Multi-source learning reveals process–aging–performance relationships in Xuan paper

Xuan paper durability reflects raw materials, papermaking, storage history, and aging conditions. We integrated raw-material ratios and process descriptors with artificial aging, near-infrared (NIR) spectroscopy, hyperspectral imaging (HSI), and degree of polymerization (DP) measurements from 21 independent handmade Xuan paper samples. NIR–HSI fusion yielded the highest spectral-only estimates for DP, degree-of-polymerization retention (DP retention ), and chain-scission index (SCI). Cluster-bootstrap intervals supported fusion gains for DP retention and SCI, but not DP; SCI reached a paper-sample-grouped cross-validation R ² of 0.840. Stable spectral responses occurred mainly at 1974–1993 nm in NIR and 1079–1086 nm in HSI. The full process–aging interaction model (PAIM) achieved R ² values of 0.893, 0.962, and 0.966 for DP, DP retention , and SCI, respectively. Aging duration accounted for most PAIM performance, and only the SCI gain over the aging-only model remained consistently positive. This initial framework supports process refinement and conservation repair-paper research.