Speaker
Description
Eddy Current Thermography (ECT) has emerged as a powerful technique in industrial environments for detecting surface breaking cracks, offering a highly automatable alternative to traditional inspection methods such as fluorescent magnetic particle testing or liquid penetrant inspection. Unlike these conventional techniques, ECT provides a high level of detectability and lends itself well to automation, making it attractive for modern quality assurance processes.
The true potential of ECT becomes evident when applying frequency resolved analysis methods, such as lock in amplification or Fourier analysis. These approaches exploit the characteristic and highly recognizable patterns that appear when analyzing the phase of thermal oscillations. In the case of short surface breaking cracks, a distinct multilobed pattern typically emerges, also known as ‘Butterfly Pattern’, enabling inspectors to identify and locate defects. Moreover, the orientation of the lobes correlates with the orientation of the crack, allowing experimenters to intuitively estimate its direction with relative ease. These inherent patterns make ECT particularly well suited for AI enabled automatic detection, opening the door to advanced fully automated inspection systems in the near future.
In previous work, we demonstrated that modifying the orientation of induced currents during thermographic testing can significantly enhance defect detectability. When the current direction is rotated, the multilobed pattern rotates synchronously, while the background remains largely unaffected. This behavior allows the current orientation to be treated as a phase parameter, enabling the combination of multiple complex images into a single enhanced image through a weighted summation similar to the optical phase shifting (PS) technique widely used in interferometry.
In this contribution, we propose viewing the multilobed pattern as a vortex like field interacting with a background. By applying PS, we effectively remove the background and isolate the vortex-like structure. Furthermore, using cross correlation with an appropriate vortex containing basis, we enhance defect visibility even further, producing a “dark field” image where only the features of interest remain. This approach not only improves clarity but also facilitates interpretation and has the potential to enhance automated defect detection. The methodology was validated using both artificial and real defects, demonstrating its applicability to practical industrial scenarios where reliable crack detection is critical.
