Automatic multi-frequency rotating-probe eddy-current data analysis

An automatic scheme for analyzing multi-frequency rotating-probe eddy-current data is proposed herein. This system integrates signal/image-processing algorithm with pattern recognition methods for accomplishing the objectives. The eddy current signals are acquired from several kinds of frequency multiplexed eddy-current rotating-probes. The problem involves the detection of the flaw signals, classifying the defect format and sizing/characterizing the defect profile. The preprocessing steps include conversion of one-dimensional data to obtain a two-dimensional image, removing background noise, suppressing structure-masking signals, and calibration. Optimal thresholding of calibrated signal based on probability of detection (POD) concepts are discussed in detail. Feature extraction and signal classification steps are then implemented to discriminate signals produced by defects or non-defects, axial or circumferential defects, tight or volumetric defects. Finally, wavelet basis function neural networks are used for estimating defect profile.;Further analysis of the statistical properties of potential defect signals is needed to discriminate between different kinds of defects. A model for characterizing the amplitude and phase probability distributions is developed. The squared amplitudes and phases of the potential defect signals are modeled as independent, identically distributed (i.i.d.) random variables following gamma and von Mises distributions, respectively. A maximum likelihood (ML) method is employed for estimating the amplitude and phase distribution parameters from measurements corrupted by additive complex white Gaussian noise. Newton-Raphson iteration is utilized to compute the ML estimates of the unknown parameters. Cramer-Rao bounds (CRBs) for the unknown parameters are computed. The obtained estimates can be utilized for maximum a posteriori (MAP) signal phase and amplitude estimation as well as efficient feature extractors in a defect classification scheme. Numerical examples of both real and simulated data are presented to demonstrate the performance of the proposed method