Neural network approach to locating acoustic emission sources in nondestructive evaluation

Many methods have been proposed for the nondestructive evaluation (NDE) of objects or structures. The analysis of acoustic emission (AE) signals produced during cracking in a material is a promising approach for NDE. We discuss the advantages and disadvantages of AE testing and describe some of the difficulties in applying classical signal processing (deconvolution) techniques to AE analysis of a complex system. In particular, determining the location of the source of an AE is a highly nonlinear inversion problem for which classical deconvolution approaches are inapplicable. We present a neural network (NN) approach that has the potential to overcome many of the difficulties in nonlinear inversion. The approach is demonstrated on a steel I-beam. Although work remains in taking the concept to a large-scale practical implementation, the approach seems in principle to address some of the most vexing aspects of this challenging problem