摘要
为满足茶叶加工过程中对茶叶品质成分在线检测的迫切需求,基于近红外光谱技术,设计了茶叶品质在线检测仪.文章系统介绍了检测仪的整机结构、工作原理、主要部件和配套分析软件.采集不同杀青程度的杀青叶样品 280 个,采用偏最小二乘法(PLS)建立杀青叶咖啡碱、水浸出物、茶多酚、含水率和游离氨基酸含量的检测模型,并对模型进行内部交叉验证和外部验证.结果表明,含水率模型的校准标准误差(SEC)和交叉验证标准误差(SECV)分别是 0.44、0.47,两者偏差最小,为 0.03,校正相关系数(RC)为 0.92;咖啡碱模型的SEC和SECV分别是0.43、0.35,RC为 0.95;且两者的预测值与真实值间的线性回归系数均在 0.90 以上.含水率和咖啡碱模型外部验证的预测值与真实值之间的平均绝对偏差在 1.00%以内,分别是 0.68%、0.80%.设计的茶叶品质在线检测仪的稳定性和精度满足现场实时检测需求,为茶叶加工设备智能化提供了技术支撑.
Abstract
To meet the urgent need for online detection of tea quality components during tea processing,a tea quality online detection instrument was developed based on near-infrared spectroscopy technology.This paper systematically introduces the overall structure,working principle,key components,and accompanying analysis software of the detection instrument.280 samples of fixed leaves with different degrees of fixation were collected,and a detection model was established for caffeine,water extract,tea polyphenols,moisture content,and free amino acid content in fixed leaves using partial least squares(PLS).Internal and external cross validation of the model were conducted.The results showed that the standard error of calibration(SEC)and standard error of cross validation(SECV)for the moisture content model were 0.44 and 0.47,respectively,with a minimal difference of 0.03,and the calibration coefficient(RC)reached 0.92;For the caffeine model,the SEC and SECV were 0.43 and 0.35,respectively,and the RC reached 0.95;The linear regression coefficients between predicted values and actual values for both models exceeded 0.90.The external validation of the moisture content and caffeine models showed that the mean absolute deviation between predicted and actual values was within 1.00%,specifically 0.68%and 0.80%,respectively.In summary,the developed tea quality online detection instrument exhibits good stability and precision,meeting the needs of real-time field detection and providing technical support for the intelligentization of tea processing equipment.
国家科技期刊平台
NSTL
维普
万方数据