An Efficient Phenomenological Inversion Method for Reconstruction of Crack Depth Profile in a Metal From ACFM Probe Output Signals

Abstract

The paper proposes an efficient phenomenological inversion method to determine the depth profile of a surface-breaking crack in a metal from the output signal of an alternating current field measurement (ACFM) probe. The proposed method utilizes a conjugate gradient algorithm to minimize an objective function, representing the difference between probe predicted and actual signals in an iterative manner. The objective function is derived explicitly in terms of crack depth variables by considering a polynomial function for the field distribution in the depth direction and applying appropriate Green's functions. This approach enhances the accuracy and computational efficiency of the inversion process, regardless of the choice of initial crack depth profile or presence of noise in the measurement system. The validity and efficiency of the proposed method are demonstrated by comparing the reconstructed depth profiles of several simulated and machine-made cracks with their actual data, and those obtained using the conventional phenomenological approach based on an efficient stochastic optimization scheme along with a fast pseudoanalytic ACFM probe output simulator.

Fast and efficient inversion methods for crack sizing using Bz magnetic signature

“…A large data set consisting of crack signals is necessary for such a learning process and is often hard to achieve experimentally. In addition, the MLPs are impaired by noisy data, even with a large data set of crack signals 5,43 …”

“…Commonly, the fatigue crack accumulates and develops in the longitudinal direction, 2,3 resulting in leakage and failure with a depth profile. 4,5 The eddy current testing (ECT) technique 6 and the alternating current field measurement (ACFM) technique 7,8 represent the two most important electromagnetic methods that can be used for crack inspection in industrial devices' metal structures. In both techniques, the metal surface is interrogated by the ECs induced by an AC carrier coil.…”

“…Calibration curves and artificial neural networks (ANNs) represent typical algorithmic methods. 5 In this context, different mathematical models and detection methods have been proposed. In Reference 18, Postolache proposed a neural network algorithm for rapid defect clustering in aluminum plates, such as cracks and holes.…”

“…The development of these cracks is typically initiated from the metal surface experiencing high stresses. Commonly, the fatigue crack accumulates and develops in the longitudinal direction, 2,3 resulting in leakage and failure with a depth profile 4,5 …”

“…Algorithmic methods, on the other hand, use a suitable signal processing technique that maps the crack signal with its geometric parameters. Calibration curves and artificial neural networks (ANNs) represent typical algorithmic methods 5 . In this context, different mathematical models and detection methods have been proposed.…”