Orca3D Support

We are often asked how to reduce the time that it takes to run a simulation without sacrificing the accuracy of the results.  Let's face it; CFD simulations require intensive computations and take a long time compared to empirical prediction methods. While getting a faster computer helps, it is not always the most practical approach to reducing simulation time.

We know that the finer the grid and the smaller the time step, the more accurate the results will generally be, but the consequence is longer running times. The discussions below address changes that you can make in each of these areas to potentially reduce run times, without sacrificing accuracy.

CFD Grid Size

One way to reduce simulation time is to make the CFD grid coarser. By default, the Orca3D Marine CFD template uses the "Normal" grid setting in SimericsMP. This grid setting has been designed to give accurate and reliable results over a wide range of vessel types. (While you will also find "Coarse" and "Medium Coarse" settings in the dialog, our recommendation is that in general these settings only be used as a way to speed up the simulation when you are learning to use the software, for example. They should not be used if you expect to obtain accurate results.)

But how do you know that "Normal" is good enough? It is always very important at the beginning of a project to conduct a "grid refinement" or "grid convergence" study in order to determine the appropriate grid setting to use for your particular design. At the highest speed or RPM that you will be analyzing, run multiple simulations where everything is held constant except the CFD Grid Size option. You should run at least three grid size settings (e.g., Normal, Medium-Fine, and Fine, although it's OK to include Coarse,  Medium Coarse, and Extra Fine if you have time, to better understand the relationship of the grid fineness to the results):

Compare the results (e.g., resistance, speed, pitch, heave) vs. the cell count for each of these simulations. You should find that at some point, increasing the cell count does not significantly change the results, so there isn't anything to be gained by going to a higher cell count. A word of caution here; while a coarse grid may seem to give results that are "satisfactory" in the sense that you might accept a 5% difference in drag, it's possible that the coarse grid is "missing" an important hydrodynamic phenomenon such as ventilation/non-ventilation of a step, separation of flow at the transom, etc. Therefore, it's not safe to assume that all of your results at other speeds and with various hull modifications will produce valid results.

In the example below, calm-water resistance runs for a 32-foot planing hull at 30 knots were done for six different Grid Refinement settings: Coarse, Medium Coarse, Normal, Medium Fine, Fine, and Extra Fine. In each case, the simulation time was seven seconds. The Resistance “Error” is the resistance at each of the settings divided by the resistance at the Extra Fine setting. 

The plot of Resistance data shows that results are converging to a consistent value as the number of cells increases. The same is true for the Pitch and Heave results. These results would confirm that in this example, the Normal grid setting is appropriate.

Notice that moving from Normal to Medium Fine roughly tripled the Run Time, even though there is only a 56% increase in the number of cells. This is because the time step was reduced to accommodate the smaller cells (you want to avoid particles of fluid “skipping” past cells entirely from one time step to the next, because that can cause the simulation to become unstable). 

Again, note that while the resistance for the Coarse grid was only 6.6% different than the Extra Fine grid, that doesn’t necessarily imply that it is suitable, even for a rough estimate. The grid may be too coarse to accurately capture a hydrodynamic phenomenon that could cause the results to be very different (e.g., a step ventilating with a Coarse grid when it would not with a Normal grid or in the physical world).

Time Step

Just as the simulation is spatially discretized into small cells, it is also temporally discretized into small time steps. The smaller the time step, the longer the simulation will take, but just as with the cell size, the more accurate the simulation will generally be. The Orca3D Marine CFD template determines a time step that is appropriate for the type and speed of your vessel. Sometimes, the time step tends to be on the conservative (i.e., smaller) side, because if the time step is too large the results will not be reliable and the simulation may even diverge and crash (sometimes dramatically, as if the vessel is going over a waterfall).

While you can also adjust the time step and conduct a "time step" convergence study, there is a better approach in SimericsMP that will adjust not only the time step, but also other settings that will help to make the simulation stable.

Rather than directly adjusting the time step, you can change the Numerical Option from the default of "Medium (Steady)" to "Aggressive (Steady)." This setting can make a significant change in run time. Note that this discussion applies to CFD simulations that reach a steady state such as calm water Resistance and Self-Propelled runs; this change should not be made when running in waves, with 6 DOF, or any other CFD simulation that is very dynamic in nature such as stepped planing hulls. For this same reason, if you intend to remove damping in order to investigate porpoising, it is best to stay with the Medium (Steady) option or revert to the Medium (Steady) option when you remove the damping. In some cases, even the Medium (Steady) setting may not have a small enough time step to accurately predict porpoising; if you see pitch oscillation when you remove damping, the next thing to try is changing from Medium (Steady) to Conservative (Steady) to see if the oscillation dies out.

This approach needs to be used with some caution, and not without a test at or close to the highest speed that you intend to use. This means running a simulation with Medium and Aggressive settings at the highest speed and comparing the results. If the Aggressive results are very close to the Medium results, then you can safely use Aggressive for your other speeds (but with the caveats regarding “dynamic” simulations noted previously). It's also important to note that with the larger time step the results, when plotted vs. time, may not be as smooth (although increasing the number of iterations to greater than 5 can sometimes help). If you are seeing oscillations that you don't expect, you can stop the simulation and change to "Conservative (Steady)" and re-start the simulation, to see whether the oscillations are physical or simply numerical. This will decrease the size of the time step and increase the number of iterations.

Ideally, the computer hardware that you are using with Orca3D Marine CFD is such that the default “Medium (Steady)” Numerical Option will provide accurate results in a timely fashion. But if you need to run your simulations more quickly, our general recommendation for steady state resistance and self-propelled runs in calm water is that it's better to have a finer grid (such as you get with a refinement zone, or with a Medium Fine or Fine grid setting) and use the Aggressive (Steady) setting, than it would be to use a coarser grid and the Medium (Steady) setting for the Numerical Option. 

To change to the Aggressive option, click on the Model tab and select the Marine Module. Then in the Properties pane, change from Template mode to Extended Mode. Finally, under Analysis Type, change from Medium (Steady) to Aggressive (Steady). We  have found that at times the simulation time and number of time steps that will actually be run differ from what is shown; for this reason, it's a good idea to save the file and then re-open it (the symptom of this is that the simulation will stop before you expected it to; you can just hit the Start button again if this happens). 

In the example below, calm-water resistance runs for a 32-foot planing hull at 30 knots were done for six different Grid Refinement settings; Coarse, Medium Coarse, Normal, Medium Fine, Fine, and Extra Fine, and each was run with both the Medium (Steady) and Aggressive (Steady) Numerical options. In all cases, the simulation time was seven seconds. The Resistance “Error” is the resistance at each of the settings divided by the resistance at the Extra Fine Grid setting using the Medium (Steady) Numerical option.

In this example, changing from Medium to Aggressive doubled or nearly doubled the size of the time step (among other changes) and reduced the time to run the simulation by more than half, with very similar results. Again, in this example the Normal Grid Refinement setting looks appropriate, and using the Aggressive (Steady) Numerical option is valid. 

Note that with the Aggressive setting, a Normal grid took only slightly longer to run than a Medium Coarse grid with the Medium setting, but with much better results. 

In summary, Orca3D Marine CFD is designed to provide accurate and reliable results in virtually all steady state simulations and is therefore sometimes a little conservative with its settings. This is done so that you can use it comfortably without the need to be a CFD specialist. But if you take some time at the beginning of a project to do a Grid Refinement study and look at both Medium and Aggressive Numerical options, you may be able to save significant time in your simulations without sacrificing accuracy.