Rapid and non-destructive determination of moisture and numbing substances in dried and fresh Zanthoxylum bungeanum Maxim. fruit using NIR spectroscopy and chemometrics

This study developed prediction models for moisture and numbing substance (hydroxy‑β-sanshool) content in both dried and fresh Z. bungeanum fruit using near-infrared spectroscopy technology. The results indicated that the determination coefficients (R p 2 ) for predicting moisture and numbing substances in both dried and fresh Z. bungeanum fruit were 0.9426, 0.6913, 0.6800, and 0.5978, respectively. The root mean square errors of prediction (RMSEP) were 0.2071, 6.6548, 0.5481, and 0.9149, and the residual predictive deviations (RPD) were 4.30, 1.80, 2.41, and 1.90, respectively. Among these, the moisture content model for dried Z. bungeanum demonstrated superior prediction ability, high stability, accuracy, and a high degree of fit, indicating its suitability for predicting moisture content in unknown samples. This study offers a technological foundation for the non-destructive, rapid, and eco-friendly assessment of key quality indicators in Z. bungeanum. It also provides technical support for standardizing quality grading, ensuring transparency in the trading process, and advancing the intellectualization of processing methods.