Fluids

Fluids

Workbench Scripting for Linux Cluster

    • MaxF
      Subscriber
      Hello,nI?m a graduate student starting a project that requires large parametric studies using Fluent on the school?s Linux cluster. It seems easy enough to do this from Workbench on my personal computer, but I?m struggling to do the same calculations on the cluster. All the information I?ve seen online requires writing Iron Python scripts that create geometry, mesh, and setup in a .wbjn file. Is it possible to simply set up a project file with a parametric study set up, move that file to the cluster, and run it? Is there any related documentation I might find useful?nThanks,nMaxn
    • Rob
      Ansys Employee
      What are you actually trying to do? We tend to use optimisers and/or morpher technology over massive parametric runs with CFD as it's generally more economic. Full parameter sweeps are better for digital twins though. n
    • MaxF
      Subscriber
      Rob, I want to optimize the geometry of static mixers to maximize heat transfer / minimize cost. I've heard of Design Explorer and just recently of Design Point Service, but I'm not sure what will work best for me. Offloading the computations to my school's Linux cluster (or other compute resource) is crucial. Direct optimization would certainly be nice to implement, but simply doing a parameter sweep and fitting that data to a model like a neural network would also suffice. What do you suggest?n
    • Rob
      Ansys Employee
      OK, I assume you're using DesignModeler? If so, set up the initial model including output parameters on the Fluent case and run locally (if possible). Then set up the parametric domain, archive the project before putting it on the cluster, unarchive & run. Remember to retain the data. n
    • MaxF
      Subscriber
      I now have the archive file set up and in my working directory on the cluster as you instructed. I can open Ansys but I don't know how to unarchive and run the .wbpz file using the text user interface. n
    • DrAmine
      Ansys Employee
      Add more details: is the cluster using a job scheduler? Are you checking with cluster administrators?.One can run workbench from command line per script or journal but better via Remote Solver Manager.n
    • MaxF
      Subscriber
      I figured out how to run workbench on the cluster (with a GUI) and update the design points, but I suspect these are running on the head node of the cluster. The cluster uses the SLURM job scheduler and the administrator has noted that processes on the head nodes are terminated if ran for too long. Would RSM still be a good option?n
    • Rob
      Ansys Employee
      You may need a SLURM script or control code: unfortunately I have no idea how that works so IT may need to look at the SLURM guides. n
    • DrAmine
      Ansys Employee
      RSM supports SLURM and so you can couple WB updates via RSM and let the jobs being scheduled via SLURM.nnYour Cluster Admin might have a look into this: https://ansyshelp.ansys.com/account/secured?returnurl=/Views/Secured/corp/v202/en/wb_rsm/rsm_custserversideint.htmln
Viewing 8 reply threads
  • You must be logged in to reply to this topic.