论文部分内容阅读
目的:建立羧甲司坦片快速定量分析模型,并探讨近红外定量分析模型中校正集样本数的选择,提高模型的预测能力。方法:收集全国19家生产企业共140批羧甲司坦片样品,采集近红外光谱图,以聚类分析法分别将总样本集分成120类、115类、111类、106类、102类、96类、91类、86类、81类、76类、71类、66类、61类、57类、52类、47类和42类,从每一类中选出一个样本作为训练集,逐一建立模型,利用OPUS光谱软件中定量2方法选择最优建模参数,建立羧甲司坦片通用型定量分析模型,并进行方法学验证。结果:选取61个样本建立羧甲司坦片定量模型,模型交叉验证均方根误差(RMSECV)为1.17%,检验集验证均方根误差(RMSEP)为1.08%,相关系数为0.995;模型能够快速准确预测片剂中羧甲司坦的质量分数,范围为22.28%~85.15%。结论:选取合适数量的样本作为模型的校正集,可有效提高模型预测能力;本文所建立的羧甲司坦片通用型快速定量分析模型准确度好,耐用性强。
OBJECTIVE: To establish a rapid quantitative analysis model of carbocisteine tablets and explore the choice of the number of calibration samples in the quantitative analysis model of near infrared to improve the predictive ability of the model. Methods: A total of 140 samples of Carboxymethyl Stearate were collected from 19 manufacturing enterprises in China. NIR spectra were collected. The total samples were divided into 120 categories, 115 categories, 111 categories, 106 categories, 102 categories, 96, 91, 86, 81, 76, 71, 66, 61, 57, 52, 47 and 42 from each category as a training set, one by one The model was established, and the optimal modeling parameters were selected by Quantitative 2 method in OPUS spectral software to establish the universal quantitative analysis model of Carboxystate, and the methodological verification was carried out. Results: The quantitative model of Carbamate was selected from 61 samples. The RMSECV of the model was 1.17%. The root mean square error of validation (RMSEP) was 1.08% and the correlation coefficient was 0.995. The model could The mass fraction of carbocisteine in tablets was quickly and accurately predicted, ranging from 22.28% to 85.15%. Conclusion: Selecting the appropriate number of samples as the calibration set of the model can effectively improve the predictive ability of the model. The universal rapid quantitative analysis model of carcotliptin established in this paper has good accuracy and durability.