Transfer Function Models in the System Identification App

Hello,

My question is whether the "Transfer Function Models in the System Identification App" can identify nonlinear system? Based on my knowledgae the Transfer Function is only defined for linear systems, but in my simulation the "Transfer Function Models in the System Identification App" could provide a good estimate (92% accuracy) of my nonlinear system. Does anyone know whether this "Transfer Function Models in the System Identification App" employs any algorithms capable of identifying nonlinear systems?

Answer 1

[Jon]

The transfer function models in the System Identification Toolbox are linear systems. If, for the range of inputs, and possibly initial conditions, that you are interested in, the behavior of the actual, non-linear system, closely resembles the simpler linear transfer function system, then the transfer function may provide a reasonable approximation. Without being too precise, for systems with smooth non-linearities, the behavior will always be approximated by a linear system for a small enough range of excitation.

Answer 2

[Rajiv Singh]

The name "Transfer function" is reserved for linear transfer functions (equivalently, linear ODEs) in System Identification Toolbox. The app does offer nonlinear identification tools - Nonlinear ARX and Hammerstein-Wiener Models. There is also grey box modeling capability available that applies to both linear and nonlinear systems (that is, estimating parameters of pre-specified linear or nonlinear ODEs to fit data).

See: Nonlinear Model Identification - MATLAB & Simulink (mathworks.com)