Attention

Following the recent operating system upgrade, software listings and module load commands now reflect the upgraded system. Some examples and notices are still being updated and may contain outdated guidance.

Fluent

Guidance last updated: 16 October 2026

ANSYS Fluent is a Computational Fluid Dynamics (CFD) code for modelling fluid flow, heat transfer, mass transfer and chemical reactions. Fluent has interfaces to other pre and post processing packages supplied by ANSYS such as ICEMCFD or Ensight and can incorporate user developed models via user defined functions. Fluent can make use of built in MPI to utilise multiple cross node CPU and can scale to hundreds of cores.

Caution

ANSYS users can now use a maximum of 400 concurrent cores, across all devices on campus, including personal machines and HPC clusters, and regardless of the number of ANSYS applications they have open. Please see ANSYS license restrictions for more detail.

module load ANSYS/2026R1
module load ANSYS/2023R2

Batch jobs

You can submit Fluent batch jobs to the scheduler. A Fluent batch job is composed of a Fluent journal file which tells Fluent what to do and a submission script which tells the scheduler how much resource to request and how to start Fluent.

Fluent Journal files

A Fluent journal file is effectively a sequence of TUI and/or Scheme commands that you’ll supply to Fluent instead of using the GUI / TUI. A more thorough explanation and tutorial on how to make a Fluent journal file can be found on the following page: Writing Fluent journal files.

Important

It is critical that you consult with the Fluent journal syntax documentation linked above in order to ensure features such as autosaving and intermediate save states are turned on and functioning correctly!

Batch Submission Script

On Stanage, Fluent jobs can either be run on just one node (SMP) or across multiple nodes, (but will use its in-built MPI communications for both).

Example SMP Fluent Scheduler Job Script

The following is an example batch submission script, cfd_job.sh, to run the executable fluent with input journal file test.jou, and carry out a 2D double precision CFD simulation. The script requests 4 cores with a runtime of 60 mins and 8 GB of real memory per node.

Hint

  • The 2ddp argument is used to specify a 2D double precision simulation. Valid values include: 2d, 2ddp, 3d and 3ddp.

  • The #SBATCH --nodes=1 asks the scheduler for 1 node.

  • The #SBATCH --ntasks-per-node=1 asks the scheduler for 1 task per node.

  • The #SBATCH --cpus-per-task=4 asks the scheduler for 4 cores in the single task.

  • The host file supplies short node names with -cnf to avoid Fluent passing fully qualified hostnames that Slurm does not recognise.

  • The arguments -gu and -driver null instruct Fluent that it will be running with no GUI to avoid errors caused by plot / figure export.

  • The argument -sifile=./"$SLURM_JOB_ID"_fluent_server_info.txt tells Fluent to create a file in the working directory with the remote visualisation server info.

#!/bin/bash
#SBATCH --nodes=1
#SBATCH --ntasks-per-node=1
#SBATCH --cpus-per-task=4
#SBATCH --mem=8000
#SBATCH --job-name=name_fluent_smp_4
#SBATCH --output=output_fluent_smp_4
#SBATCH --time=01:00:00
#SBATCH --mail-user=a.person@sheffield.ac.uk
#SBATCH --mail-type=ALL
module load ANSYS/2026R1

scontrol show hostnames "$SLURM_JOB_NODELIST" > "hosts.$SLURM_JOB_ID"

fluent 2ddp -t"$SLURM_CPUS_PER_TASK" -cnf="hosts.$SLURM_JOB_ID" -gu -driver null -sifile=./"$SLURM_JOB_ID"_fluent_server_info.txt -i test.jou

Tip

You may have noted that this SMP job uses different arguments from the MPI jobs below. The use of a single task is intentional and is in order to ensure all requested memory within a job is available to all Fluent application processes in a single CGroup. This should avoid asymmetric memory requirements for the host Fluent process during start up in some models causing crashes due to insufficient memory being available and the scheduler terminating the job.

The job is submitted to the queue by typing:

sbatch cfd_job.sh

Example MPI Fluent Scheduler Job Scripts

As SLURM is capable of making MPI job resource requests very specifically, two scripts are provided below. The first is a “generic” job submission with the scheduler allocating MPI tasks and cores anywhere it can freely find them available in the cluster. The second script is a “specific” example with the scheduler being explicitly told to allocate tasks with cores specifically across 4 nodes.

Tip

A generic request with #SBATCH --ntasks=n allows the scheduler to choose the placement of tasks, which may reduce queue times.

The following is an example of a “generic” batch submission script, cfd_job.sh. It will run the executable fluent with input journal file test.jou, and carry out a 2D double precision CFD simulation. With --ntasks=4, Fluent uses four cores allocated on one or more nodes. The script requests 4 cores, 1 core per task (the default) with 4 tasks, with a runtime of 60 mins and 2 GB of real memory per core.

Hint

  • The #SBATCH --ntasks=4 asks the scheduler for 4 tasks (each with a single core by default) which are allocated on any node across the cluster as no further specific options are supplied.

