Simple digital image processing applied to thermographic data for the detection of cracks via eddy current thermography

Abstract

In the present work, our objective is to obtain reliable results for the detection of cracks through the use of simple digital image processing methods as an alternative to computationally intense and time-consuming data processing techniques, such as the discrete Fourier transform. These methods were applied to existing experimental data (raw thermograms) obtained via eddy current thermography on aluminum plates for crack detection. The techniques employed are the depiction of: (1) the norm of the spatial derivatives of the thermal field at various time instants, (2) the mean temperature value over a time window, and (3) the norm of the spatial derivatives of the mean value of temperature. Moreover, we visualized certain operations between the norms of the spatial derivatives, such as the square of the norm of the spatial derivatives and the product of the norm of the spatial derivatives between consecutive time instants. Through the proposed methods, the location and shape of the cracks were detected and identified both within the excitation period and after the excitation period had ended. The cracks were successfully detected regardless of their location in the aluminum plates and their orientation with respect to the heat flow. In certain cases, the results obtained from the operations between the norms of the spatial derivatives of temperature showed that the pattern of the exciting coil was eliminated and the shape of the crack was enhanced. A key parameter for the effectiveness of the techniques is to achieve temperature gradients around the crack that are stronger than those that are due to the non-uniform heating of the work-piece by the coil.