多维数据分析法研究多组分同时测定

Simultaneous multicomponent determinations by means of multidi- mensional data analysis were studied in this project. Combination of chemistry, mathematics, statistics and computer science etc. was applied to resolve overlapping peaks and optimize analytical methods. Simultaneous multicomponent determinations were studed by our developed methods such as wavelet multiresolution analysis and partial least squares, wavelet-based latent variable regression, soft thresholding wavelet-based kernel partial least squares, radial basis function neural networks, Elman recurrent neural network, generalized regression neural networks. Combining application of some up-to-date techniques were focused specially. Three new methods (artificial neural networks with maximum likelihood principal component analysis, wavelet packet transform based Elman recurrent neural network method, wavelet packet transform based generalized regression neural network) were developed firstly and applied successfully to simultaneous multicomponent determinations. The developed methods were applied to the environmental water samples of our district in order to improve these methods further. 22 papers were published, in which one paper was published in Talanta and was indexed by SCI, 5 papers were published in national kernel journals, 15 papers were published by journal or publishing company, 7 papers were collected in the proceeding of international conference. This project has theoretical values and realistic meaning.

本项目拟研制小波数字滤波方法，正确扣除噪音及导入增秩方法确定组分数；编制不同的小波函数和小波变换程序，解析重叠峰，优化测定方法；并根据系统论，小波变换，多元分辨等理论研制伏安，动力分析，光度同时测定方法探讨多维数据分析法对多组分体系同时测定。同时对本地区没有预知组分数目，种类和浓度范围的环境水样进行测定。