基于低场核磁共振技术构建韧性饼干中水分含量无损定量预测模型

目的基于低场核磁共振技术(low-field nuclear magnetic resonance, LF-NMR)弛豫特性建立韧性饼干水分含量的预测模型,探讨LF-NMR无损快速预测韧性饼干水分含量的可行性。方法采用LF-NMR对韧性饼干的CMPG(Carr-Purcell-Meiboom-Gill)序列信号进行采集,利用直接干燥法检测韧性饼干中水分的实际含量,建立多元回归分析、K最邻值回归分析(K-nearestneighbor,KNN)和高斯过程回归(Gaussianprocess regression,GPR)模型。结果 3种模型中KNN回归模型预测集决定系数最大为0.9932,均方根误差最小为0.2542,因此韧性饼干中水分与低场核磁弛豫特性的分析中, KNN回归分析的模型最优。结论使用LF-NMR分析仪预测韧性饼干中水分含量是一种无损、快速的方法,对于韧性饼干样品中水分快速预测建模结果良好,因此利用低场核磁共振技术快速预测韧性饼干样品中的水分含量方法可行。

Construction of non-destructive quantitative prediction model of the moisture content of tough biscuits based on low-field nuclear magnetic resonance

Objective The characteristics of low-field nuclear magnetic resonance (LF-NMR) relaxation and the moisture content of the biscuits was detected through analytical methods regression analysis, to explore whether it is feasibility that the LF-NMR technology predict moisture content of semi hard biscuits. Methods We used LF-NMR to collect the CMPG sequence signal of biscuits and detected the actual content of the moisture in the biscuits using the drying method. The characteristics of LF-NMR relaxation and the moisture content of the biscuits was detected through multiple regression analysis, K-neighbor regression analysis and gaussian process regression analysis, to explore the LF-NMR in measuring moisture content of biscuits. Results The predictive set of KNN regression models in the three models has the largest decision coefficient is 0.9932, the smallest average square root error is 0.2542. The model of KNN regression analysis of moisture and LF-NMR relaxation characteristics in biscuits is optimal. Conclusion LF-NMR was a useful nondestructive and fast detection method of the moisture content prediction in semi hard biscuits. The rapid moisture detection model in the same type of semi hard biscuit samples are good. It is feasible of rapid detection of moisture content in biscuit samples by using LF-NMR.