-
Notifications
You must be signed in to change notification settings - Fork 631
FDS Validation Process
This is a summary of how to run and process the FDS validation cases. These cases are simulations of a variety of experiments. These cases are run and processed with each minor release of FDS.
Note that if all you want to do is compile the FDS Validation Guide, make sure you have all the software except the Fortran compiler, check out the FDS Repository, and skip to the last step, "Compiling the FDS Validation Guide."
All of the input files are located in the Validation folder of the FDS Repository. The routines used to automate this process are mainly shell scripts that are appropriate for a unix or linux operating system. Running and processing the Validation Suite on a Windows PC would require the various scripts to be converted to ".bat" files.
- Git client and an account on GitHub
- Fortran 2008 compliant compiler (You need this only if you are going to run cases)
- Open MPI or other MPI libraries (You need this only if you are going to run cases)
- Matlab
- LaTeX, preferably PDF-LaTeX.
The process has a number of prerequisite steps:
- Check out the fds, exp, and out repositories under the Fire Models organization. The instructions can be found at Git User Workflow.
- If you plan to run the validation cases, compile FDS by running the script
make_fds.shin the appropriate folder of the FDS_Compilation directory. You may have to modify the run scripts if your executable is different than that which is set. The default is currently for 64 bit linux.
Each sub-directory under Validation represents an experimental test series. The experimental data is kept under version control in a neighboring repository with the same sub-directory structure called "exp". In the fds repo under Validation/CASENAME/ you will see a folder called FDS_Input_Files. It is suggested that you not run the cases within this folders because if you run the cases again and again, you will end up over-writing previous results that you might want to consult later on. Instead, create a folder called Current_Results at the same level as the original three. On a linux or unix platform you can link Current_Results to whatever other folder you'd like. The name Current_Results is important in the automation process.
Unlike the Verification Suite, there is no single script that runs all the Validation cases. There are too many, and it requires running parallel jobs over a cluster of computers. Instead, do the following:
- cd to the Validation folder of the Repository.
- One by one, descend into the folders, create a
Current_Resultsfolder (or link), copy the FDS inputs there, and run them. This can be scripted if you want. Each validation case contains a linux/unix shell script calledRun_All.shwhich automatically runs the input files in theCurrent_Resultsfolder. These scripts were designed for a linux batch queuing system.
- cd to the Validation folder of the Repository.
- Type:
./Process_All_Output.cshor process individual cases by running./Process_Output.sh. Make sure that the file is executable. Otherwise, just type:chmod +x *sh. The Process_Outpus.sh script will copy the output files from Current_Results into the corresponding sub-directory in the neighboring "out" repository (where output files are kept under version control).
You do not have to rerun the Validation Suite if you just want to compile the FDS Validation Guide. But you do have to run the Matlab script FDS_validation_script.m to make all the figures. Here is how you do it:
- Start up Matlab and cd to the
Utilities/Matlabfolder of the Repository. - Type:
FDS_validation_scriptat the Matlab command prompt. - After about 15 to 30 minutes (depending on your computer speed), cd to
Manuals/FDS_Validation_Guidein the FDS Repository and at the command prompt type:make_guide.sh. This script runs pdflatex and bibtex such that the FDS Validation Guide is fully compiled with all cross-references properly established.