Detection of holes in a plate using global optimization and parameter identification techniques

In this work, an inverse problem of damage identification and localization is modelled using independent techniques, both for the direct model and for the inverse model. The damage is characterized by a hole in the structure, which modifies existing temperature and stress fields. Direct models for the thermal (conduction) and elastostatic problems are needed to obtain the distribution of these quantities in the domain, for a particular configuration. The boundary element method is used here as the direct problem, and two different and independent techniques are used for the inverse problem, in order to localize and to identify a damage in a structure. The first technique adopted is a global optimization technique using genetic algorithms and the second approach is a parameter identification technique using artificial neural networks. The identification and localization of a hole in the structure is performed using these two techniques for the inverse problem, with comparable results.