An experimental artificial-neural-network-based modeling of magneto-rheological fluid dampers

J. C. Tudón-Martínez, J. J. Lozoya-Santos, R. Morales-Menendez, R. A. Ramirez-Mendoza

Research output: Contribution to journalArticle

13 Citations (Scopus)

Abstract

A static model for a magneto-rheological (MR) damper based on artificial neural networks (ANNs) is proposed, and an intensive and experimental study is presented for designing the ANN structure. The ANN model does not require time delays in the input vector. Besides the electric current signal, only one additional sensor is used to achieve a reliable MR damper structure. The model is experimentally validated with two commercial MR dampers of different characteristics: MR 1 damper with continuous actuation and MR 2 damper with two levels of actuation. The error to signal ratio (ESR) index is used to measure the model accuracy; for both MR dampers, an average value of 6.03% of total error is obtained from different experiments, which are designed to explore the nonlinearities of the MR phenomenon at different frequencies by including the impact of the electric current fluctuations. The proposed ANN model is compared with other well known parametric models; the qualitative and quantitative comparison among the models highlights the advantages of the ANN for representing a commercial MR damper. The ESR index was reduced by the ANN-based model by up to 29% with respect to the parametric models for the MR 1 damper and up to 40% for the MR 2 damper. The force-velocity diagram is used to compare the modeling properties of each approach: (1) the Bingham model cannot describe the hysteresis of both MR dampers and the distribution function of the modeled force varies from the experimental data, (2) the algebraic models have complications in representing the nonlinear behavior of the asymmetric damper (MR 2 ) and, (3) the ANN-based MR damper can model the nonlinearities of both MR dampers and presents good scalability; the accuracy of the results supports the use of this model for the validation of semi-active suspension control systems for a vehicle, by using nonlinear simulations. © 2012 IOP Publishing Ltd.
Original languageEnglish
JournalSmart Materials and Structures
DOIs
Publication statusPublished - 1 Aug 2012
Externally publishedYes

Fingerprint

dampers
Neural networks
Fluids
fluids
indexes (ratios)
Electric currents
electric current
actuation
nonlinearity
static models
Distribution functions
Hysteresis
Scalability
Time delay
Suspensions
vehicles
time lag
distribution functions
hysteresis
diagrams

Cite this

@article{2baba0d7230640f89064b5d1abcd73c2,
title = "An experimental artificial-neural-network-based modeling of magneto-rheological fluid dampers",
abstract = "A static model for a magneto-rheological (MR) damper based on artificial neural networks (ANNs) is proposed, and an intensive and experimental study is presented for designing the ANN structure. The ANN model does not require time delays in the input vector. Besides the electric current signal, only one additional sensor is used to achieve a reliable MR damper structure. The model is experimentally validated with two commercial MR dampers of different characteristics: MR 1 damper with continuous actuation and MR 2 damper with two levels of actuation. The error to signal ratio (ESR) index is used to measure the model accuracy; for both MR dampers, an average value of 6.03{\%} of total error is obtained from different experiments, which are designed to explore the nonlinearities of the MR phenomenon at different frequencies by including the impact of the electric current fluctuations. The proposed ANN model is compared with other well known parametric models; the qualitative and quantitative comparison among the models highlights the advantages of the ANN for representing a commercial MR damper. The ESR index was reduced by the ANN-based model by up to 29{\%} with respect to the parametric models for the MR 1 damper and up to 40{\%} for the MR 2 damper. The force-velocity diagram is used to compare the modeling properties of each approach: (1) the Bingham model cannot describe the hysteresis of both MR dampers and the distribution function of the modeled force varies from the experimental data, (2) the algebraic models have complications in representing the nonlinear behavior of the asymmetric damper (MR 2 ) and, (3) the ANN-based MR damper can model the nonlinearities of both MR dampers and presents good scalability; the accuracy of the results supports the use of this model for the validation of semi-active suspension control systems for a vehicle, by using nonlinear simulations. {\circledC} 2012 IOP Publishing Ltd.",
author = "Tud{\'o}n-Mart{\'i}nez, {J. C.} and Lozoya-Santos, {J. J.} and R. Morales-Menendez and Ramirez-Mendoza, {R. A.}",
year = "2012",
month = "8",
day = "1",
doi = "10.1088/0964-1726/21/8/085007",
language = "English",
journal = "Smart Materials and Structures",
issn = "0964-1726",
publisher = "IOP Publishing Ltd.",

}

An experimental artificial-neural-network-based modeling of magneto-rheological fluid dampers. / Tudón-Martínez, J. C.; Lozoya-Santos, J. J.; Morales-Menendez, R.; Ramirez-Mendoza, R. A.

