Vous êtes sur la page 1sur 2

RECONSTRUCTION OF IMPACT FORCE VIA BAYESIAN APPROACH

1 1* 2 S. Samagassi1 , A. Khamlichi , A. Driouech and E. Jacquelin

Laboratory Communication systems Abdelmalek Essadi University, Tetouan, MOROCCO saora82@yahoo.fr, khamlichi7@yahoo.es, adriouach@hotmail.com Laboratory LMM University of Lyon, Villeurbanne, FRANCE eric.jacquelin@univ-lyon1.fr Keywords: Inverse problem, Force reconstruction, Non punctual impact, Bayesian approach
2

ABSTRACT Knowledge of a force exerted on a mechanical system constitutes very important information in various applications dealing with dynamics of structures. However, direct measure of the force by transducers is so difficult in practice, and even impossible in many situations. This is why indirect methods are preferred in this field. These consist in reconstruction of the force by solution of the inverse problem. Various methods have been developed during last decades which enable identification of force characteristics through measuring the response in some locations of the elastic structure [1]. Considering the time domain approach to the problem, great attention has been paid to the stability issue of inversion operation like for instance applying singular value decomposition followed by Tikhonov regularization [2]. In this work, force reconstruction is considered within the context of Bayesian approach which considers the unknowns as random variables [3-5]. This approach provides a rigorously probabilistic framework which incorporates all sources of uncertainties such as those resulting from the actual model used or those associated to noise during measurement. Bayesian reconstruction of the force is conducted through Markov Chain Monte Carlo sampling method (MCMC). The Bayesian based incorporates both the a priori available information and the a posteriori information provided by the measured response. Estimation of the inverse problem solution can then be obtained by the conditional mean of the a posteriori density of probabilities. A planar beam having a symmetric section and loaded orthogonally to its mean fiber in its plane of symmetry is considered. It is assumed to be made from a homogeneous elastic material and having a uniform cross section. The beam is subjected to a force resulting from a non punctual impact where the acting pressure is assumed to be uniform on the whole impact zone which is considered to be perfectly known. The inverse problem was then only about reconstruction of the pressure time signal from the measured strain. A linear beam model that integrates the presence of some additive measurement noise was considered. The applied

and Hall, 1995

pressure was assumed to be of the form p(t) = at 2e bt . Estimation of the inverse problem solution was then performed through the conditional mean of pressure knowing the measured response. Evaluation of this expectation was performed through the ergodic mean of a homogeneous Markov chain process that was generated by operating Hastings-Metropolis algorithm on the a posteriori density of probabilities. . REFERENCES [1] H. Choi, A. Thite, D. Thompson, Comparison of methods for parameter selection in Tikhonov regularization with application to inverse force determination, Journal of Sound and Vibration 304: 894-917, 2007. [2] E. Jacquelin, A. Bennani, P. Hamelin, Force reconstruction: analysis and regularization of a deconvolution problem, Journal of Sound and Vibration 265: 81-107, 2003. [3] P. Lee, Bayesian Statistics: An Introduction, Arnold Publication, 1997. [4] W. Hastings, Monte Carlo sampling methods using Markov chains and their applications, Biometrika 57: 97109, 1970. [5] W. Gilks, S. Richardson, D. Spiegelhalter, Markov Chain Monte Carlo in Practice, Chapman

Vous aimerez peut-être aussi