next up previous
Next: Outlook Up: The Graphical User Interface Previous: Aims of the New

Description of the GUI

The key to simple use is simple handling, which can be supported by the graphical capabilities of MATLAB. The core of the GUI is a flow chart, on which the processing of the data can be followed (Fig. [*]). It can be started from the command line by issuing fdtool .


  
\begin{figure}
\vspace{1cm}
\end{figure}
Figure: Starting graph of the Graphical User Interface of the Frequency Domain System Identification Toolbox (fdident)

At start the helps  offer assistance. General help is offered under the menu items ``Getting Started$\ldots$ '', ``Fdident GUI'', and ``This window'' (Fig. [*]).


  
\begin{figure}
\vspace{1cm}
\end{figure}
Figure: The help menu

Special help is available under ``Help on Object'' . When selecting this, the cursor changes to a question mark. When an object is selected by this mark, a help window opens up with an explanation of the object.

The task of the block diagram is to illustrate the whole identification procedure, and provide control of it. The main steps are: read data (from file, or as a result of measurement or simulation), data preprocessing, identification, and finally validation, or maybe comparison of several models.

An important menu item of the basic diagram is ``UserLevel'' . When this is set to ``Basic'', the number of questions to the user is minimized: the program decides whatever it can. More details are shown and can be influenced in the setting ``Intermediate''. In the following part we assume the setting ``Intermediate''.

Let us follow now steps of an identification procedure. Let us open first the frequency domain data acquisition window (Read / Measure Time Domain Data). Figure [*] shows the situation after reading in a data object.


  
\begin{figure}
\vspace{1cm}
\end{figure}
Figure: The Read / Measure Time Domain Data window

The data are periodic, so the Segmentation window offers segmentation  (Fig. [*]).


  
\begin{figure}
\vspace{1cm}
\end{figure}
Figure: The segmentation window

Segmentation is illustrated in a separate window (Fig. [*]).


  
\begin{figure}
\vspace{1cm}
\end{figure}
Figure: Illustration of segmentation

After segmentation, first conversion to the frequency domain, then a selection possibility among the frequency lines follow. These are not illustrated here.

Now the Variance Analysis  block can be opened. Figure [*] illustrates the situation after variance analysis.


  
\begin{figure}
\vspace{1cm}
\end{figure}
Figure: Variance Analysis window


  
\begin{figure}
\vspace{1cm}
\end{figure}
Figure: Computer Aided Model Selection window

When data preprocessing is finished, parameter estimation   follows. The straightforward step is to open the Estimate Plant Model block. However, when experimentation with different model orders is desirable, Computer Aided Model Scan is the right tool (Fig. [*]). In this block we can select the desired combinations of the numerator/denominator orders, and run the estimation function for each of them.

Here we would like to point at two modifier checkboxes . The transfer function is often badly conditioned , both from estimation and from evaluation points of view, especially when the numerator/denominator orders of the system transfer function are above 30-40. This means that roundoff errors may deteriorate the results. A remedy is the use of orthogonal (Forsythe) polynomials   [9]. This is switched on by checking the checkbox ``Improved numerical stability''. The user does not notice any difference in the interface, but inside a more involved code runs. This is typical for the GUI: a modification which seems to be simple for the user may significantly change the execution of the program.

The situation is similar for the checkbox ``Transient elimination''. This makes the identification algorithm deal also with a transient component beside the periodic one, while the user does not feel much difference in the interface [5, 6].

During the run of the iterative minimization, information about the run is shown (ELiS Run Info, Fig. [*]).


  
\begin{figure}
\vspace{1cm}
\end{figure}
Figure: Information window of elis

When every desired identification result is at hand, the Evaluate and Compare Plant Models block can be opened (Fig. [*]). The figure illustrates the situation after copying Model Set #1 into Set #2.


  
\begin{figure}
\vspace{1cm}
\end{figure}
Figure: Compare and Evaluate Models window

If we are interested in details of the selected plot, we can zoom into it by double-click.

Further steps of model validation will not be illustrated here. Rather, we go back to the first block we skipped at the beginning, the Excitation Signal Design  block (Fig. [*]).


  
\begin{figure}
\vspace{1cm}
\end{figure}
Figure: Excitation Signal Design, with tooltipstring under the radiobutton ``Linear''

Here we may observe another help facility, present in all windows: when the cursor is moved above a ``uicontrol'' , after a short time a one-line help, the so-called tooltipstring  (balloon help)  appears. This gives the quickest information about graphics objects.

The Excitation Signal Design block assists in designing the desired input signals. Since here several parameters are in interrelation, there is an Adjust button to assure consistency of all parameters.

If a multisine  is designed with components at 100Hz, 300Hz, $\ldots$ 2300Hz, with equal amplitudes, the corresponding windows are as given in Figs. [*]-[*].


  
\begin{figure}
\vspace{1cm}
\end{figure}
Figure: Amplitude spectrum


  
\begin{figure}
\vspace{1cm}
\end{figure}
Figure: Information window of crest factor minimization

A binary signal with similar properties can also be designed by selecting DIBS (Fig. [*]).


  
\begin{figure}
\vspace{1cm}
\end{figure}

Coming back now to the already rather colorful basic chart (Fig. [*]), every active element can be selected, and this starts the corresponding actions. Blocks open up to windows, and at arrows an action selection window appears (Fig. [*]). Thus it is easy to export/import data, perform graphical inspection or display information about them.


  
Figure: Status of the main window after identification
\begin{figure}
\vspace{1cm}
\end{figure}
Figure: Illustration of DIBS design


  
\begin{figure}
\vspace{1cm}
\end{figure}
Figure: Arrow actions window


next up previous
Next: Outlook Up: The Graphical User Interface Previous: Aims of the New
István Kollár
1999-07-06