Quantitative Detection And Analysis Of Flour BPO Based On NIR Technology

The Near Infrared Reflectance Spectroscopy(NIRS) which has been widely used in the quantitative and qualitative analysis of crops and their by-products, is characterized by speed, non-damage and non-pollution. As is known that incidents of food safety occur frequently in China, but the traditional chemical metrology method can not realize the speedy detection in the field. Under such circumstance, this research proposed designing a prediction model based on Radial Basis Function(RBF) neural network to predict the concentration of Benzoyl Peroxide(BPO) in flour. 100 different concentration samples we got by adding BPO to plain flour were analyzed, and the original spectra were pre-processed using Standard Normal Variate(SNV) method and abnormal sample elimination to strengthen the spectral feature by removing the scatter, the shift, etc. A NIRS model was designed to predict the concentration of BPO in flour in the 36 samples pre-processed by means of Partial Least Squares(PLS), BP neural network and RBF, respectively, with prediction correlation coefficient(R), root mean squared error of prediction(RMSEP) and ratio of performance to standard deviate(RPD) reaching 0.9937, 15.5095 and 8.8216, respectively. The results demonstrated that the predictive method proposed by this study had high prediction accuracy and feasibility providing quality evaluation and dynamic monitoring service for quality inspection department and consumers.