Sparse Principal Component Thermography for Structural Health Monitoring of Composite Structures

Non-destructive testing (NDT) techniques play an important role in structural health monitoring (SHM) of composite structures, among which infrared thermography (IRT) is popular because it is easy to operate, enables rapid inspection of large areas, and presents results as easily interpreted thermal images. In order to achieve noise reduction, feature extraction, and data compression, principal component thermography (PCT) was developed for thermographic data processing. However, each principal component in PCT is a linear combination of all the original pixel values, making the results difficult to interpret and hence affecting defect identification. In this work, sparse principal component thermography (SPCT) is proposed as an improved version of PCT, which provides more interpretable analysis results owing to its structure sparsity and leads to a better defect detection. The feasibility of SPCT is illustrated with two case studies.