Experimental verification in static tests of a computational method for detecting the position of a defect

Abstract

This paper presents the results of an experimental study of the detection of a defect in steel plates based on an optimisation algorithm using small amplitude homogenisation. In a first approximation using numerical simulation, this method detected inclusions using boundaries measurements obtained from static tests in the context of linear elasticity. We study the performance of such method by introducing direct measurements of real experiments. The defects in the specimens generate very small perturbations in displacements and therefore, it was necessary to perform several experimental realisations of the test procedure for dealing with random and systematic influences. We included in the analysis practical engineering considerations such as bolted and welded connections, a typical experimental setup and limited instrumentation. Detection results using experimental measurements shows that it is possible to use this method in real applications and that the results are improved when using an iterative scheme, although with limitations. For instance, the need to use larger measurement zones than those used in numerical simulations in order to avoid potential bias in the detection of defects due to errors in measurement. In this sense, it is also necessary the validation of support and loading conditions used in the simulations.