Combining spectral super-elements, genetic algorithms, and massive parallelism for computationally efficient flaw detection in complex structures

Abstract

An iterative method to determine the location and size of flaws using dynamic response measurements is presented. The solution procedure combines frequency-domain problem solving methods (deconvolution and the spectral element method) with a stochastic iterative search technique based on a genetic algorithm. Experimentally measured acceleration and strain responses from example structures are used to verify the method.