Modelling friction in a nonlinear dynamic system via Bayesian inference

Abstract

This work is concerned with the system identification of a real nonlinear system with Duffing-type and friction nonlinearities. With friction being a complex nonlinear phenomenon for which a variety of models have been developed, the identification problem investigated in this paper is one of model selection as well as parameter estimation. Consequently, through the comparison of experimental results with the output of various digital simulations the parameters of several different friction models (Coulomb, hyperbolic tangent and LuGre) are estimated using Bayesian inference in conjunction with Markov Chain Monte-Carlo methods. The performance of each model is then analysed using the Deviance Information Criterion which rewards the ability of the model to replicate the experimental behavior while penalising model complexity. The potential benefits of tackling model selection and parameter estimation problems using a Bayesian framework are discussed. © The Society for Experimental Mechanics, Inc. 2013.