复杂环境下基于近红外光谱的挂面生产过程水分快速检测方法研究
Research on Rapid Moisture Detection Method in Noodle Production Process Based on Near Infrared Spectroscopy Under Complex Environment
何锦漪 1林颢 1齐雅静 1何文森 1徐斌1
作者信息
- 1. 江苏大学食品与生物工程学院,镇江 212013
- 折叠
摘要
水分含量是面条的重要品质指标,水分快速无损检测对挂面企业提质增效有重要意义.研究利用近红外技术分别检测挂面生产过程面片与挂面水分,根据不同外界环境下近光谱的波动探究近红外检测面片与挂面水分的影响因素,采用Si-PLS变量筛选和CARS竞争性自适应重加权2种算法筛选出于水分相关的特征波长,用线性回归与偏最小二乘法2种方法建模并比较模型性能,所建立最佳模型预测误差为0.011 6,相关系数R为0.999 2.对于实际环境中温度变化导致光谱波动的情况,采用多元线性回归和支持向量回归2种方法分别建立温度补偿模型,所建立的补偿模型误差为0.087 8,相关系数R为0.989 5,证明了在挂面生产过程中利用近红外光谱技术进行水分快速检测的可行性.
Abstract
Water content is an important index of noodle quality,and rapid non-destructive testing of water content is of great significance for maintaining noodle quality.In this paper,near-infrared technology was used to detect the moisture content of noodles and dry semi-finished noodles in the production process,and the influencing factors of near-infrared detection of the moisture content of noodles and dry semi-finished noodles were explored according to the fluctuations of near-spectrum under different external environments.Two algorithms,Si-PLS vari-able screening and CARS competitive adaptive reweighting,were used to screen the characteristic wavelengths related to water.Linear regression and partial least square method were used to model and compare the model performance.The prediction error of the best model was 0.011 6,and the correlation coefficient R was 0.999 2.For the spectrum fluctuation caused by temperature change in the actual environment,multiple linear regression and support vector re-gression were used to establish temperature compensation models,respectively.The error of the compensation model established was 0.087 8,and the correlation coefficient R was 0.989 5.The feasibility of using near infrared spec-troscopy to detect moisture quickly in the process of noodle production was proved.
关键词
挂面/水分/近红外光谱/温湿度补偿模型Key words
vermicelli/hydration/near infrared spectrum/temperature and humidity compensation model引用本文复制引用
基金项目
苏州市农业科技创新项目(SNG2022040)
出版年
2024
NETL
NSTL
万方数据