Stochastic identification of imperfections in a submerged shell structure

• Published in: Computer methods in applied mechanics and engineering Vol. 272; pp. 58 - 82
• Main Authors: Reed, H.M., Earls, C.J., Nichols, J.M.
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 15-04-2014
• Subjects: Bayesian estimation; Computer simulation; Damage identification; Defects; Estimates; Fluid structure interaction; Markov chain Monte Carlo; Markov chains; Mathematical models; Monte Carlo methods; Reversible jump Markov chain Monte Carlo; Shell structure; Shells; Submerged; Bayesian estimation; Damage identification; Fluid structure interaction; Shell structure; Reversible jump Markov chain Monte Carlo; Markov chain Monte Carlo
• ISSN: 0045-7825; 1879-2138
• DOI: 10.1016/j.cma.2014.01.003

•Geometric imperfection parameters are estimated in a fluid–structure interaction system.•Inverse problem solved using Markov chain Monte Carlo sampling.•Non-contact approach to damage identification is explored.•Uncertainty surrounding the parameters is quantified.
Accurate predictions of the buckling load in imperfection sensitive shell structures requires precise knowledge of the location and magnitude of any geometric imperfections in the shell (e.g. dents). This work describes a non-contact approach to identifying such imperfections in a submerged shell structure. By monitoring the acoustic pressure field at discrete points proximal to a shell structure excited by a cyclic membrane (i.e. in-plane) loading, it is noticed that parameters, describing small scale denting, can be identified. In order to perform the identification, a fluid-structure model that predicts the spatio-temporal pressure field is required. This model is described in detail and includes the predicted effects of the imperfection on the observations. A Bayesian, Markov chain Monte Carlo approach is then used to generate the imperfection parameter estimates and quantify the uncertainty in those estimates. Additionally: for cases involving the occurrence of an unknown number of dents, reversible jump Markov chain Monte Carlo (RJMCMC) methods are employed in this work.