An approach to the characterization of anomalies in tubes using acoustic tests

Abstract

Pipelines are common components in several sectors of industry and in particular in refineries. Their maintenance is vital due to the fact that any non planned stoppage is associated to loss of efficiency and money; moreover, it could also lead to environmental and safety issues, like leakages and explosions. In this context, investigations regarding the improvement or development of non-destructive techniques to assess the structural integrity of pipelines is of interest of industry and academia. This work proposes and approach to investigate pipeline anomalies based on acoustic signals. A Finite Element Model (FEM)is built to describe the forward problem where both leakage and blockage may be properly described by variations of the cross-section area of the pipe. As the cross-section area variation may take place along the pipe, it is adopted a set of parameterizations to describe feasible scenarios. The adopted parameterizations also bring some level of regularization to the inverse analyses. The inverse problem is formulated in the Bayesian framework and the sampling of the posterior density of model parameters is performed by means of the Transition Markov Chain Monte Carlo (TMCMC) algorithm. A set of numerical examples are presented to assess the effectiveness of the approach.