System Identification Toolbox

An interactive environment for building accurate, simplified models of complex systems from noisy time-series data

The System Identification Toolbox provides tools for creating mathematical models of dynamical systems based on observed input/output data. The Toolbox features a flexible graphical user interface that aids in the organization of data and models. The identification techniques provided with this Toolbox are useful for applications ranging from control system design and signal processing to time-series analysis and vibration analysis.

Features

Easy-to-use GUI
Preprocessing of data, including bias removal, data detrending, and prefiltering
Data range selection
Model estimation: ARX, ARMAX, Box-Jenkins, output-error, and subspace methods
Time and frequency response plots
Pole-zero plots
Simulations and residuals for model validation

Highlights

Convenient GUI. The GUI simplifies the preprocessing of data as well as the iterative process of estimating models and evaluating their goodness-of-fit. Operations such as loading/saving data, selecting data range, bias removal, and detrending are quickly performed with minimal effort from pull-down menus made available in the GUI.

Graphical Organization. Data sets and identified models are organized graphically so that previous analyses can be recalled easily during the iterative system identification process. For the novice, the graphical organization provided by the GUI enables you to view the next available steps in the process. For the identification expert, the main user interface organizes the data to display what has been done already. This facilitates a rapid comparison of model estimates and provides a graphical means of retrieving prior models and inspecting their performance (model output, frequency response, etc.).

Parametric Models. Starting with measurements of a system's input and output data, a parametric model can be determined to mathematically describe the dynamic behavior of a system. The System Identification Toolbox supports virtually all standard model structures including AR, ARX, ARMAX, output-error, Box-Jenkins, ARARX, ARMA, ARARMAX, etc. The Toolbox supports general linear state-space models that can be defined in discrete-time and continuous-time. These models can include an arbitrary number of inputs and outputs.

Simulation and Validation. Functions are provided that allow identified models to be simulated using test data as inputs.

Broad Range of Applications. Linear model identification from test data is commonly used in the design of control systems where a plant model is required. In signal processing applications, derived models can be used for model-based and adaptive signal processing. System identification techniques have also been used successfully in financial applications.

Stefan Steinhaus, webmaster@steinhaus-net.de