Physically interpretable unsupervised thermographic clustering for structural alteration diagnostics in ancient jade artifacts

Ancient jade artifacts often develop complex weathering and surface-to-near-surface alterations that are not apparent under visual inspection. This study proposes an unsupervised thermographic framework integrating pulsed thermography (PT), long-pulsed thermography (LPT), and Self-Organizing Map (SOM) clustering for non-invasive, label-free assessment. A differential Thermographic Signal Reconstruction (TSR) encoding separates depth-sensitive information: zero-order PT-TSR retains surface-dominated responses, whereas first-order LPT-TSR suppresses surface effects and enhances deeper diffusion contrasts. SOM directly operates on high-dimensional thermal sequences without aggressive feature compression, preserving response topology to cluster alteration-related thermophysical patterns and depth-dependent heterogeneity. Validation on a reference sample demonstrates reliable discrimination between shallow and deeper thermal features. Applied to a Shang-dynasty jade dagger, the method maps alteration-depth variations, identifies vulnerable regions, and reveals previously undocumented traces consistent with hafting, use-related contact, or burial contact–induced alteration (qin). The approach provides a physically interpretable and broadly applicable workflow for mineral-based cultural heritage diagnostics.