Crack Characterization Method with Invariance to Noise Features for Eddy Current Inspection of Fasterner Sites

Abstract

There is a need for research in eddy current (EC) nondestructive evaluation (NDE) to improve the reliability to detect, locate and size cracks around fastener sites in multi-layer structures while minimizing the overall cost of inspection. The objective of this work is to develop feature extraction and classification algorithms for crack characterization with invariance to noise features for eddy current inspection of fastener sites. Model-based parametric studies were first performed to explore potential features under a wide array of crack, noise and material conditions. Through these studies, several features were identified to have some invariance to the characteristic asymmetric response due to gaps between the fastener and hole, probe liftoff variation, and probe skew. In particular, a promising feature with noise invariance to all non-flaw conditions considered in this study was found through investigating changes in the eddy current response along a circumferential direction in an annulus region away from the hole center. To obtain a measure of this localized crack feature, an approach was developed using a fit of a characteristic function to the data through nonlinear least squares estimation. A model-based optimization approach was also implemented to evaluate the best signal processing algorithm design to distinguish between several classes of crack size. Using this approach, an optimized measure was found to be well correlated with subsurface crack size and insensitive to noise conditions included in this study.