Discrimination Of Royal Jelly Storage Conditions And Rapid Quality Detection Based On FTIR Spectroscopy

The royal jelly collection process is cumbersome and time-consuming,and is greatly affected by storage conditions.The quality of royal jelly on the market is uneven.The existing detection methods cannot meet the requirements of rapid and accurate detection at the same time.Since the chemical bond information in mid-infrared spectroscopy can reflect the quality of royal jelly,and its conventional indicators(moisture,water-soluble protein,total sugar,acidity,viscosity,color difference)related to changes,this subject takes royal jelly as the research object,and establishes a qualitative and quantitative rapid detection method for the storage conditions of royal jelly and mid-infrared spectroscopy,which has theoretical significance and application value.The main research contents and results are as follows:1 Qualitative identification of royal jelly storage conditions based on conventional indicators and mid-infrared spectrum informationIn order to compare conventional indicators and mid-infrared spectroscopy to distinguish the quality of royal jelly,and to study the feasibility of replacing conventional indicators with mid-infrared spectroscopy,corresponding classification models of royal jelly under different storage conditions was established by the conventional indictors,mid-infrared spectrum and the information fusion of the two,respectively.In order to make a qualitative judgment on the storage conditions of royal jelly.Principal component analysis is used to extract feature information.The recognition rate of the temperature(frozen-normal temperature)two-classification model verification set established by the support vector machine is 88.89%,and the recognition rate of the three-classification model verification set of the time(7-14-30d)is 87.5%.The recognition rate of the verification set of the temperature two-classification model based on the mid-infrared spectrum information was94.44%,and the verification set of the time three-classification model was 94.44%.The recognition rate of the verification set based on the temperature two-classification model after information fusion is 97.22%,and the verification set recognition rate of the time three-classification model is 100%.The results show that the prediction model based on mid-infrared spectroscopy is more accurate than the conventional index,and the combination of the two can further improve the recognition rate.2 Linear quantitative prediction of royal jelly quality indicators based on PLSRepresentative selection of royal jelly quality indictors(water,water-soluble protein,total sugars),are used to established the infrared spectrum information corresponding partial least squares(PLS)model for quantitative prediction.By comparing three different pretreatments prediction effects,the best preprocessing methods of the model are SG smoothing + first derivative + SNV,SG smoothing +first derivative,SG smoothing + second derivative,respectively.Besides,by comparing different wavelength selection methods,the competitive adaptive weighting method(CARS)has the optimal prediction effect.When the selected characteristic wavelengths were 96,84 and 111,the correlation coefficients of the established partial least squares model were 0.9555 0.9631 and 0.9850,and the relevant root mean square errors were 0.0046,0.0913 and 0.4244,respectively.This result shows that processing the mid-infrared spectroscopy information by the PLS method can better predict the conventional indicators of royal jelly and can be applied to the rapid detection of royal jelly quality.3 Non-linear quantitative prediction of royal jelly quality parameters based on BP neural networkIn order to compare the quantitative prediction effects of various spectrum processing algorithms,the BP neural network algorithm is used to process the mid-infrared spectrum information,and the absorption value of the mid-infrared spectrum characteristic wavelength selected in part 2 is used as the input layer neuron of the BP neural network to establish three representative quality indictors prediction models predicting the quality parameters of royal jelly quantitatively.The results show that compared with the commonly used preprocessing PLS linear model,the BP neural network model obtained by evaluating different network structure combinations shows a significant decrease in the root mean square errors of water,water-soluble protein,and total sugar at 0.0032,0.0058,and 0.0069,respectively.While ensuring the accuracy of the calibration set,its correlation coefficient can also be maintained at 0.9353,0.9533,0.9563,which has a good prediction effect.