Defect localisation and quantitative identification in multi-layer conductive structures based on projection pursuit algorithm

ABSTRACT

Pulsed eddy current (PEC) technology has become a burgeoning method for detection and analysis of multi-layer conductive structures owing to rich time and frequency domain information presented by PEC signals. In this study, PEC technique is applied to characterise hidden-defect parameters while nondestructively inspecting multi-layer structures. A projection pursuit (PP) feature extraction method based on the information divergence index is investigated to effectively analyse PEC signals. An improved accelerating genetic algorithm is adopted to find the optimal projection direction. The signal’s dimension is reduced with minimal information loss while the data’s structure is preserved to the greatest degree. The features extracted on the basis of PP are simultaneously employed in crack localisation and crack length quantitative evaluation combined with a SVM classifier. The theoretical analysis and experimental results demonstrate that compared with the principal component analysis method, the features extracted by the presented PP algorithm work better for simultaneously characterising crack’s depth and size information and it reflect the inherent laws of the data, which make the features more physically interpretation meanwhile. Inversion accuracy for smaller and deeper cracks is enhanced obviously which will be helpful for crack localisation and quantitative identification of crack parameters in difficult situations.

KEYWORDS:

• Pulsed eddy current
• multi-layer conductive structure
• hidden cracks
• projection pursuit algorithm

Acknowledgments

This work was supported by the Natural Science Foundation of China [Grant No. 61174005] and the Technology Major Project of China [Grant No. 2016ZX0517-003 and No. 31300028-18-ZC0613-0002].

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This work was supported by the Technology Major Project of China [No. 2016ZX0517-003,No. 31300028-18-ZC0613-0002];the Natural Science Foundation of China [No. 61174005];