Harnessing infrared thermography and unsupervised clustering for the non-destructive evaluation of delamination in Chinese cloisonné enamels

Chinese cloisonné enamels are valued for their beauty but are prone to interfacial defects due to thermal mismatch between enamel and bronze. This study uses pulsed infrared thermography (PT) to detect delamination in these layered structures. A thermal quadrupole method is applied to model heat transfer and identify thermal resistance caused by defects. To enhance interpretation, techniques like thermographic signal reconstruction (TSR) are combined with two classification strategies. The first is a rule-based method using peak timing from thermal response curves. The second integrates deep autoencoders (DAE) and Gaussian Mixture Models (GMM) for unsupervised clustering. Together, these methods offer a fast, non-contact solution for evaluating cloisonné artifacts, supporting their conservation and restoration.