Simultaneous determination of vegetable oil frying frequency and peroxide value based on the three-dimensional fluorescence spectroscopy and machine learning

Huihui Yang; Yutang Wang; Bei Fan; Qing Chen; Meifeng Wu; Fengzhong Wang; Long Li

doi:10.1016/j.foodchem.2024.142729

Simultaneous determination of vegetable oil frying frequency and peroxide value based on the three-dimensional fluorescence spectroscopy and machine learning

Food Chem. 2024 Dec 31:471:142729. doi: 10.1016/j.foodchem.2024.142729. Online ahead of print.

Authors

Huihui Yang¹, Yutang Wang², Bei Fan², Qing Chen², Meifeng Wu², Fengzhong Wang³, Long Li⁴

Affiliations

¹ Institute of Food Science and Technology, Chinese Academy of Agricultural Sciences (CAAS), Beijing 100193, PR China; Weifang Institute of food science and processing technology, Weifang 261000, PR China.
² Institute of Food Science and Technology, Chinese Academy of Agricultural Sciences (CAAS), Beijing 100193, PR China.
³ Institute of Food Science and Technology, Chinese Academy of Agricultural Sciences (CAAS), Beijing 100193, PR China; Weifang Institute of food science and processing technology, Weifang 261000, PR China. Electronic address: wangfengzhong@sina.com.
⁴ Institute of Food Science and Technology, Chinese Academy of Agricultural Sciences (CAAS), Beijing 100193, PR China; Weifang Institute of food science and processing technology, Weifang 261000, PR China. Electronic address: llzgnydx@163.com.

PMID: 39788013
DOI: 10.1016/j.foodchem.2024.142729

Abstract

The practice of deep-frying introduces various health concerns. Assessing the quality of frying oil is paramount. This study employs three-dimensional fluorescence spectroscopy to evaluate the peroxide value of vegetable oils after varying frying times. Three feature preprocessing techniques combined with three machine learning methods to predict the frying frequency and the peroxide value. For the prediction of frying frequency, SG-RF model performed the best for soybean oil, while for peanut and corn oils, SG-PLS excelled, with the Rp² of 0.98, 0.98, and 0.97, respectively. Regarding the quantification of peroxide value, normalize-kNN performs the best for soybean frying oil, for peanut frying oil SG-RF are optimal, and the optimal combination is normalization-RF for corn frying oil, with the Rp² of 0.87, 0.89, and 0.93, respectively. This rapid assessment method enables early detection of oil safety issues, protecting consumer health by preventing the use of degraded oils.

Keywords: Fluorescence spectroscopy; Frying oil; Machine learning; Peroxide value.