涂层
主成分分析
材料科学
数学
计算机科学
统计
复合材料
作者
Qing Tao,Jin Zheng-Ji,Luo Xiao-jian,Xiao-Yong Rao,H E Yan
出处
期刊:PubMed
[National Institutes of Health]
日期:2020-10-01
卷期号:45 (19): 4625-4632
标识
DOI:10.19540/j.cnki.cjcmm.20200810.301
摘要
To conduct endpoint determination of tablet coating and analyze coating process by near-infrared spectroscopy(NIRS). Firstly, the layered k-fold cross validation was used to discuss the optimal combination of spectral interval and preprocessing method, and the optimal result was used as the modeling condition. Secondly, in above condition, the qualitative model was proposed based on the throught of principal component analysis(PCA) and confidence interval, this model and the conformity test model were established respectively in this study, and the discrimination performance of coating eligibility was calculated. Then, the internal cross validation was used to obtain the coating end-point conforming rate as the threshold for determining the endpoint of tablet coating. Finally, the results of predicted conforming rate and discrimination parameters for the test samples of different batches were calculated with the above models, which were then further used for end-point detection and process analysis. As compared with the conformity test, this method proposed in this study was more accurate and stable in determining the coating conformity, and the coating endpoint can be accurately determined when 95% of the eligibility rate was used as the discriminant threshold. Meanwhile, in the process analysis, the change trend of the process parameters(model discrimination parameters, prediction conforming rate) was basically consistent with that in the conformity test. The results indicate that the method proposed in this study has a good and stable performance, which can be used to determine the endpoint of coating and analyze the process. It is of great significance to reduce the difference between batches and improve the product consistency. Meanwhile, this research method also lays a foundation for the future researches by online near infrared measurements.
科研通智能强力驱动
Strongly Powered by AbleSci AI