X-ray absorption spectrum combined with deep neural network for on-line detection of beverage preservatives

Aiming at the problem that the common preservative detection methods are too cumbersome and costly, a rapid on-line detection method based on X-ray absorption spectrum (XAS) for beverage preservatives is proposed. Benzoic acid, potassium sorbate, sodium dehydrogenate, and propyl p-hydroxybenzoate were selected to form standard concentration and excess concentration solutions. XAS data were collected by an X-ray spectrum detection system, and the leave one out cross-validation method was used to divide the data set. Principal component analysis was used to extract the main features of the spectrum. Deep neural network (DNN) and support vector machine were established to classify samples of different powders and different concentrations of solutions. DNN reached 100% and 94.17% recognition rates on preservative powders and potassium sorbate solutions, respectively. The experimental results show that XAS combined with DNN can identify different types of preservative powder and detect whether the preservative content of the beverage exceeds the standard, which provides a new way to achieve fast on-line detection of preservative content in beverages.