低场核磁共振技术结合化学计量学建立鲍鱼水分和脂肪含量预测模型

Prediction Models Established of Moisture and Fat Content by Low Field NMR Combined With Chemometrics in Abalone

为研究鲍鱼的无损检测技术,提出了一种基于低场核磁共振技术的快速、无损预测鲍鱼水分和脂肪含量的新方法,采用CPMG序列测定横向弛豫时间T2,结果表明:鲍鱼肉中含有3种不同组分水,分别是结合水(T2b)、肌原纤维内水(T21)和肌原纤维间隙水(T22)。并结合化学计量学建立2种关于鲍鱼水分和脂肪含量的预测模型:校正集(Calibration)用于预测样品水分和脂肪含量,而交互验证集(Validation)被用来验证预测结果的稳健性和预测标准误差(SEP)值。PCR和PLSR预测模型相关系数R2均大于0.9。然而PLS预测模型则给出了最好的预测结果:在水分预测模型中R2val=0.993 5,SEP=0.140 1;脂肪含量预测模型中R2val=0.967 5,SEP=0.265 0。由此表明,2种预测模型都具有评价鲍鱼品质的良好潜力,并且PLS预测模型的精准性和稳健性优于PCR预测模型。

In order to study abalone nondestructive testing technology, this paper proposes a fast, non-destructive prediction abalone new methods of moisture and fat content,based on the technology of low field nuclear magnetic resonance （NMR）.This research adopts the CPMG sequence determination of transverse relaxation time T2,The result showed that the abalone meat contains three different components of water,respectively ,combined water （T~） ,myofibril inland waters （T21） and myofibril interstitial water （T22）.Combined with chemometrics to establish two kinds of prediction model about abalone moisture and fat content.Grade correction （Calibration） model is used to predict the sample moisture and fat content and interaction level verification （Validation） model is used to verify the robustness of results and standard error （SEP） value.PCR and PLSR model correlation coefficient R2 were greater than 0.9.However PLS forecast model gives the best results：R~=0.993 5 in moisture prediction model, SEP=0.140 1 ;Fat content prediction model of R2~=0.967 5,SEP=0.265 O.Thus the two kinds of prediction model has good potential of evaluation quality of abalone, and precision and robustness of the PLS model is better than that of PCR prediction model.