Prediction of Frying Oil Quality Based on Polynomial Curve Fitting

It is cumbersome to determine the quality of frying oil by measuring the physical and chemical indicators of frying oil.It is necessary to find a way to quickly determine the quality of frying oil. The physicochemical index data such as acid value, carbonyl value and polar component were obtained by performing soybean frying experiments at five temperatures of 160, 175, 190, 205, and 220℃, and polynomial fitting modeling was performed. On the basis of the initial fitting model, the abnormal data with the absolute value of the studentized residual > 3 or the Cook distance > 1 were deleted, and the polynomial fitting modeling was performed again with the remaining data. The model was used to predict the physical and chemical physics of other unknown time points. Both the acid value and polar component were fitted by a quadratic polynomial, and the carbonyl value was fitted by a cubic polynomial. By fitting the result, R² was above 0.8628. Three unknown oil samples were taken at each temperature, and their three physical and chemical indicators were measured as the measured values, the fitting equation obtained by each model was calculated as the fitted value for the verification experiment, and the correlation was above 0.9968. It indicated that the model was stable and could forecast the unknown oil sample well.