关键词: 低场核磁共振, 总极性化合物含量, 向后多元回归, 主成分回归, 偏最小二乘回归

Abstract:

The contents of total polar compounds (TPC) in frying oils were predicted by using low field nuclear magnetic
resonance (LF-NMR). LF-NMR T2 relaxation parameters (the relaxation time: T21, T22, and T23, the corresponding peak
areas: S21, S22, S23 and the single component time T2W) of the samples were collected. The TPC contents of frying oil were
also determined by column chromatography method as a reference standard. Three mathematic models were established to
quantitatively analyze TPC contents, using backward multiple linear regression (BMLR), principal component regression
(PCR) and partial least squares regression (PLSR), respectively. Comparing the correlation coefficients and root mean square
errors of calibration set and prediction set, the BMLR model showed the best reliability in predicting TPC contents, with a
correlation coefficient of prediction sets of 0.928, root mean square error of prediction (RMSEP) of 0.568%.

Key words: low field nuclear magnetic resonance (LF-NMR), total polar compounds (TPC) content, multiple linear regression (BMLR), principal component regression (PCR), partial least squares regression (PLSR)