A new invertible model of magnetorheological damper based on sigmoid function

Abstract

As an intelligent device, the magnetorheological (MR) damper has received extensive attention in many fields. However, in the real-time control system, it is still extremely necessary to develop a simple, accurate and invertible dynamic model of the MR damper. In this work, a new invertible MR dynamic model based on sigmoid function is proposed. According to the experimental data of a common MR damper, it is assumed that the damping force at zero velocity under constant current is only related to current and independent of the excitation frequency. The preliminary sigmoid model based on this hypothesis can better adapt to different frequency excitations while reducing the complexity of the model. In addition, according to the parameter analysis of the proposed preliminary model, the relationship between the parameters and the input current is reasonably simplified, so that the final model can meet reversibility requirements in the control system while ensuring accuracy. In order to evaluate the performance of the proposed model, it is compared with the existing sigmoid models from the literature. The results show that the proposed model has certain advantages in the number of parameters, accuracy and reversibility, which is suitable for the real-time control system.

Realization of desired damping characteristics based on an open-loop-controlled magnetorheological damper

“…The sigmoid model (Lu et al, 2020) is used to describe the dynamic behavior of the MR damper, as shown in Equations (2a) (2b).…”

“…However, common reversible models usually cannot accurately describe the dynamic behavior of MR damper at different frequencies, such as polynomial models (Dong et al, 2009, 2010; Xiao Min et al, 2009) and hyperbolic tangent models (Hu et al, 2017; Kecik et al, 2014), or the recognition processes of the models are complicated (Yu et al, 2017), such as Adaptive Neuro-Fuzzy Inference Inverse Model (ANFIS) (Nguyen et al, 2017; Zong et al, 2012, 2013). Recently, Zhang et al proposed a reversible sigmoid model (Lu et al, 2020), which has good fitting accuracy for damping forces at different frequencies and different currents, but the actual control effect of this model has not yet been experimentally verified.…”

“…To achieve the desired damping characteristics of the MR damper, this study adopted the sigmoid model (Lu et al, 2020) to obtain the desired current, and designed a tripolar current driver to improve the response of the current, and finally used the open-loop control approach to achieve the control target. The linear and nonlinear damping characteristics are uniformly expressed in the form of a piecewise power function, which is used as the control target.…”

Magnetorheological damper modeling based on a refined constitutive model for MR fluids

“…It is noted that the Bouc-Wen models are accurate in predicting the dynamic response of MR damper but they are usually highly computationally demanding due to the involvement of many parameters and differential equations (Spencer et al, 1997; Yang et al, 2004). Moreover, the existence of differential equations makes it harder to obtain the inverse models of Bouc-Wen models which are necessary for the current identification of MR dampers in real-time control systems (Cheng et al, 2020; Lu et al, 2020). The Bingham plastic model, which is originally proposed for characterizing the materials that behave as a rigid body at low shear rates and as a viscous fluid at high shear rates, is not only computationally efficient than the Bouc-Wen models but also easy to be inverted to achieve effective and efficient real-time control.…”

Shaking Table Attached to Magnetorheological Damper: Simulation and Experiments for Structural Engineering

“…As for static models, several researchers have suggested either nonlinear functions or soft-computing techniques. For instance, researchers have utilized the static sigmoid function [ 20 , 24 ], Gompertz function [ 25 ], hyperbolic tangent function [ 26 ], and neural networks [ 27 , 28 , 29 ]. These investigations came with a drawback—they require a large database to tune in the corresponding functions.…”