiency(%)Concept 3D Design Just hydraulic parts,no gaps Main CAD parameters:=inlet angle=outlet a

2、ngleNumber of blades.Detailed Design“As built”All gaps/details includedEarly 3D SimulationEarly feedback on hydraulic performanceEarly feedback on structural performanceInform concept designFinal 3D ValidationFinal hydraulic performance validationFinal structural performance validationConcept 1-2D D

3、esign Rough hydraulic design Analytical calculations via Euler equations,no losses included Many steps in the engineering design process will be AI-augmentedAI in the engineering design process Instant Result PreviewAI models trained large amounts of generic simulation data can be used to preview PD

4、E-based solver results.(Currently in development at SimScale)AI OptimizationSurrogate models derived from simulation data can be used to optimize without geometry creation at all.Example:KSB,Caeses&SimScaleRequirements EngineeringLLMs help structure requirements faster and cleaner and break it down

5、into atomic requirements.Example:Valispace AIDesign AutomationBased on requirements,geometry can(partially)be generated automatically using LLMs and CAD APIs,implicitly building a true digital thread Example:BlenderGPTAI Use Cases(Examples)AI SimulationAI models trained on design-specific,small amou

6、nts of simulation data can be used to replace PDE-based solvers.Example:Navasto FEASimulation AutomationProcedures during the simulation setup will be automatically completed/suggested based on previous human completion.Example:ADSK Assembly MatingSetup SuggestionSuggested setupBest template to useS