HS-3015: Automated Fit from CSV Data

Learn how to create a Lookup model to link to tabulated data in an external .csv file, run a DOE of type Run Matrix to import the data in the lookup .csv file, and build a predictive model using FAST (Fit Automatically Selected by Training).

Before you begin, copy the model files used in this tutorial from <hst.zip>/HS-3015/ to your working directory.

Review CSV Data

Open the FAST_data.csv file and review its contents.
The .csv file contains two variables (x and y) and three responses.

Perform the Study Setup

  1. Start HyperStudy.
  2. Start a new study in the following ways:
    • From the menu bar, click File > New.
    • On the ribbon, click .
  3. In the Add Study dialog, enter a study name, select a location for the study, and click OK.
  4. Go to the Define Models step.
  5. Add a Lookup model by dragging-and-dropping the FAST_data.csv file from the Directory into the work area.

    Figure 1.
  6. Import variables.
    1. Click Import Variables.
      The Import Variables dialog opens.
    2. In the Number of design variables field, enter 2.
    3. Click OK.
    The input variables are expected in the first two columns, and the remaining columns are interpreted as output responses.

    Figure 2.
  7. Go to the Define Input Variables step.
  8. Review the input variables.
    The bounds of the input variables are based on the FAST_data.csv file’s contents. The nominal values are set to the first entry in the .csv file.

    Figure 3.

Perform Nominal Run

  1. Go to the Test Models step.
  2. Click Run Definition.
    An approaches/setup_1-def/ directory is created inside the study Directory. The approaches/setup_1-def/run__00001/m_1 directory contains the input file, which is the result of the nominal run.

Review Output Responses

  1. Go to the Define Output Responses step.
  2. Review the output responses.
    One output response is named Highly Nonlinear and two are polynomials.

    Figure 4.

Run a Run Matrix DOE

  1. Add a DOE.
    1. In the Explorer, right-click and select Add from the context menu.
    2. In the Add dialog, select DOE and click OK.
  2. Go to the DOE 1 > Specifications step.
  3. In the work area, set the Mode to Run Matrix.
  4. From the Settings tab, Matrix File field, navigate to your working directory and select the FAST_data.csv file.

    Figure 5.
  5. Click Apply.
    The DOE matrix populates with the input variable values from the FAST_data.csv file.
  6. Go to the DOE 1 > Evaluate step.
  7. Click Evaluate Tasks.

Run FAST Fit

  1. Add a Fit.
    1. In the Explorer, right-click and select Add from the context menu.
    2. In the Add dialog, select Fit and click OK.
  2. Import matrix.
    1. Go to the Fit 1 > Specifications step.
    2. Click Add Matrix.
    3. In the work area, set Matrix Source to Doe 1 (doe_1).
    4. Click Apply.
  3. Define specifications.
    1. Verify that the Fit Type assigned to each output response is FAST – Fit Automatically Selected by Training.

      Figure 6.
      By default, FAST automatically selects the best Fit type from all available Fits. You can manually select the Fit types FAST can choose by highlighting one or more responses in the work area and selecting Fits from the Settings tab.

      Figure 7.
    2. Click Apply.
  4. Evaluate tasks.
    1. Go to the Fit 1 > Evaluate step.
    2. Click Evaluate Tasks.
      Note: The choices for the best available Fit vary for each output response, which can cause these loops to be time consuming compared to when you select a single specific Fit. The steps for each output response are mutually exclusive, therefore you can use the Multi-Execution option to accelerate this process.
  5. Go to the Fit 1 > Post-Processing step.
  6. Review diagnostics.
    1. Click the Diagnostics tab.
      The Highly Nonlinear response uses RBF, while the other responses use LSR. In each case, FAST selected the specifics to have the highest validation R-square value. The R-Square can be interpreted as the % of the data’s variance that can be explained by the model.

      Figure 8.
    2. Click Regression Terms and compare Poly1 and Poly2 by selecting them individually in the work area.
      Poly1 and Poly2 are using stepwise regression, which means that the coefficients of the regression are reduced to a minimal set that sufficiently models the data. Poly1 uses only x, whereas Poly2 uses x and y^2.

      Figure 9. Poly1

      Figure 10. Poly2
    3. If required, copy the data from the Fit Type and Fit Specifics columns in the Diagnostics table and paste it into the Fit Type and Fit Specifics columns in the Specification step.
      This step explicitly sets the Fit specifications to the results determined from FAST; if the Fit must be re-run, this step can save time because FAST does not need to search for the best settings.
  7. Click the Trade-Off Tab to plot all the functions and see the predicted versus the known data points.
    In each case, the Fit model follows the data closely regardless of the sinusoidal functions in the Highly Nonlinear response to the simple planar data of the polynomial responses.

    Figure 11.