Tiny scale defect contour recognition on curved structure based on electromagnetic thermography

Precision spherical pressure vessels are used in critical pressure-bearing structures due to their excellent structural strength. The surface of the spherical shell may experience corrosion in long-term service, leading to the formation of tiny pitting defects, which cause a risk to the structural integrity. A high-precision, non-destructive method is required for tiny defects detecting and contour recognition. In this paper, firstly, a new sensor consisting of a rotatable yoke and spherical adaptive flexible material with high permeability (FMHP) is designed to improve the performance of the electromagnetic thermography, which achieves the detection of tiny scale defects in spherical shell surface with diameter of 40μm. Secondly, super-resolution algorithms based on machine learning and deep learning are developed to realize the contour recognition of tiny defects, indicating that the generative adversarial network has an optimum performance. Then, to address the distortion phenomenon in infrared imaging of spherical structures, a coordinate transformation-based image correction algorithm is developed, enabling the accurate reconstruction of defect contours.