Three-dimensional subsurface defect reconstruction for industrial components using pulsed thermography

Pulsed thermography is a promising method for detecting subsurface defects, but most pulsed thermographic inspection results are represented in the form of 2D images. Such a representation can limit the understanding of where the defects initiate and how they grow by time, which is a key to predict the remaining use of life of component and feedback to the design to avoid such defects. Threedimensional subsurface defect visualisation is a solution that can unlock this limitation. A straightforward approach to reconstruct 3D subsurface defect is conducting two inspections on both front and rear sides. However, the deployment of this approach can be limited because 1) one side of the inspected component could be inaccessible; 2) the accuracy of measurement could be compromised if the defect thickness is very thin due to extreme closed values of defect depths from two inspections; and 3) if the defect is too deep for one side, the defect could be missed. Addressing the challenge of 3D subsurface defect reconstruction and visualisation, this thesis proposes a novel technique to measure defect depth and estimate defect thickness simultaneously through estimating the thermal wave reflection coefficient value achieved by introducing a modified heat transfer model based on a single-side inspection method. The proposed method is validated through model simulations, experimental studies, and a use case. Four composite samples with different defect types, sizes, depths and thicknesses, are used for experimental studies; a steel sample with a ‘s’ shape triangular air-gap inside is used for a use case. The simulation results show that under the noise level of 25 dB, the percentage error of the developed depth measurement method is 0.25% whilst the minimum error of the best existing method is 2.25%. From the experimental study results, the averaged percentage error of the defect thickness estimation is less than 10% if the defect thickness is no more than 3 mm. For the use case, the reconstructed defect shape is similar to the X-ray image.