Tutorial: Design and Modeling Scripts

On this page…
Processes You Can Automate Scripting Demos

Processes You Can Automate

The following description illustrates an example engine modeling process you can automate with the command-line Model-Based Calibration Toolbox product. You can assemble the commands for these steps into an easy-to-use script or graphical interface. This is a guideline for some of the steps you can use to model engine data.

1. Create or load a project — CreateProject; Load

2. Create a new test plan for the project using a template set up in the Model Browser — CreateTestplan

3. Create designs that define data points to collect on the test bed — CreateDesign. Work with classical, space-filling or optimal designs: CreateConstraint; CreateCandidateSet; Generate; FixPoints; Augment.

4. Create or load a data object for the project and make it editable — CreateData; BeginEdit

5. Load data from a file or the workspace — ImportFromFile; ImportFromMBCDataStructure

   You can instead specify the required data file when you call CreateData; you must still call BeginEdit before you can then make changes to the data.

6. Work with the data:

   • Examine data values — Value

   • Modify the data to remove unwanted records — AddFilter; AddTestFilter

   • Add user-defined variables to the data — AddVariable

   • Add new data — Append

   • Group your data for hierarchical modeling by applying rules — DefineTestGroups; DefineNumberOfRecordsPerTest

   • Export your data to the workspace — ExportToMBCDataStructure

7. Save your changes to the data, or discard them — CommitEdit; RollbackEdit

8. Designate which project data object to use for modeling in your test plan — AttachData

9. Create and evaluate boundary models, either in a project or standalone. You can use boundary models as design constraints. See Boundary Model Scripting.

10. Create models for the data; these can be one- or two-stage models and can include datum models — CreateResponse

11. Work with your models:

    • Examine input data and response data — DoubleInputData; DoubleResponseData

    • Examine predicted values at specified inputs — PredictedValue; PredictedValueForTest

    • Examine Predicted Error Variance (PEV) at specified inputs — PEV; PEVForTest

    • Examine and remove outliers — OutlierIndices; OutlierIndicesForTest; RemoveOutliers; ; RestoreData

    • Create a selection of alternative models — CreateAlternativeModels

    • Choose the best model by using the diagnostic statistics —- AlternativeModelStatistics; DiagnosticStatistics; SummaryStatistics

    • Extract a model object from any response object (Model Object), then:

      • Fit to new data ()

      • Create a copy of the model, change model type, properties and fit algorithm settings (CreateModel; ModelSetup; Type (for models); Properties (for models); CreateAlgorithm)

      • Include and exclude terms to improve the model (StepwiseRegression)

      • Examine regression matrices and coefficient values (Jacobian; ParameterStatistics)

      • If you change the model you need to use UpdateResponse to replace the new model back into the response in the project.

• For two-stage test plans, once you are satisfied with the fit of the local and response feature models (and have selected best models from any alternatives you created), you can calculate the two-stage model — MakeHierarchicalResponse.

• Now you can also examine the predicted values and PEV of the two-stage model.

• You can export any of these models to MATLAB or Simulink software — Export

This overview is not an exhaustive list of the commands available. For that, see Function Reference and Commands — Alphabetical List in the Model-Based Calibration Toolbox Reference.

Back to Top

Scripting Demos

See Model-Based Calibration Toolbox Demos for a selection of command-line demos. Run the demos to learn about:

• Loading and Modifying Data

• Designing experiments and constraining designs

• Gasoline engine modeling scipt to automatically generate a project for the gasoline case study, including:

  • Grouping and filtering data

  • Boundary modeling

  • Response modeling

  • Removing outliers and copying outlier selections

  • Creating alternative models and selecting the best based on statistical results

• Point-by-point diesel engine modeling to automatically generate a project for the diesel case study, including:

  • Defining engine operating points

  • Creating designs for each operating point

  • Augmenting designs to collect more data

  • Building a point-by-point boundary model

  • Create response models using the local multiple model type

Back to Top

Introduction to the Command-Line Interface

Understanding Model Structure for Scripting

© 1984-2012- The MathWorks, Inc.    -   Site Help   -   Patents   -   Trademarks   -   Privacy Policy   -   Preventing Piracy   -   RSS