Eigen Background Subtraction for Industrial Flaw Detection: Application to High-Pressure Turbine Blade CT Scans
Clément Remacha, G. Redoulès, Axel Aublet · Journal of Nondestructive Evaluation · 2023
Abstract We propose a state-of-the-art approach that is the first to use Eigen background subtraction to reveal flaws in three-dimensional Computed Tomography data. Our method is composed of two main steps. During the first step, principal component analysis (PCA) is applied on flaw-free blade stack data. From a statistical perspective, a series of “flaw-free” characteristic functions is extracted. The second step consists of decomposing the blade of interest according to the functions calculated from PCA. This projection allows the construction of a synthetic blade without any flaws. A subtraction between the synthetic blade and real blade highlights the abnormal variations. The main advantage of this technique is that the processing remains applicable even when the measurement system or parts of the system have variability that is greater than the flaw size.