Skip to content

Create a mapping

A Mapping fits a surrogate model to the results of a set of Runs, so you can predict performance between the simulated Attitudes, plot it over two parameters, and search for optimum settings. For how the model works and when to trust it, see mappings and surrogate models.

Before you start

  • Simulated Attitudes that cover the space. You need at least two Runs whose results include force data. For an aero map, generate the Attitudes with a design of experiments such as Optimised Latin Hypercube. See generate a sweep.
  • Parameters that vary. Each parameter you map must have a value in every selected Run and must change across them.
  • A table view with the performances. The values you can map are the columns of a Results table view, such as drag, lift and balance. See create a table view.
  • Select a Programme and a Project in the header. Mappings belong to a Project.

Create the Mapping

  1. In the sidebar, under Visualisation & Data, click Mappings.
  2. Click New Mapping. Create Mapping opens.

  3. Fill in the form:

    Field Required Notes
    Name Yes Up to 200 characters. Must be unique in the Project
    Runs Yes Only Runs with results are listed ("Runs must have results to be used in a mapping")
    Parameters Yes The model's inputs. Map parameters, or Run parameters that are set in every selected Run and aren't the same in all of them
    Apply Smoothing No Off by default. Lets the model smooth out scatter in the results instead of passing exactly through every point
    Multi Component Mapping No Off by default. See multi-component mappings
    Force Settings Unless multi-component The table view whose columns are offered as performances. Defaults to the view loaded on the Results page
    Performances Unless multi-component The outputs to model: columns of the Force Settings view that aren't parameters

  4. Click Create. It's disabled until the form is complete.

Bramble trains one model per performance in the background. The Status column on the Mappings page updates as it goes: "Calculation in progress", then "Calculation successful" or "Calculation failed".

If the Parameters list is empty, its placeholder says why:

Message Meaning
"At least two valid cases are required" Fewer than two Runs in the Project have results with force data
"More than one case must be used in a mapping" Only one of the selected Runs has usable results
"No valid cases selected" None of the selected Runs has usable results
"No valid parameters for this selection" No parameter is set in every selected Run and varies across them

To rename a Mapping, click its pencil icon. To delete it, click its delete icon ("Cannot delete Mapping in progress" while it's still training).

Multi-component mappings

With Multi Component Mapping on, Bramble models each zone's force components separately and adds them up, instead of modelling the totals. Force Settings and Performances are disabled, and the performances are fixed: Drag, Lift, Front Lift, Rear Lift, Balance (as a percentage) and L/D. Lift is front plus rear lift, balance is front lift divided by lift, and L/D is lift divided by drag.

Multi-component mappings work with OpenFOAM results only.

Open a Mapping

Click a Mapping on the Mappings page. The header shows its name and a Multi-Component or Smoothed tag.

The table lists every point in the Mapping. Type is Case for the simulated Attitudes and Interpolation for predictions you add, followed by the parameters and performances. Case rows can't be removed.

Predictions aren't saved

Interpolation rows only exist on the page. They're lost when you reload the page or navigate away. To keep them, download the CSV or add them to a Map.

Predict

Button Does
Predict one point Predict Single Point: enter a value for every parameter, then click Predict and add to table
Predict Linear Sweep Predict - Linear Sweep: set From, To and Number of points for every parameter, then click Calculate
Optimise a Performance Finds the parameter values that minimise or maximise one performance within bounds

Linear sweep. Each parameter's column heading shows the range in the data, and From and To default to it. Number of points is a whole number of at least 2. All the parameters move together, from the From values to the To values: it's a line through parameter space, not a grid. For example, From 50, To 70 and Number of points 3 gives 50, 60 and 70. From must be less than To.

Optimise a performance.

  1. Click Optimise a Performance.
  2. Choose the Performance to optimise, and Maximise (the default) or Minimise.
  3. Set From and To for each parameter to bound the search.
  4. Click Confirm.

The search runs in the background; a spinner shows while it's running. When it finishes you get Optimisation Success: <performance> and the optimum is added to the table, named after the performance, the direction and the bounds. "Optimisation Failure" means no optimum was found.

Plot a contour map

Plot Interpolated Point plots one performance over two parameters as a heatmap. It needs a Mapping with at least two parameters ("2 or more Parameters are required").

  1. Click Plot Interpolated Point.
  2. Select Variables: choose the X-axis Variable, Y-axis Variable and Dependent Variable (the performance). Click Next.
  3. Set Bounds: for each axis, set From, To and Number of points (2 to 200, default 50). From and To default to the range in the data; set both or leave both empty. Click Next.
  4. Constants: if the Mapping has more than two parameters, give a fixed value for each of the others.
  5. Click Plot.

Turn on Show normalised uncertainties to plot the variance of the prediction instead of the prediction itself. It's low near the simulated Attitudes and grows away from them, so it shows where the Mapping needs more data.

Click a point on the plot to add that Attitude to the table ("Attitude Added to Mapping").

Explore trade-offs

Pareto front. Plots the best trade-off between two performances. It needs at least two performances and isn't available for multi-component mappings.

  1. Click Generate Pareto Data. Bramble works out the front for every pair of performances in the background.
  2. When it's ready, click Plot Pareto Data.
  3. Choose the X-axis Performance and Y-axis Performance, then click Plot.

Cloud. Plots 5,000 random predictions as a scatter of two performances.

  1. Click Generate Cloud Data, then Plot Cloud Data when it's ready. Multi-component mappings generate the cloud automatically, so they only have Plot Cloud Data.
  2. Choose the X-axis Performance and Y-axis Performance, then click Plot.
  3. Optionally set From and To for each parameter and click Filter plot. Points are coloured Within limits or Outside of limits.

Both searches go 25% beyond the parameter range of the simulated Attitudes. Pareto points outside that range are coloured Outside of original range: they're extrapolations, so treat them with care.

Click a point on either plot to add it to the table.

Run the predicted Attitudes

To check a prediction with CFD, add it to a Map and run it:

  1. In the table, click Add to Map on an Interpolation row.
  2. In Add interpolation to a Map, choose the Map in Select a Map, then click Add to Map.

A new Attitude with the row's parameter values is added to that Map. Only Maps you can edit are listed, and Restricted Users don't have the row actions.

A Mapping doesn't update when new Runs finish. To include them, create a new Mapping with the extra Runs.

Export

Click Download CSV file to download the table, simulated and predicted rows, as <mapping name>_mapping.csv.

For an AeroMap Excel workbook of a Mapping, contact support@bramblecfd.com.

Related: Mappings and surrogate models · Generate a sweep · Create a table view · Create a Map