| System Identification Toolbox | |
Detrending
Detrending the data involves removing the mean values or linear trends from the signals (the means and the linear trends are then computed and removed from each signal individually). This function is accessed under the pop-up menu Preprocess, by selecting item Remove Means or Remove Trends. More advanced detrending, such as removing piecewise linear trends or seasonal variations cannot be accessed within the GUI. It is generally recommended that you remove at least the mean values of the data before the estimation phase. There are however situations when it is not advisable to remove the sample means. It could for example be that the physical levels are built into the underlying model, or that integrations in the system must be handled with the right level of the input being integrated.
Selecting Data Ranges
It is often the case that the whole data record is not suitable for identification, due to various undesired features (missing or "bad" data, outbursts of disturbances, level changes etc.), so that only portions of the data can be used. In any case, it is advisable to select one portion of the measured data for estimation purposes and another portion for validation purposes. The pop-up menu item Preprocess > Select Range... opens a dialog box, which facilitates the selection of different data portions, by typing in the ranges, or marking them by drawing rectangles with the mouse button down.
For multivariable data it is often advantageous to start by working with just some of the input and output signals. The menu item Preprocess > Select Channels... allows you to select subsets of the inputs and outputs. This is done in such a way that the input/output numbering and names remains consistent when you evaluate data and model properties, for models covering different subsets of the data.
Prefiltering
By filtering the input and output signals through a linear filter (the same filter for all signals) you can, e.g., remove drift and high frequency disturbances in the data, that should not affect the model estimation. This is done by selecting the pop-up menu item Preprocess > Filter... in the main window. The dialog is quite analogous to that of selecting data ranges in the time domain. You mark with a rectangle in the spectral plots the intended passband or stop band of the filter, you select a button to check if the filtering has the desired effect, and then you insert the filtered data into the GUI's Data Board.
Prefiltering is a good way of removing high frequency noise in the data, and also a good alternative to detrending (by cutting out low frequencies from the pass band). Depending on the intended model use, you can also make sure that the model concentrates on the important frequency ranges. For a model that will be used for control design, for example, the frequency band around the intended closed-loop bandwidth is of special importance.
If you intend to use the data to build models both of the system dynamics and the disturbance properties, it is recommended to do the filtering at the estimation phase. That is achieved by selection the pop-up menu item Estimate > Parametric Models, and then select the estimation Focus to be Filter. This opens the same filter dialog as above. The prefiltering will however apply only for estimating the dynamics from input to output. The disturbance model is determined from the original data.
Resampling
If the data turn out to be sampled too fast, they can be decimated, i.e., every
k-th value is picked, after proper prefiltering (antialias filtering). This is obtained from menu item Preprocess > Resample.
You can also resample at a faster sampling rate by interpolation, using the same command, and giving a resampling factor less than one.
Quickstart
The pop-up menu item Preprocess > Quickstart performs the following sequence of actions: It opens the Time plot Data view, removes the means from the signals, and it splits these detrended data into two halves. The first one is made Working Data and the second one becomes Validation Data. All the three created data sets are inserted into the Data Board.
Multi-Experiment Data
The toolbox allows the handling of data sets that contain several different experiments. Both estimation and validation can be applied to such data sets. This is quite useful to deal with experiments that have been conducted at different occasions but describe the same system. It is also useful to be able to keep together pieces of data that have been obtained by cutting out "informative pieces" from a long data set. Multi-experiment data can be imported and used in the GUI as any iddata object. Selecting specific part of a multi-experiment data set is done from the pop-up menu item Preprocess > Select Experiment. To merge several data sets in the Data board (obtained, e.g., by cutting out nice portions from other data sets) use the pop-up menu item Preprocess > Merge Experiment.
| | Taking a Look at the Data | Checklist for Data Handling | |