In: Smart Materials and Structures, 01.08.2012.

Research output: Contribution to journalArticle

TY - JOUR

T1 - An experimental artificial-neural-network-based modeling of magneto-rheological fluid dampers

AU - Tudón-Martínez, J. C.

AU - Lozoya-Santos, J. J.

AU - Morales-Menendez, R.

AU - Ramirez-Mendoza, R. A.

PY - 2012/8/1

Y1 - 2012/8/1

N2 - A static model for a magneto-rheological (MR) damper based on artificial neural networks (ANNs) is proposed, and an intensive and experimental study is presented for designing the ANN structure. The ANN model does not require time delays in the input vector. Besides the electric current signal, only one additional sensor is used to achieve a reliable MR damper structure. The model is experimentally validated with two commercial MR dampers of different characteristics: MR 1 damper with continuous actuation and MR 2 damper with two levels of actuation. The error to signal ratio (ESR) index is used to measure the model accuracy; for both MR dampers, an average value of 6.03% of total error is obtained from different experiments, which are designed to explore the nonlinearities of the MR phenomenon at different frequencies by including the impact of the electric current fluctuations. The proposed ANN model is compared with other well known parametric models; the qualitative and quantitative comparison among the models highlights the advantages of the ANN for representing a commercial MR damper. The ESR index was reduced by the ANN-based model by up to 29% with respect to the parametric models for the MR 1 damper and up to 40% for the MR 2 damper. The force-velocity diagram is used to compare the modeling properties of each approach: (1) the Bingham model cannot describe the hysteresis of both MR dampers and the distribution function of the modeled force varies from the experimental data, (2) the algebraic models have complications in representing the nonlinear behavior of the asymmetric damper (MR 2 ) and, (3) the ANN-based MR damper can model the nonlinearities of both MR dampers and presents good scalability; the accuracy of the results supports the use of this model for the validation of semi-active suspension control systems for a vehicle, by using nonlinear simulations. © 2012 IOP Publishing Ltd.

AB - A static model for a magneto-rheological (MR) damper based on artificial neural networks (ANNs) is proposed, and an intensive and experimental study is presented for designing the ANN structure. The ANN model does not require time delays in the input vector. Besides the electric current signal, only one additional sensor is used to achieve a reliable MR damper structure. The model is experimentally validated with two commercial MR dampers of different characteristics: MR 1 damper with continuous actuation and MR 2 damper with two levels of actuation. The error to signal ratio (ESR) index is used to measure the model accuracy; for both MR dampers, an average value of 6.03% of total error is obtained from different experiments, which are designed to explore the nonlinearities of the MR phenomenon at different frequencies by including the impact of the electric current fluctuations. The proposed ANN model is compared with other well known parametric models; the qualitative and quantitative comparison among the models highlights the advantages of the ANN for representing a commercial MR damper. The ESR index was reduced by the ANN-based model by up to 29% with respect to the parametric models for the MR 1 damper and up to 40% for the MR 2 damper. The force-velocity diagram is used to compare the modeling properties of each approach: (1) the Bingham model cannot describe the hysteresis of both MR dampers and the distribution function of the modeled force varies from the experimental data, (2) the algebraic models have complications in representing the nonlinear behavior of the asymmetric damper (MR 2 ) and, (3) the ANN-based MR damper can model the nonlinearities of both MR dampers and presents good scalability; the accuracy of the results supports the use of this model for the validation of semi-active suspension control systems for a vehicle, by using nonlinear simulations. © 2012 IOP Publishing Ltd.

U2 - 10.1088/0964-1726/21/8/085007

DO - 10.1088/0964-1726/21/8/085007

M3 - Article

JO - Smart Materials and Structures

JF - Smart Materials and Structures

SN - 0964-1726

ER -