EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Data-Driven FTIR–PLS Modeling of Ethanol in Agave Distillates: Chemometric Design Factors and Matrix Sensitivity

Deyani Nocedo-Mena, Melanie Sarahí Mateo González, Orestes Boffill Beltrán, Eduardo Gerardo Pérez-Tijerina

AI Chemistry · 2026

DOI: 10.3390/aichem1040018 ↗

Kurz gefasst

Fachartikel in der Zeitschrift AI Chemistry, 2026. Verfasst von Deyani Nocedo-Mena und 3 weiteren Autor:innen. Die Arbeit nennt 40 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Secretariat of Science, Humanities, Technology and Innovation.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Agave distillates are chemically heterogeneous matrices whose composition may affect the performance and transferability of spectroscopic calibration models. Accurate determination of their ethanol content is therefore essential for quality control and regulatory compliance. This study established a comparative ATR-FTIR chemometric framework for ethanol prediction using 46 agave-distillate samples measured in triplicate. Thirty-five samples were assigned to calibration and 11 to a validation set. Eighteen principal PLS configurations were developed by varying spectral-data strategy, spectral-region combination, and number of factors. Replicate dispersion, first- and second-derivative preprocessing, and model dimensionality were also evaluated, and PLS was benchmarked against PCR and PCA–SVR. Models based on averaged spectra produced lower RMSEP and higher rPred values than their corresponding individual-replicate configurations in all nine pairwise comparisons. The best averaged-spectrum PLS model achieved an RMSEP of 0.96% v/v and an rPred of 0.9897, whereas the best individual-replicate configuration yielded an RMSEP of 2.37% v/v. The benchmark showed that model performance depended on dimensionality and validation criterion; consequently, no regression approach was universally superior. Application of representative PLS, PCR, and PCA–SVR models to three additional samples produced similar descriptive mean absolute errors of 2.16 to 2.32% v/v. Overall, ATR-FTIR combined with multivariate regression provides a promising proof-of-concept strategy for the rapid estimation of ethanol in agave distillates. These findings support the future development of accessible and reliable quality-control protocols for agave spirits and other complex distilled beverages.

Förderung

  • Secretariat of Science, Humanities, Technology and Innovation

Zitieren

APA

Nocedo-Mena, D., González, M. S. M., Beltrán, O. B., & Pérez-Tijerina, E. G. (2026). Data-Driven FTIR–PLS Modeling of Ethanol in Agave Distillates: Chemometric Design Factors and Matrix Sensitivity. AI Chemistry. https://doi.org/10.3390/aichem1040018

Vancouver

Nocedo-Mena D, González MSM, Beltrán OB, Pérez-Tijerina EG. Data-Driven FTIR–PLS Modeling of Ethanol in Agave Distillates: Chemometric Design Factors and Matrix Sensitivity. AI Chemistry. 2026. doi:10.3390/aichem1040018

BibTeX
@article{NocedoMena2026,
  author = {Nocedo-Mena, Deyani and González, Melanie Sarahí Mateo and Beltrán, Orestes Boffill and Pérez-Tijerina, Eduardo Gerardo},
  title = {Data-Driven FTIR–PLS Modeling of Ethanol in Agave Distillates: Chemometric Design Factors and Matrix Sensitivity},
  journal = {AI Chemistry},
  year = {2026},
  doi = {10.3390/aichem1040018}
}
Kostenlos starten →