Abstract: The paper presents a new method for detecting phospholipid content in soybean oil based on electrochemical analysis and to establish accurate and reliable calibration models. A total of 40 mixed soybean oil samples with different contents of phospholipids were prepared, and the standard values of phospholipid contents were precisely detected by molybdenum blue colorimetry. Considering no electron transfer during the phospholipase hydrolysis process, we developed an electrochemical sensor based on a multiple modified enzyme electrode to obtain electrochemical signals by cyclic voltammetry measurement through an electrochemical workstation. The original electrochemical data were denoised by either Savitzky-Golay smoothing filter or Daubechies (dbN) wavelet series. Through the comparative analysis, it was found that the best denoise was achieved based on db6 wavelet three-layer decomposition. Finally, four methods were used respectively to set up regression models between the electrochemical data and phospholipid concentration, i.e., linear fitting between reductive peak current and phospholipid concentration, principal component regression (PCR) model, partial least squares regression (PLSR) model and support vector machine regression (SVMR) model. Comparison of these methods showed that the prediction accuracy of the SVRR model based on radial basis kernel function was the highest. A good linear relationship was observed in the phospholipid concentration range of 5.87–304.89 mg/L, and the detection limit (LOD) was 1.68 mg/L (RSN = 3). The coefficient of determination (R2), root-mean-square error of prediction (RMSEP) and relative standard deviation (RSD) were 0.998 7, 0.288 9 and 2.55%, respectively, all of which met the requirements of practical detection.

Key words: soybean oil, phospholipid content, electrochemical analysis, multiple modified enzyme electrode, wavelet denoising, calibration model