Reconstruction of Natural Crack Shapes from the ECT Signals by Using an Artificial Neural Networks Forward Model and an Optimization Approach

This paper presents work on a natural crack identification problem from eddy current testing (ECT) signals. ECT is a widely used in-service nondestructive testing (NDT) technique. A crucial problem in ECT is to inverse flaw profile from testing signals. Iterative inversion algorithms are commonly used to solve this problem. Typical iterative inversion approaches use a numerical forward model to predict the measurement signal from a given defect profile. But the use of numerical models is computationally intensive. In this study, the reconstruction of natural crack shapes from the ECT signals is realized by utilizing artificial neural networks as the forward solver and applying a metaheuristics-based optimization method. The crack is successfully reconstructed that verified both the efficiency of the artificial neural network forward scheme and the feasibility of the metaheuristics-based inversion method.