Design And Implementation Of Online Near-Infrared Quality Surveillance System In Key Productive Processes Of TCM

Quality control of traditional Chinese medicine injections is the important and difficult research topic in the construction of modernization of Traditional Chinese Medicine (TCM). It is difficult to guarantee the stable quality of TCM injection because of the traditional extraction using the experience physical parameters (for example, set decoction time, temperature, pressure, etc) to control the extraction process, which is influent by the complex components of TCM and its unstable qualities as a result of different batches and fluctuations during productive processes. In order to improve this situation, this paper designed and developed an online near-infrared quality surveillance system of TCM, which can be used in the key extraction processes of TCM injection, such as the processes of water extraction, ethanol extraction and column chromatography, in order to monitor the dynamic variation information of medicine extract. According to the information we can flexibly adjust the extraction scheme and optimize the productive technology so as to obtain the optimal extract in order to guarantee the homogeneous, stability and controllability of the quality of TCM injection, and achieve the goal of energy conservation and emission reduction, which can improve the benefit of the enterprise.The main research content of this paper includes the following aspects:1. Considering the finiteness of the Chinese Medicine components and the high dimensionality of its spectral data, according to the basic idea of manifold learning algorithm this thesis proposed LPP-PLS(Local Preserving Projection-Partial Least Square) spectral analysis modeling new method in order to find the low dimension manifold structure characteristics existing in the spectrum data. From the results of the experiments of building gardenia spectrum analysis model with this new method we can verify its effectiveness and advantage after comparing with other model building algorithms.2. Considering the multiple correlation and superfluous information existing in the spectral data, this paper proposed a new wavelength selection algorithm—Experience Interval Windows Moving Related Coefficient Division Combination ? ? Algorithm. From the results of the experiments of building isatis tinctoria spectrum analysis model with this new method we can verify its effectiveness and advantage after comparing with other model building algorithms.3. Through the analysis of existing chemometrics software insufficiency and the fusion of existing spectral analysis modeling method, according to the need of project implementation, this thesis researched and developed a new chemometrics software which can not only provide both offline analysis and online predicting, but also provide the convenience for users to establish the optimal model by the way of an automatic optimization guide. By the way of application practice and comparing with the OPUS software, author proved the effectiveness and advantage of this chemometrics software.4. Based on the analysis of the production spot information of project implementation project, the author designed and developed a main-control software for the extraction process. This software has many useful functions, such as communication with auto-control system and online near-infrared spectrometer, collecting spectrum, controlling preprocessing samples system, displaying the status of productive process, managing spectrum and model, etc., in order to trace the producting processes and supply the scientific theoretical material for enterprise users to flexibly adjust the productive scheme and optimize the productive technology.