EACS 2016 paper - Parameter Identification of Hysteresis Using Duffing-Like Model

This paper discusses a relatively new and standard system identification approach for hysteresis system using the Duffing-like model. As well-known in literature, many structural and mechanical components display the hysteresis behaviour of forward-backward loop on the input-output graph, such as magnetorheological damper and rubber isolator. The existing hysteresis models in literature usually involve non-ideal functions, such as coupled parameters and absolute values. Difficult-to-determine parameters sometimes cause the design, analysis, and compensation work inefficient and ineffective for pragmatic purpose. The proposed Duffing-like model is developed based on the Duffing equation, showing a scenario of continuous nonlinear ordinary differential equation. Therefore, the parameters are decoupled and the identification work can be carried out in a more systematic and deterministic manner. In addition, stability of the Duffing-like model can be proven using the Lyapunov and invariant set theorems, which offer a basis for parameter bounding and control design. Modelling and identification results of a magnetoresistive sensor are shown as an example to demonstrate the proposed techniques; the physical meaning of the parameter is also discussed.