Forward modeling of cracks detection for RFEC inspection

Because of the poor prior knowledge and constraints, the quantitative inspection of pipeline cracks was an ill-posed problem in Remote Field Eddy Current Inspection. Some significant correlations between the cracks and the features of the magnetic field signals had been discovered through adequate Finite Element simulations on the axisymmetric defects of the pipeline here. Based on the correlations above, two forward models, which can quantitatively map the defects size to the features of the magnetic field signals, were proposed. By contrast, the model based on Back-Propagation Neural Networks had better approximation accuracy and generalization ability. It seems to be an effective reference to the quantitative inverse of the pipeline defects.