Infrared Thermal Imaging-Based Turbine Blade Crack Classification Using Deep Learning

Abstract Non-destructive testing is widely applied for the detection and identification of defects in turbine blades modern aircraft engines. Cracks can affect performance pose a risk to safety service life. For Original Equipment Manufacturers it is, therefore, essential be able identify all defects. Heat flow thermography offers, compared often used penetrant testing, potential improve contact-free, reproducible, quick apply, automated. With induction (heat flow) thermography, even possible detect cracks that lie below surface therefore are not externally visible. However, manual inspection images very time-consuming. By automating image classification procedure with deep learning technique, speed accuracy improved over manually performed classification. The development objective this AI application expected support assist highly skilled experienced specialists medium term. Our solution based on convolutional neural networks. Several challenges training process, including data imbalance, small dataset, extremely large addressed.