|
| 氢谱核磁共振与随机森林算法联用快速检测大豆油过氧化值 |
| Rapid determination of peroxide value of soybean oil using 1H NMR spectroscopy combined with random forest algorithm |
| |
| DOI： |
| 中文关键词: 氢谱核磁共振 过氧化值 随机森林算法 定量预测 大豆油 |
| 英文关键词:1H nuclear magnetic resonance (1H NMR) peroxide value random forest algorithm quantitative prediction soybean oil |
| 基金项目:广东高校省级重点平台及重大科研项目（青年创新人才类）（2017KQNCX250）；北京市科技计划（Z1711000 01317004） |
|
| Hits: 859 |
| Download times: 324 |
| 中文摘要: |
| 基于氢谱核磁共振技术与随机森林算法联用，建立了大豆油过氧化值的快速检测方法。通过比对大豆油氢谱核磁共振数据，分析鉴别出5种氧化产物，分别为（Z,E）型和（E,E）型共轭烯烃类氢过氧化物、直链烷醛、（E,E）-共轭二烯醛、（E）-共轭烯醛。根据氢谱核磁共振测定了油样氧化产物的含量，同时采用比色法测定了过氧化值，采用交叉验证的方式构建了大豆油过氧化值快速检测的定量预测模型。结果表明：建模参数决策树的数量（ntree）为700、决策树的最大深度（max_depth）为5、决策树最小样本节点数量（min_samples_split）为4时，所建的预测模型标准均方根偏差（SRME）为2.851%，模型过氧化值预测值与比色法（标准方法）测定结果相关系数(R2)为0.943，模型表现出良好的稳定性和预测精度。氢谱核磁共振与随机森林算法联用可快速在线检测大豆油过氧化值，为油脂品质指标监测提供技术支持。 |
| 英文摘要: |
| A rapid detection method of peroxide value of soybean oil was established by 1H nuclear magnetic resonance (1H NMR) spectroscopy combined with random forest algorithm. By comparing the 1H NMR data of soybean oil, five kinds of oxidation products were identified, such as (Z,E)-conjugated form, (E,E)-conjugated form, E-alkenals, (E,E)-alkadienals and n-alkanals. According to the content of five kinds of oxidation product in soybean oil samples determined by the 1H NMR data and the peroxide value of soybean oil samples detected by colorimetry, the quantitative prediction model for the rapid detection of peroxide value was constructed by cross-validation. The results showed that the model could present the stronger learning ability and accuracy when the optimization parameters of ntree 700,max_depth 5 and min_samples_split 4, and the model’s coefficient(R2) and root mean square error (RMSE) were 0.943 and 2.851%, respectively. The method based on 1H NMR spectroscopy combined with random forest algorithm could determine peroxide value rapidly, which provided scientific support for monitoring and evaluating the oil quality online. |
| 查看全文 View/Add Comment Download reader |
| Close |
|
|
|
