Reading Time: 20 min

Per-Simulation Time: 25 s

Per-Simulation Cost: 0.4

Target audience: new-to-Luminary

In this tutorial we will build from the single simulation you ran in the first NACA 0012 Tutorial. This Tutorial assumes you have a NACA 0012 Project with at least one converged solution produced.

The goal here is to describe a few different ways to run your next simulation, because most engineering is not about one Simulation, but about analyzing and comparing many simulations. Most of the basic principles about setting up the case were tackled in the first Tutorial, and will not be reproduced here. What we will introduce is:

What you should have in your Project in a fully completed simulation setup (Physics, Outputs, Solver) as well as at least one simulation available in the Results->All Results table:

To create a new simulation, simply go to any of the Physics, Outputs, or Solver tabs and make an adjustment to a parameter. One easy example to try is to adjust the angle-of-attack for the simulation and run again.

Action:

Congratulations! As simple as that, you've now run your second simulation with this mesh, but at a new flow condition. Comparing the pressure fields, you should see small differences in the shock location. Below we show an image of the two results overlaid with transparency:

Both the shock on the suction-side (top surface) as well as the presence of a weak shock on the pressure-side (bottom surface) are different between the two cases.

What the Physics/Outputs/Solver - collectively the Setup environment - tabs do are define the Simulation Settings: configuration information that is sent to the flow solver. Once a simulation has started, those settings are immutable (can not be modified), and are associated metadata of the simulation result.

When you go back to any of the Setup tabs, you are changing the Simulation Settings that will be used for the next simulation; these have no impact on any of the previously computed results.

You may want to save the settings state you currently have in the Setup environment. You can do this by clicking the "..." button on the upper right of the Checklist panel and clicking Save settings to Library:

You can add a name and metadata to this file when you save it, and it will subsequently be available to Import in any Project. Note that importing a settings file from the Library will over-write any existing settings inputs. You can also save settings from the Results tab of any completed simulation using the same mechanism.

While this approach is good when the number of parameter changes are few, this manual process doesn't scale well to running many different simulation settings across different parameters. To efficiently run and sample a larger design space, you can leverage the Design of Experiments (DOE) interface.

This interface is exposed once a simulation has been completed, and is based off of the Settings for that particular simulation result. For any simulation result tab you should see this button available:

This will open a new interface that is designed to help you quickly define how your parameter design space will be sampled. This is controlled primarily through the DOE panel on the right:

Here you can specify how each dimension of your parameter space is sampled (the dimensions are defined in the next step: Inputs). The options are:

For the purpose of this tutorial we will stick to the Full Sweep option.

Action: none - default "Full Sweep" is used

This is where you define (a) what variables to use in the design space, and (b) how each dimension is sampled. The interface provides a curated list of variables that can be accessed within the Design of Experiments. Let's setup an angle of attack sweep at more granular intervals.

Action:

You will now see where the sampling for this dimension (angle of attack) is defined: the Type dropdown. The options available are:

If you have a large number of non-uniform samples, we suggest using the Manual Exploration Policy and upload a CSV that describes the samples.

Action:

As you adjust the sampling strategy and number of dimensions, you will see a badge on the Run Simulation button that indicates the total number of new simulations that will be run. If the button is purple, then you are ready to run your ensemble of cases!

The Luminary platform will allocate resources and run all of these jobs, executing as many in parallel as possible, without you needing to define anything related to the computational resources.

Each simulation run within a DOE is treated equal to any other Simulation - they are listed in the Results->All Results table and can be opened in their own tab as well. The only difference that you'll observe is that these simulations are aggregated in the All Results table into a container called Design of Experiments, which you can rename.

Additionally, this table will be augmented with the value of the DOE parameters (see column 2), so you can quickly glance and see how each row is different.