论文部分内容阅读
建立基于仪器分析方法的质量标准是中药进入国际市场的必要条件。近红外光谱技术以其能够反映样品的多种信息、易于在线应用的优势,应用于中成药生产的在线质量监控,可以提高中成药的质量控制标准,加快中药现代化的进程。但在近红外光谱检测中存在着各成分谱图重叠严重,光谱信息冗余,特征吸收区域不明显的问题,需要对采集到的波长进行优选,以达到提高模型预测精度和简化模型的目的。从近红外光谱方法测量中药有效成分的基础研究入手,以冰片含量的检测为例,尝试采用遗传算法与模拟退火算法结合的模拟退火遗传算法及物理意义相对明确的多链逐步选择法对校正模型的波长进行优选。结果表明,波长选择的方法可以使模型采用的波长数减少的同时提高预测精度,波长选择最多可将波长数减少84%,预测精度提高47.6%。
Establishing quality standards based on instrument analysis methods is a necessary condition for Chinese medicines to enter the international market. Near-infrared spectroscopy technology can be applied to the online quality monitoring of Chinese patent medicine production with its advantages of being able to reflect a variety of information of samples and being easy to apply online. It can improve the quality control standards of Chinese patent medicines and accelerate the modernization of traditional Chinese medicine. However, in the near infrared spectroscopy detection, there is a problem that the spectrums of the components overlap, the spectral information is redundant, and the feature absorption area is not obvious. Therefore, the collected wavelengths need to be optimized in order to improve the model prediction accuracy and simplify the model. The basic research on the measurement of effective components of traditional Chinese medicine by near-infrared spectroscopy was started. Taking the detection of borneol content as an example, simulated annealing genetic algorithm combining genetic algorithm with simulated annealing algorithm and multi-chain gradual selection method with relatively clear physical meaning were used to calibrate the model. The wavelength is preferred. The results show that the wavelength selection method can reduce the number of wavelengths used in the model and improve the prediction accuracy. The wavelength selection can reduce the number of wavelengths by up to 84% and the prediction accuracy by 47.6%.