Principal component analysis for fast and automated thermographic inspection of internal structures in sandwich parts

Abstract. Rising demand and increasing cost pressure for lightweight materials – such as sandwich structures – drives the manufacturing industry to improve automation in production and quality inspection. Quality inspection of honeycomb sandwich components with infrared (IR) thermography can be automated using image classification algorithms. This paper shows how principal component analysis (PCA) via singular value decomposition (SVD) is applied to compress data in an IR-video sequence in order to save processing time in the subsequent step of image classification. According to PCA theory, an orthogonal transformation can project data into a lower dimensional subspace with linearly uncorrelated principal components preserving all original information. The effect of data reduction is confirmed with experimental data from IR-video sequences of simple square-pulsed thermal loadings on aramid honeycomb-sandwich components with CFRP/GFRP (carbon-/glass-fiber-reinforced plastic) facings and GFRP inserts. Hence, processing time for image classification can be saved by reducing the dimension of information used by the classification algorithm without losing accuracy.