Intelligent Resolution System For Complicated Multi-component Chemical System

The aim of this thesis is to make new intelligent expert system, which can be used to cope with complicated multi-component chemical systems. Researches on new chemometric algorithms, knowledge discovery in database and software development of the system are presented in this thesis. So far, as the first stage of the whole system, part of the works has been finished as follows: 1. The research on fundamental chemometric algorithms Works on this aspect include the researches on automatic deconvolution of two-way data from hyphenated chromatography and the novel library search methods for mass spectrum of pure substance and/or mixture, respectively. In the research on automatic deconvolution method, we put forward a new method to select key spectra and then to determine chemical rank automatically by comparing the similarity of the potential key spectra. Then, based on the determined real key spectra, we put forward a new algorithm, the stepwise key spectrum selections, which can implement resolution of overlapped chromatographic profiles automatically. The library search of mass spectrum of pure substance is focused on the similarity search. By emphasizing the substructure similarity, we put forward a new algorithm to calculate the similarity index, which can give a rather good result in similarity search. Library search of mass spectrum of a mixture is used in the situation when pure spectrum can not be obtained, such as high boiling point substance or overlapped chromatographic peaks. An index called weighted spectrum existence index is put forward to select potential substance from mass library. Then, non-negative least square regression is used to determine the real substance. III (C- 2. Knowledge discovery in database Research on this aspect has two goals, the first is to accumulate enough knowledge for the knowledge database of the intelligent expert system; the second is to make a primary approach on how to obtain useful chemical knowledge from a big database. By statistic analysis of NIST 62 mass library, we firstly find that the differences between the measured (M+1)/M ratios and the calculated ones fit logarithm normal distribution. Our works on this aspect shows that a wide and deep inquiry in chemical database is a meaningful work not only for the proof of the existing knowledge, but also for the discovery of new chemical knowledge. 3. Software development of the intelligent system The work in this part is to construct the body of the intelligent expert system. Because it is a long period of work to develop such an intelligent system, the whole system is divided into two parts ?curve resolution system and mass spectrum expert system. The curve resolution system integrates methods presently available in resolution of two-way data from hyphenated chromatography. It can either be an independent software itself, or be the forepart of the intellig鏽t expert system. For the effectiveness in run time and the availability for future transferring of codes, we take C++ object orientated technology in the developing process. We also develop a matrix class, which integrates matrix calculations, to simplify the codes and enhance the effective o...