Voids indentification by isogeometric boundary element and neural network algorithms

This paper investigates the potential of the concomitant use of both Isogeometric Boundary Element Method (IGABEM) and Artificial Neural Networks Algorithm (ANN) to determine the number, position and geometric shapes of voids in a plate subjected to lateral pressure. In the proposed approach the boundary conditions are given, and the displacements of a finite number of points provide the information required to define the geometric characteristics of one or more internal voids. Exploiting the potentialities of IGABEM, it is possible to achieve also complex geometries with a level of accuracy unthinkable with the shape functions commonly used in other numerical methods. Besides, the richness of the space of configurations obtainable with the isogeometric approach can be successfully handled by the ability of ANN to solve inverse problems with a high level of complexities. The concurrent use of both returns a powerful tool whose potentialities in solving inverse problems are here explored and discussed.