基于中红外光谱法检测煎炸油极性组分

Detection of polar components in frying oil based on mid-infrared spectroscopy

目的 利用中红外光谱技术实现对煎炸油极性组分的快速检测。方法 根据光谱-理化值共生距离分类法对煎炸油中红外光谱数据进行样本划分,从而得到校正集和预测集。采用卷积(Savitzy-Golay,S-G)平滑+一阶导数预处理手段,利用竞争自适应重加权算法(competitive adaptive reweighted sampling, CARS)进行特征提取,建立煎炸油极性组分含量的偏最小二乘回归(partial least squares regression, PLSR)预测模型,并利用BP神经网络算法对模型进行优化。结果 BP神经网络算法建立的模型校正集决定系数为0.9032,校正集均方根误差(root means quare error of calibration,RMSEC)为0.1264,预测集决定系数为0.8569,预测集均方根误差(root mean square error of prediction, RMSEP)为0.0625。经BP神经网络算法优化后,均方根误差明显减小,提高了预测模型的准确性。结论 结合BP神经网络算法的中红外光谱技术是一种检测煎炸油极性组分的有效方法,为食用油品质的快速检测提供理论指导和技术支撑。

Objective To rapidly detect polar components in frying oil by mid-infrared spectroscopy.Methods According to the sample set partitioning based on joint X-Y distance, the samples of infrared spectral data in frying oil were divided, and the correction set and prediction set were obtained. Moreover, competitive adaptive reweighted sampling(CARS) algorithm was used for feature extraction after Savitzy-Golay(S-G)smoothing+1stderivative preprocessing. The prediction model of polar components had been established by partial least squares regression(PLSR) methods, which was optimized by BP neural network algorithm. Results The model was established by BP neural network algorithm, the coefficient of determination and root mean square error of calibration(RMSEC) were 0.9032 and 0.1264, respectively. The coefficient of determination and root mean square error of prediction(RMSEP) were 0.8569 and 0.0625, respectively. After optimized by BP neural network algorithm, the root-mean-square error decreased obviously, which improved the accuracy of the prediction model.Conclusion Mid-infrared spectroscopy combined with BP neural network algorithm is an effective method to detect polar components of frying oil, which provides theoretical guidance and technical support for rapid quality detection of edible oil.