6. Normalizing Validation Data

In the Splitting Data exercise, you created two sets of stimulus and response data. You can use one set, consisting of Estimation Stimulus and Estimation Response, to estimate the model for the DC motor. You can use the other set, consisting of Validation Stimulus and Validation Response, to validate the model you estimate. If you normalize the data set you use for model estimation, normalize the data set you use for model validation as well.

Complete the following steps to normalize the data you want to use for model validation.

1. Press the <Ctrl> key and click and drag the Normalize Data step to the empty white area of the Project View to create a copy of the Normalize Data step.
2. In the Project View, click the blue down arrow on the Normalized Stimulus input of the new Normalize Data step and select Validation Stimulus to specify Validation Stimulus as the stimulus signal you want to normalize.
   
   Notice that when you click the blue down arrow, you can choose from a list of ten available inputs. These available inputs are the output signals from the steps above this Normalize Data step in the Project View.
3. Click the blue down arrow on the Normalized Response input and select Validation Response to specify Validation Response as the response signal you want to normalize.
4. On the Mean Source Configuration tab in the Configuration View, select the Value from another block option in the Stimulus mean source section and verify that Stimulus Mean is selected in the Value from another block pull-down menu. Selecting this option to normalize the stimulus signal ensures that this step uses the same mean for both the estimation data and validation data.
5. Select the Value from another block option in the Response mean source section and verify that Response Mean is selected in the Value from another block pull-down menu. Selecting this option to normalize the response signal ensures that this step uses the same mean for both the estimation data and validation data.
6. On the Standard Deviation Source Configuration tab, select the Value from another block option in the Stimulus standard deviation source section and verify that Stimulus Standard Deviation is selected in the Value from another block pull-down menu. Selecting the Stimulus Standard Deviation and Response Standard Deviation to normalize the stimulus and response signals, respectively, ensures that this step uses the same standard deviation for both the estimation data and validation data.
7. Select the Value from another block option in the Response standard deviation source section and verify that Response Standard Deviation is selected in the Value from another block pull-down menu.

   Note  You want the System Identification Assistant to use the means and standard deviations of the estimation data to normalize the validation data so that the two sets of data have the same range. Because the System Identification Assistant does not use the actual mean and standard deviation of the validation data to normalize the signals, the normalized signals corresponding to Validation Stimulus and Validation Response may not have a mean of exactly 0 or a standard deviation of exactly 1.

8. In the Project View, rename the Normalized Stimulus 1 output to Test Stimulus. You can use Test Stimulus as the stimulus signal to validate the model you estimate.
9. Rename the Normalized Response 1 output to Test Response. You can use Test Response as the response signal to validate the model you estimate.
10. Select File»Save Project to save the project.