近红外光谱分析技术快速测定石粉中钙含量的研究

Research on rapid determination of calcium content in stone powder using near infrared spectroscopy

石粉是一种无机物,是畜禽日粮中钙的重要来源之一。通过研究利用近红外光谱技术(NIRS)建立石粉钙含量定量模型,以期建立一种快速可靠的石粉钙含量测定方法,同时为其他无机饲料样品的近红外分析提供参考。研究共收集248个石粉建模样品,并进行近红外光谱采集和湿化学检测,利用建模软件建立了石粉钙近红外定量模型,另外随机收集21个盲样,对模型进行盲样验证。结果显示:(1)石粉钙近红外预测模型决定系数R^(2)为85.28,交叉验证均方根误差RMSECV为0.308,残留预测偏差RPD为2.61,说明模型预测能力较好。(2)盲样验证结果显示,模型可以有效识别出异常报警样品,20个正常样品的湿化学值与预测值的绝对误差在0.01%~0.56%,验证集标准偏差RMSEP为0.229,平均偏差Bias为0.077,残留预测偏差RPD为3.98,各项数据表明模型验证效果较好。

Stone powder is a kind of inorganic substance that serves as an importance source of calcium in livestock and poultry rations.In this paper,a quantitative model for determining calcium content in stone powder is proposed based on near infrared spectroscopy(NIRS)technology in the hope to find out a way of determining calcium content in stone power with high precision and reliability,which may offer reference for NIRS-based analysis of other inorganic feed samples.In the study,altogether 248 stone powder modeling samples were collected to undergo NIRS and wet chemical tests.A quantitative model of near infrared was established using modeling software to determine the calcium content in stone powder.Furthermore,21 blind samples were randomly collected to verify the model.The results suggest:(1)the determination coefficient of the proposed model R^(2) is 85.28,the root mean square error of cross-validation(RMSECV)is 0.308,and the residual prediction deviation(RPD)is 2.61,which collectively prove the competent predicting performance of the proposed model;(2)as shown by the blind sample verification results,the model could effectively identify abnormal alarm samples,because for the 20 normal samples,the relative errors between wet chemical values and predicted values are within 0.01%—0.56%,the root mean square error of prediction(RMSEP)is 0.229,the average bias is 0.077,and the residual prediction deviation(RPD)is 3.98.All the data have testified the competent verification effect of the proposed model.