Description:

CFD simulations at HPC scale are costly, and the design space is too vast to be sufficiently analyzed by grid search methods. Therefore, the focus of this workshop is to explore surrogate-based optimization strategies to efficiently guide the design decision process and parallel scalable deep learning methods to reduce the simulation costs. This workshop will provide an overview of the methods currently developed and used in the EU-funded Center of Excellence (CoE) Research on AI- and Simulation-Based Engineering at Exascale (RAISE). In a live demonstration, these methods will be exemplarily applied to CFD data provided by wall-resolved large-eddy simulations of the active drag reduction of turbulent boundary layer flows using spanwise traveling transversal surface waves.

Audience:
Academic and industry researchers

 

Agenda:

14:00-14:30 Introduction to surrogate-based optimization and parallel scalable deep learning (Fabian Hübenthal, RWTH)

14:30-15:00  Deep learning based analysis (Rakesh Sarma, FZJ)

15:00-15:30 High Performance Computing: Hardware and software (Eray Inanc, FZJ)

15:30-15:50 Live demo session

15:50-16:00 Q&A

 

Register here.