Welcome to www.adaptics.com

Contents of www.adaptics.com homepage.


IMAGINE the POSSIBILITIES

DATA IN -- MODEL OUT


ADAPTx

Automated Multivariable System Identification

and Time Series Analysis Software

Introducing ADAPTx, the next generation software for modeling complex dynamical systems from measured data. ADAPTx automatically determines optimal statistical models in a robust and stable computation. The accuracy of the fitted model is described by computed accuracy bounds. This complete procedure is entirely automated, or you may wish to use the interactive menus to investigate the effect of making choices other than those considered statistically optimal by ADAPTx.

Completely Automatic. Supply the measured data and ADAPTx identifies a state space model.

Imagine the Possibilities in:

MATLAB® Version. Runs under the industry standard productivity tool for signal processing, control design, simulation and analysis.

Runs Under C++. For embedded applications in self-tuning control and signal processing using a C++ matrix library in Windows or UNIX.

Simulation and Real Data. ADAPTx has been demonstrated extensively using complex high-fidelity simulations as well as real systems data for structural vibration, aircraft wing flutter, stirred tank reactor, distillation column, autothermal reactor, power plant boiler, automotive steering.

General Model Class. These systems involve multiple inputs and outputs, high state order, stiff dynamics, unstable dynamics, unknown feedback, colored state and measurement noise, bias and trends in the measurements. ADAPTx applies to general linear stochastic systems with no assumed prior structural form.

Stable Computation. ADAPTx computation involves primarily the singular value decomposition which is always stable and accurate. The state space model description is always well conditioned. No use is made of iterative parameter optimization that may not converge. Required computation is predetermined by the problem size.

Optimal Model Fitting. The model fitting is close to the optimum achievable accuracy. The model state order is chosen to minimize the statistically optimal Akaike information criterion (AIC) . This optimality is approximated even for small samples, and no initialization or prior information is used.

Identified Model Accuracy. Accuracy is given by confidence bands on the transfer function and power spectrum estimates, and on maximum singular value quantities. These can be used directly in robust control design.

Example: Shake Table Data. 1-input, 2-outputs, 600 sample points, optimum state order 29, 152 estimated parameters (a small sample case). Accuracy comparable to a 32,000 sample FFT averaging 512 point batches with hanning window and 50 percent overlap.

Contents


Adaptics, Inc

1717 Briar Ridge Road
McLean, Virginia 22101, USA
sales@adaptics.com
Phone: 703 532-0062; Fax: 703 536-3319
Copyright 1997 Adaptics, Inc.