Inversion of eddy current NDE signals using artificial neural network based forward model and particle swarm optimization algorithm

An inversion algorithm for the reconstruction of natural crack shape from eddy current testing signals is developed by using an artificial neural network based forward model and particle swarm optimization algorithm. Eddy current inspections are performed to measure signals caused by fatigue cracks introduced into plate specimens. The preprocessed ECT signals and the true crack shapes are used in the training of neural network. The parameters of the particle swarm optimization algorithm are modified and the results are discussed. The reconstruction results of crack shape verified both the efficiency of neural network based forward model and the promising of particle swarm optimization algorithm in crack shape inversion.