determination of butylated hydroxytoluene in edible oil using near-infrared spectroscopy combined with multivariate analysis

Edible oil is a fundamental component of human daily diet and has a wide impact on global agriculture and economy. The issue of additives in edible oil has always been a concern for its quality and safety, as excessive use of additives can be harmful to human health. This study suggests a quantitative detection technique for determining the concentration of butylated hydroxytoluene (BHT) in edible oil using near-infrared spectroscopy (NIRS) technology. Firstly, the spectral characteristics of edible oil samples with different concentrations of BHT are characterized using a near-infrared spectrometer. Then, particle swarm optimization (PSO), butterfly optimization algorithm (BOA), and fox optimizer (FOX) are employed to optimize the preprocessed spectral wavelengths. Finally, a detection model based on the optimized wavelengths is established using backpropagation neural network (BP-ANN) to achieve quantitative detection of BHT content in edible oil. The results show that the model based on the feature wavelengths optimized by the FOX algorithm has the best detection performance. This model takes 7 features as inputs, with a root mean square error of prediction (RMSEP) of 1.6116 g·kg−1 and a prediction determination coefficient (RP) of 0.9900 on the prediction set. The results demonstrate that near-infrared spectroscopy technology can accurately predict the BHT content in edible oil.