  • The srun hostname -s > "hosts.$SLURM_JOB_ID" section sets up the hostlist which is required for correct MPI task spawning in conjunction with the -cnf="hosts.$SLURM_JOB_ID" argument.

  • The 2ddp argument is used to specify a 2D double precision simulation. Valid values include: 2d, 2ddp, 3d and 3ddp.

  • The argument $SLURM_NTASKS is a SLURM scheduler variable which will return the requested number of tasks.

  • The argument -mpi=intel instructs Fluent to use the Intel MPI communication method. Consult Fluent documentation for OpenMPI instructions if applicable.

  • The argument -scheduler_tight_coupling instructs Fluent to use Slurm to efficiently and safely do task spawning.

  • The arguments -gu and -driver null instruct Fluent that it will be running with no GUI to avoid errors caused by plot / figure export.

  • The argument -pib.infinipath instructs Fluent to use the high performance Omnipath networking.

  • The argument -sifile=./"$SLURM_JOB_ID"_fluent_server_info.txt tells Fluent to create a file in the working directory with the remote visualisation server info.

#!/bin/bash
#SBATCH --mem-per-cpu=2000
#SBATCH --ntasks=4
#SBATCH --job-name=name_fluent_mpi_4_generic
#SBATCH --output=output_fluent_mpi_4_generic
#SBATCH --time=01:00:00
#SBATCH --mail-user=a.person@sheffield.ac.uk
#SBATCH --mail-type=ALL
module load ANSYS/2026R1

srun hostname -s > "hosts.$SLURM_JOB_ID"

fluent 2ddp -t"$SLURM_NTASKS" -mpi=intel -scheduler_tight_coupling -cnf="hosts.$SLURM_JOB_ID" -gu -driver null  -pib.infinipath -sifile=./"$SLURM_JOB_ID"_fluent_server_info.txt -i test.jou

The following is an example “specific” batch submission script, cfd_job.sh. Like the “generic” script, it runs fluent with test.jou for a 2D double precision CFD simulation on four CPU cores. This “specific” script requests one core on each of four nodes. The script requests 4 cores (1 core per task, 1 task per node on 4 nodes) with a runtime of 60 mins and 2 GB of real memory per core.

Hint

  • The #SBATCH --nodes=4 asks the scheduler for 4 nodes.

  • The #SBATCH --ntasks-per-node=1 asks the scheduler for 1 task per node.

  • The #SBATCH --cpus-per-task=1 asks the scheduler for 1 core in each task.

The above combined options taken together instruct the scheduler to allocate the job resources as described above, 1 core per task, 1 task per node on 4 nodes = 4 cores. These options can be combined together in other ways which may be optimal for certain Fluent models. e.g. some may benefit from higher CPU performance, greater memory bandwidth or other optimised resource requests.

#!/bin/bash
#SBATCH --nodes=4
#SBATCH --ntasks-per-node=1
#SBATCH --cpus-per-task=1
#SBATCH --mem-per-cpu=2000
#SBATCH --job-name=name_fluent_mpi_4_specific
#SBATCH --output=output_fluent_mpi_4_specific
#SBATCH --time=01:00:00
#SBATCH --mail-user=a.person@sheffield.ac.uk
#SBATCH --mail-type=ALL
module load ANSYS/2026R1

srun hostname -s > "hosts.$SLURM_JOB_ID"

fluent 2ddp -t"$SLURM_NTASKS" -mpi=intel -scheduler_tight_coupling -cnf="hosts.$SLURM_JOB_ID" -gu -driver null  -pib.infinipath -sifile=./"$SLURM_JOB_ID"_fluent_server_info.txt -i test.jou

Either job can then be submitted to the queue by typing:

sbatch cfd_job.sh

Exporting GUI Fluent plots in batch submission jobs

While in a batch job you cannot call a display of a window / figure / animation as Fluent has nowhere to direct this output (there is no available display.)

To avoid this issue while attempting to export images from a batch job you should ensure the fluent command in your submission script has the -gu and -driver null arguments.

For a batch job:

fluent 2ddp -i test.jou -gu -t$SLURM_NTASKS -driver null

You can setup your plots / graphs either in your journal file or interactively via the GUI and saving the case file.

You can then instruct Fluent to make a hard copy after each figure/animation in the you journal file with:

/display/hard-copy "myfile.tif"

Ensure that you select the correct file format as some file types will not support animations.

For an example of the journal file commands see below:

/display set-window xx ;(xx -> window number)
/display set contours surface yy ;(surface id = yy)
/display set contours filled-contours yes
/display contour zz vof 0 1 ;(contour of volume fraction field (vof) of zz phase, min = 0, max = 1)
/display hard-copy "vof-%t.jpg" ;(save the image file in .jpg format)

See Ansys’s Saving Picture Files documentation for further details.

ANSYS Fluent training and help resources

Important

Academic support requests should be directed to the IT Services’ Research and Innovation team or the Ansys academic support resources (ensure you register with your University email for priority support).

ANSYS provides numerous academic training and help resources including tutorials, video lectures and examples for computational fluid dynamics products. A short list of the resources ANSYS maintains is summarised below: