Simcenter STAR-CCM+ Simcenter STAR-CCM+: Do I benefit from running Spray A with AMR?

Simcenter STAR-CCM+ Simcenter STAR-CCM+ Virtual Reality Teamcenter Share Simcenter Cloud HPC Simcenter STAR-CCM+ Viewer Simcenter STAR-CCM+ Application Specific Solutions


This example article shows how AMR can be used for spray A RANS simulation and what (mainly run time) computational benefits it may have.


AMR was introduced in Simcenter STAR-CCM+ in 2020.1.
Starting 2020.2 you will be able to use it for reacting flows as well (please note that Simcenter STAR-CCM+ In-cylinder does not support AMR in this release). 

This example KB demonstrates and gives suggestions:
  • How to activate AMR, 
  • What values to use for the coarsening/refinement, and 
  • What scalar fields they should be based on?
Below you find a reference for the first spray A case that is run with a static grid. The results from that study will be compared to the Spray A AMR case shown here. As for the static Spray A case, the liquid penetration length and lift-off length (using OH scalar) is monitored.

Select the Adaptive Mesh model under Optional Models in the Model Selection for the Physics Continua. 
User-added image
A transition width of two cells is used.
User-added image
Two scalar fields are used for refinement. 
  1. Two-fold refinement for the fuel species C12H26 without any coarsening, and
  2. Three-fold refinement for temperature, with coarsening above and below the defined range (the temperature is never above the maximum temperature that is specified for the range limits)
Settings for the Fuel Species
User-added image

Settings for the Temperature scalar
User-added image

As indicated above, it would be possible to specify the Fuel range using the method used for Temperature, and vice versa. 
The complete settings can be found in the attached sim-file.

The computational domain

The idea here has been to try to make an example of how you would run the simulation the first time around, with no exact idea in mind of where the flame lift-off would be and how far the spray would penetrate. For that reason, the experimental lift-off has been used as a reference and it turns out that for this specific case and mechanism in use, the RANS simulation gives a good comparison to the experiments.

The static mesh is built with two refinement boxes, corresponding to the refinement that AMR would yield, thus creating quite a large cell count. Running the simulation a 2nd time around with the experience from the first run, we could easily reduce the cell count. However, and keep it in mind, the number of simulations have doubled. If we were to run 10 different spray cases, this corresponds to 20 simulations.
With AMR, we let the numerics handle where the refinement lands. 
User-added image


First, in the picture below, we see the total run time and the number of cells used by the static mesh simulation and AMR simulation.
Reading the left axis, looking at the black curves, we can make out that the total run time on 24 CPUs took approximately 20 hours on the static mesh and 4 hours on the AMR mesh.
Looking at the red curves and the right axis, it is seen that the static mesh has approximately three times more cells compared to the final AMR mesh. 
User-added image

The liquid length for the two CFD runs are compared with the experimental liquid length below. The two CFD runs have close to identical results.
User-added image

Finally, qualitatively we can compare the shape of the spray and temperature/OH fields. These too, are close to identical. 
User-added image

To answer the original question, yes it would benefit you to run spray A with AMR. 

KB Article ID# KB000042270_EN_US



Associated Components

Design Manager Electronics Cooling In-Cylinder (STAR-ICE) Job Manager Simcenter STAR-CCM+