Study On Rapid Detection Of Resveratrol In Wine By Hyperspectral Technology And Fluidized Bed Enrichment

Resveratrol has widely biological activity and is a global research hotspot.It is the most important active substance in wine.To establish a rapid and efficient method for the detection of resveratrol in wine has broad application prospects.Hyperspectral imaging technology and fluidized bed enrichment technology were used, together in this paper to solve the problem of high detection limit. An enrichment apparatus was built.With spectroscopic analysis and chemometrics methods, detection models of resveratrol in wine were established.It aimed to provide references for trace analysis in hyperspectral technology.(1)A direct injection HPLC method for determination of resveratrol in wine was established.The effects of detection wavelength,solvent strength,mobile phase pH,column temperature, sample pH and other factors on the separation of wines were examined. It showed that R2 was 0.9999, linear range was 1～20mg/L; RSD of precision experiment, stability experiment and recovery experiment were 1.16%, 2% and 1.25% respectively; Comparison result among cultivars was:Cabernet Gernischt>Merlot>Cabernet Sauvignon>Chardonnay>Riesling,and resveratrol content in Cabernet Gernischt was up to 6mg/L.(2)The enrichment effects of resveratrol in wine by macroporous adsorption resin were explored Five resins,including HPD826,DA-201,AB-8,H103 and HPD600,were studied in this paper. H103 was screened out and its adsorbent kinetic property was studied.The adsorption conditions were optimized:wine velocity was 1.5BV/h, pH constant;the desorption conditions:water was used for elution first,then 80% ethanol was used at the velocity of 1BV/h. The results showed that the adsorption rate of H103 reached 89.3%.(3) H103 was applied in the the enrichment apparatus to enrich resveratrol from wine.The results showed that there were no resveratral in liquid effluent when the resin was enough with wine cycling twice.(4) Prediction models for resveratrol in wine by hyperspectral was established. Spectral images(900-1700nm)of resin were extracted.RN was selected as the best spectra preprocessing method after comparing five methods (MSC, SNV, SG-S, RN and QN)for models of resveratrol.Then six regression models based on PLSR, SVMR (LK-SVMR, PK-SVMR, RBF-SVMR and S-SVMR) and PCR were established.Correction models based on LK-SVMR and PLSR were selected to evaluate their accuracy and stability in prediction sets further. Finally, the PLSR model was chosen as the best model. The results showed that PLSR model achieved good prediction results,its R2p was 0.8528, RMSEP was 0.0360, R2C was 0.8783, and RMSEC was 0.0330.