Three-Dimensional Morphological Feature Quantization of the Aero-Engine Turbine Disc with Super-Resolution Industrial Computed Laminography

Industrial computed laminography (ICL) is a three-dimensional (3D) non-destructive imaging method widely used in industrial digital imaging of plate-shell components. However, for imaging aero-engine turbine discs with a sizeable length-width-thickness ratio, there are problems, such as unclear feature details in the laminographic image and the inability to quantify 3D feature distribution. Deep learning is widely used in super-resolution reconstructing image details, which can solve parts of the above problems. Therefore, a 3D feature quantization technology combined with a deep learning network is proposed. By improving the resolution of the laminographic image of the aero-engine turbine disc using a deep learning network, the detailed morphological feature can be obtained and quantitatively analyzed using the 3D nearest neighbor index (NNI) mathematical model to get the multi-scale information of features. This work uses the 3D NNI combined with deep learning to analyze ICL quantitatively reconstructed internal features.