摘要
粮食质量检测中粗淀粉是耗时较长的检测项目,直接影响着生产中的投料量及出酒率。传统检测方法检测一份粮食样品的粗淀粉则需要4h,不能及时完成粮食的入库影响生产工作,检测样本耗时长也限制着不能全面检测样品。采用近红外光谱技术结合偏最小二乘法,建立和优化了同时检测5种酿酒原料粗淀粉的预测模型,在光谱预处理方法为一阶导数和多元散射校正(MSC)、波数为5300cm^(-1)-6000cm^(-1)和6200cm^(-1)-8000cm^(-1)条件下获得酿酒原料水份的最优预测模型。该模型交叉验证误差均方根(RMSECV)、决定系数(R')分别为0.83%和0.9487。随机选取验证集样品对模型预测能力进行验证,通过与传统分析方法检测结果进行比较,平均相对误差为0.82%,结果表明该模型具有较好的预测能力,可用于5种酿酒原料粗淀粉的检测。
The determination of raw starch content in grain quality discrimination usually cost relatively long time and thus the result of the determination could directly influence the loading rate and the liquor output rate. The traditional method of raw starch content determination usually costs for about 4 hours, which could delayed the grain in-taking and make it impossible for overall determination of the sample.Near infrared spectroscopy with partial least square was used to establish and optimize the model of raw starch content determination of 5 liquor-making raw materials simultaneously. The best optimized model of water prediction in raw material was obtained when first-order derivative spectrum prepares, multiple scatter correction(MSC) and a wavenumber of 5300cm^(-1)-6000cm^(-1)and 6200cm^(-1)-8000cm^(-1) were used. The root mean square error of cross validation(RMSECV), the coefficient of determination(R^2) was 0.83% and 0.9487 separately. The predictive ability of the model was evaluated by random sample and the results showed that the average relative error was 0.82% compared with the conventional method, which means the model showed a good predictability and could be qualified to detect raw starch content in 5 liquor-making raw materials.
出处
《酿酒》
CAS
2016年第2期64-67,共4页
Liquor Making
关键词
近红外光谱技术
酿酒原料
粗淀粉
模型
Near infrared spectroscopy
liquor-making raw materials
raw starch
model