Model Training

There are two ways to start the model training, either via Model > Configurations (Model Configurations window) or via Model > TrainModels (Model Configuration All Outputs window).

Starting the training via Model > Configurations allows you to set the model properties individually for each output. They concern the following topics:

  • modeling method (for this tutorial: NARX Structure)
  • method parameters

    For the NARX Structure method, these parameters are:

Setting individual model properties and starting the training

  • Select Model > Configurations.

    The Model Configurations window opens. It contains a separate tab for each output.

    Fig. 36: Model Configurations window

  • In the Modeling Method drop-down list, select NARX Structure.
  • In the Model area, do the following:

    • In the drop-down list, select a model type. See also Modeling Method and Algorithm.

      The <output> - Parameters window opens. Its content depends on the selected model type.

    • In the <output> - Parameters window, set the model parameters.

      See NARX Model Types for a description of the window.

    • Click OK to close the <output> - Parameters window.

      You can also open the <output> - Parameters window with the Edit button.

  • In the Feedback Structure area, enter the minimum and maximum time lags for inputs and outputs.

    See also Feedback Structure (NARX).

  • In the Dimensionality Reduction area, activate None.

    None is sufficient for the diesel example. For further information, see Dimensionality Reduction

  • Once you have set model parameters for all outputs, click OK or Apply to start the training.

Setting global model configurations and starting the training

Starting the training via Model > Configuration for All allows you to set model parameters for all outputs.

  • Select Model > Configuration for All.

    The Model Configuration All Outputs window opens.

    Fig. 37: Model Configuration All Outputs window

    This window contains the same elements as the Model Configurations window (Model Configurations window), except that there is only one tab for all outputs.

  • Set the model properties as described in Setting individual model properties and starting the training.

    Note  

    The settings you make affect all outputs.

  • Click OK to start the training.