11. Estimating a Model for Comparison
You estimated the parameters of and validated an ARX model for the DC motor. You also can estimate and validate other models with the same input and validation signals to determine if a different model provides a more accurate estimation for the system.
Complete the following steps to estimate the parameters of an ARMAX model using the sample signals.
- Click the Add Step button to open the Add Step menu.
- Select System Identification»Model Estimation»Parametric Estimation to add a Parametric Estimation step to the Project View. You want to add a new Parametric Estimation step so you can keep ARX Model to compare with the ARMAX model.
- On the Input Signals and Model Type tab in the Configuration View, select Normalized Stimulus from the Stimulus signal pull-down menu to specify Normalized Stimulus as the input stimulus signal for this Parametric Estimation step.
Note If the System Identification Assistant displays an error message stating The signal length is too small for computation, or the stimulus and response signals are not of the same length., you can ignore this error. This error message displays because the Stimulus signal and Response signal, which you specify in the next step, have different lengths.
- Select Normalized Response from the Response signal pull-down menu to specify Normalized Response as the input response signal for this Parametric Estimation step.
- Select ARMAX from the Type pull-down menu.
- On the Model Order Settings tab, set both A Order and B order to 2 to set up a second-order model.
- Verify that C order is set to 1. An ARMAX model incorporates the effect of previous prediction errors in the current estimation. The C coefficient adjusts the effect of the previous prediction signals. Setting C order to 1 configures the simplest case.
- Set delay to 1. Now the ARMAX model has the same input signals, orders, and delay as the ARX model.
- In the Project View, rename the SysID Model output to ARMAX Model.
- Select File»Save Project to save the project.
|Previous: 10. Validating the ARX Model||Next: 12. Validating the New Model|