High Performance Computing of gas turbine flows: current and future trends
  • Dr. Nicolas Gourdain, CERFACS, Dr. Michel Gazaix, ONERA, University of Ljubljana
  • September 2013

Slide contents

  • energy = combustion
  • context
  • Why such flows are so challenging for CFD
  • The players: codes and humans
  • 2000 people in three centers
  • CERFACS has seven shareholders
  • elsA - a structured code for aerodynamics
  • AVBP - an unstructured code for reactive flows
  • these CFD codes are not commercial software with HMI
  • they are used to design this
  • they are used to design this or this
  • a large range of aeronautics
  • what does more CPU power really mean for industry
  • data from www.top500.org
  • turbulence is not CFD-friendly
  • numerical simulation and industry
  • how to predict turbulance
  • examples of numerical results
  • parallel computing efficiency -1
  • parallel computing efficiency -2
  • why my parallel efficiency can be so bad
  • mesh-partitioning and load balancing -examples of algorithms -1
  • mesh-partitioning and load balancing -examples of algorithms -2
  • mesh-partitioning and load balancing -examples of algorithms -3
  • let's talk about real life
  • mesh-partitioning and load balancing -real life is complicated for CFD
  • communication strategies -MPI
  • communication strategies -AVBP-1
  • communication strategies -AVBP-2
  • communication strategies -AVBP-3
  • another example
  • communication strategies -elsA-1
  • communication strategies -elsA-2
  • communication strategies -elsA-1
  • communication strategies -elsA-2
  • communication strategies -elsA-3
  • communication strategies -it is mandatory to reduce...
  • let's start with some HPC bottlenecks..
  • multi-physics in real gas turbine
  • mesh partitioning for structured mesh:impact of ghost cell-1
  • mesh partitioning for structured mesh:impact of ghost cell-2
  • parallel computing:load-balancing with multi-phase flows
  • mesh partitioning: the influence of particles-1
  • mesh partitioning: the influence of particles-2
  • quality assessment
  • quality assessment-communication strategy:MPI non-blocking calls-1
  • quality assessment-communication strategy:MPI non-blocking calls-2
  • quality assessment-communication strategy:MPI non-blocking calls-3
  • quality assessment-impact of rounding errors on LES-1
  • quality assessment-impact of rounding errors on LES-2
  • how CFD and HPC can help to design better products-overview1
  • how CFD and HPC can help to design better products-overview2
  • LES for combustion chambers -target configuration
  • simulation of thermo-acoustic instabilities
  • LES for combustion chambers
  • Ignition process in combustion chamber
  • simulation of compressor instabilities
  • simulation of compressor instabilities -surge cycle simulation
  • LES in a compressor stage
  • DNS for turbines
  • LES of jet noise and control-1
  • LES of jet noise and control-2
  • quantification of uncertainy -1
  • quantification of uncertainy -2
  • conclusion
  • thanks