| 郑少帅，张倩文，徐一川，张广杰，席元庆，彭丹.基于低场核磁共振技术的大豆煎炸油理化指标
跨食材通用模型的构建研究[J].中国油脂,2026,51(8):.[ZHENG Shaoshuai, ZHANG Qianwen, XU Yichuan, ZHANG Guangjie,
XI Yuanqing, PENG Dan.Construction of cross-ingredients universal models for physicochemical indicators in soybean frying oil based on low-field nuclear magnetic resonance technology[J].China Oils and Fats,2026,51(8):.] |
| 基于低场核磁共振技术的大豆煎炸油理化指标
跨食材通用模型的构建研究 |
| Construction of cross-ingredients universal models for physicochemical indicators in soybean frying oil based on low-field nuclear magnetic resonance technology |
|
投稿时间：2025-08-07 修订日期：2026-02-05 录用日期：2025-08-29
出版日期：
|
| DOI：10.19902/j.cnki.zgyz.1003-7969.250363 |
| 中文关键词: 大豆油 煎炸 低场核磁共振 极性组分 酸值 通用模型 |
| 英文关键词:soybean oil frying low-field nuclear magnetic resonance polar components acid value universal model |
| 基金项目:河南省科技攻关项目（242102320278）；河南省A类专业创建建设专项（HN-HautFood-100）；河南工业大学“双一流”本科生科技创新能力提升专项项目（HN-HautFood IAEG-012) |
|
|
| 摘要点击次数: 3 |
| 全文下载次数: 0 |
| 中文摘要: |
| 旨在为快速判定煎炸油品质与废弃点提供技术支持，以3种食材（薯条、鸡块和豆腐）为煎炸对象，系统分析了不同食材煎炸过程中大豆煎炸油理化指标（极性组分含量和酸值）及其低场核磁共振（LF-NMR）信号的变化规律，并基于LF-NMR技术结合化学计量学方法，对煎炸油理化指标跨食材通用模型建立方法进行了优化，探究了通用模型应用的可行性。结果表明：随着煎炸时间的延长，大豆煎炸油的理化指标（极性组分含量和酸值）和LF-NMR横向弛豫曲线衰减速率均呈上升趋势，且其理化指标与单组分弛豫时间具有较好的相关性（R2>0.87）；极性组分含量和酸值通用模型的最佳建模条件为建模数据选择1~300 ms的横向弛豫曲线数据，预处理方法分别为标准正态变量变换（鸡块-豆腐模型为正交信号校正法）和正交信号校正法，建模方法均为支持向量机法，煎炸薯条-鸡块、鸡块-豆腐、薯条-豆腐和薯条-鸡块-豆腐的煎炸油的极性组分含量通用模型的预测均方根误差分别为1.491 1、1.747 3、2.567 7和1.632 8，平均相对误差均小于11%，而酸值通用模型受食材类型影响较大。综上，大豆煎炸油极性组分含量的通用模型预测性能稳定，基本满足实际检测需求，而酸值则需建立不同食材下的独立模型。 |
| 英文摘要: |
| To provide technical support for rapid evaluation of frying oil quality and determination of its discard point, the changes in physicochemical properties (total polar components, TPC; acid value, AV) of soybean oil and their low-field nuclear magnetic resonance (LF-NMR) signals were systematically analyzed during the frying of three different ingredients (fries, chicken nuggets, and tofu). Based on LF-NMR technology combined with chemometric methods, the construction approach of a cross-ingredients universal model for predicting the physicochemical indicators of frying oil was optimized, and the feasibility of applying such a universal model was explored. The results showed that as the frying time prolonging, the physicochemical indicators (TPC and AV) and the decay rate of the LF-NMR transverse relaxation curve of soybean oil both exhibited an upward trend. Moreover, the physicochemical indicators were well correlated with the relaxation time of the single component (R2>0.87). The optimal modeling conditions for the TPC and AV universal models were as follows: the transverse relaxation curve data in the range of 1-300 ms were used for model development; the preprocessing methods were standard normal variate transformation (with orthogonal signal correction for the chicken nuggets-tofu model) for TPC, and orthogonal signal correction for AV; and the modeling method was the support vector regression (SVR). The root mean square errors of prediction (RMSEP) for the TPC universal models of fries-chicken nuggets, chicken nuggets-tofu, fries-tofu, and fries-chicken nuggets-tofu were 1.491 1, 1.747 3, 2.567 7, and 1.632 8, respectively, with average relative errors less than 11%. On the other hand, the AV universal models were significantly affected by food material types. In conclusion, the universal model for predicting TPC of soybean frying oil demonstrates stable predictive performance and essentially meets the requirements for practical detection, whereas for the AV, independent models need to be established for different ingredients. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|
