论文部分内容阅读
目的建立近红外通用性模型,能对不同厂家吡嗪酰胺片的含量进行快速、无损地测定,有效监控其质量。方法采集9个浓度梯度的各3批自制样品及来源于20个不同厂家46批次的真实样品近红外漫反射光谱,并通过聚类分析方法确定校正集和预测集,考察不同预处理方法、谱段和光滑点数的影响,选择建立了最佳的吡嗪酰胺片的定量模型。结果 46个校正集样品经交叉验证建立校正模型,交叉验证均方根误差(RMSECV)为0.775,相关系数为99.4%;27个预测集真实样品的预测均方根误差(RMSEP)为0.962,预测值与真实值的相关系数为99.8%。预测值的平均回收率为99.9%(RSD为1.26%)。方法精密度RSD为0.84%(n=6),方法稳定性RSD为0.5%(n=5)。对6个厂家6批真实样品含量测定,相对误差均小于1.53%。结论所建立的定量模型能够对不同厂家不同规格的样品作出准确、快速的含量分析。
Objective To establish a near-infrared universal model that can detect pyrazinamide tablets from different manufacturers rapidly and non-destructively and monitor their quality effectively. Methods Three batches of self-made samples with nine concentration gradients and 46 real-time samples of near-infrared diffuse reflectance spectra from 46 batches of 20 different manufacturers were collected. The calibration sets and prediction sets were determined by cluster analysis. Different pretreatment methods, Spectral bands and smooth points, we chose to establish the best quantitative model of pyrazinamide tablets. Results The calibration model was established by cross validation of 46 calibration samples. The RMSECV was 0.775 and the correlation coefficient was 99.4%. The RMSEP of real samples from 27 prediction sets was 0.962. The predicted The correlation coefficient between the value and the true value is 99.8%. The average predicted recovery was 99.9% (RSD 1.26%). The accuracy of the method was 0.84% (n = 6), and the stability of the method was 0.5% (n = 5). 6 batches of 6 real samples were determined, the relative errors were less than 1.53%. Conclusion The established quantitative model can accurately and quickly analyze the content of different samples from different manufacturers.