Start Date

2016 12:00 AM

Description

Giant Magneto-Resistive (GMR) sensors have recently emerged as a promising tool for Eddy Current (EC) inspection of multi-layer riveted structures. However, the detection of defects under fasteners remains a challenging problem, since the distortion of the induced magnetic field introduced by subsurface flaws is masked by the strong responses from steel fasteners and structural edges. Multi-frequency EC-GMR measurements can be used to suppress the rivet signal, while retaining defect indications and hence increase the detectability of defects. In this paper, we propose a novel generic filter for processing EC-GMR measurements acquired at a single or multiple excitation frequencies to suppress the rivet signal with noise, and retain the defect indication. The proposed algorithm is based on linear combinations of the complex measurements, but controlled nonlinearity is introduced by applying the non-linear transformations, such as a power law, before feeding the signals to the filter. The coefficients of the linear combinations are iteratively updated to minimize a cost function. Measurements from a calibration sample with known rivet geometry are used in this step such that rivet indications are minimized while defect indications are retained. Performance of the proposed nonlinear mixing algorithm is evaluated on experimental GMR measurements obtained with rotating eddy current excitation. Initial results of performance of the algorithm are promising.

Language

en

File Format

application/pdf

Share

COinS
 
Jan 1st, 12:00 AM

Nonlinear Mixing Algorithm for Rivet Signal Suppression in EC-GMR Inspection of Multilayer Structures

Giant Magneto-Resistive (GMR) sensors have recently emerged as a promising tool for Eddy Current (EC) inspection of multi-layer riveted structures. However, the detection of defects under fasteners remains a challenging problem, since the distortion of the induced magnetic field introduced by subsurface flaws is masked by the strong responses from steel fasteners and structural edges. Multi-frequency EC-GMR measurements can be used to suppress the rivet signal, while retaining defect indications and hence increase the detectability of defects. In this paper, we propose a novel generic filter for processing EC-GMR measurements acquired at a single or multiple excitation frequencies to suppress the rivet signal with noise, and retain the defect indication. The proposed algorithm is based on linear combinations of the complex measurements, but controlled nonlinearity is introduced by applying the non-linear transformations, such as a power law, before feeding the signals to the filter. The coefficients of the linear combinations are iteratively updated to minimize a cost function. Measurements from a calibration sample with known rivet geometry are used in this step such that rivet indications are minimized while defect indications are retained. Performance of the proposed nonlinear mixing algorithm is evaluated on experimental GMR measurements obtained with rotating eddy current excitation. Initial results of performance of the algorithm are promising.