Research On Quantitative Detection Of Iodine Value And Saponification Value Based On Spectroscopy And Type Identification Based On Characteristic Values In Edible Oil

Vegetable oil can provide the body with unsaturated fatty acids,vitamins which are needed as essential nutrients.After the completion of the oil processing,their quality will be affected by many aspects.Then we need to take measurements to control quality.Iodine value and saponification value can be used as parameters of oil quality analysis.They can be measured by chemical method,but this method has the features,including time consuming,high cost,and easy to pollute the environment.And requirements of the experimental operating are relatively high.Some chemical reagents will even produce some harm to the human body and taking time is not conducive to the realization of tracking measurement.Therefore,it is important to find a new method to determine the parameters of oil.Near Infrared and Raman spectroscopy developed rapidly in recent years that have the advantages of fast speed,high efficiency and non-preprocessing have been increasingly used in oil and food industry.It is study on oils and fats based on spectroscopic techniques.The main contents and conclusions are included:(1)the models to predict saponification value were established using parameters optimization algorithm including grid search(GS),particle swarm optimization(PSO),genetic algorithm(GA)combined with pretreatment methods based on NIR spectral analysis technique.Comparing with three kinds of optimization methods,the results showed GS and GA combined with SNV-DT method to establish the models had the best correlation coefficient of calibration set and prediction set,which are 93%,99% separately.(2)The detection models of iodine content were established separately based on NIR and Raman,which were done with the pre-processing and feature variables extraction in order to eliminate the useless information and simplify model.The result of NIR quantitative prediction showed the performance of model equipped with MSC and successive projections algorithm(SPA)was excellent.The correlation coefficient R of the prediction set was 98.8761% and the number of variables was 12;The prediction results of Raman quantitative models showed the excellent correlation coefficient R of the prediction set was 95.6785% and the number of variables was 307 when the Raman model was applied by savitzky golay15-asymmetric least squares-normalization(SG15-ALS-Nor)pretreatment and interval partial least squares(i PLS).Comparing with the two spectral models,the prediction results of NIR were more satisfacting.And it also shows that the spectral preprocessing method and the variable selection algorithm have a significant effect on the prediction results.(3)Four characteristic values including iodine value,palmitic acid,oleic acid and linoleic acid were applied to establish classification models.It indicated the results of three kinds of identification models based on the four characteristic values as input variables were very obvious.The accuracy of the PSO-SVC calibration and prediction sets were both 100%.There was no wrong in the process of training and predicting.The results show that it is feasible and effective to identify the oil species by using multiple characteristic values,and the method of using the characteristic value is of positive reference value in the simple and rapid identification oil type.