Establishment of Rapid Identification Model of Edible Oil Adulterated with Frying Oil by Ultraviolet Spectrum Combined with BP Neural Network Algorithm
To establish a method for rapidly detecting adulteration in frying oils,ultraviolet spectroscopy was em-ployed to distinguish adulterants.In this study,soybean oil,corn oil,and sunflower seed oil were selected as repre-sentative oils and were subjected to frying.Adulterated oil samples were prepared by blending pure oils with frying oils for different frying durations(0-6 h)and varying levels of adulteration(0％-90％).The ultraviolet spectra were pre-processed by using a second-derivative transformation,and a model for detecting adulteration in frying oils was established by combining the processed spectral characteristics with the Backpropagation(BP)neural net-work algorithm.This model was used to analyze the type of adulterant,frying time,and adulterant content.The re-sults indicated that after the second-derivative transformation,the spectral characteristic peaks for adulterated frying oils were as follows:446 nm and 462 nm for soybean oil,268 nm and 274 nm for corn oil,and 280 nm and 288 nm for sunflower seed oil.In accordance with these peak positions and values,the recognition rates of the Levenberg-Marquardt algorithm(LMA),Momentum Gradient Descent(MGD),and Elastic Gradient Descent(EGD)for iden-tifying the adulteration model were 98.15％,91.67％,and 95.52％,respectively.
NETL
NSTL
万方数据
