混合物定性定量分析的分枝定界算法

A Bound-Branch Algorithm for Qualitative and Quantitative Analyses of Mixtures

巳知被测体系可能存在的物种范围,但其确含何种物质及相应浓度必须经分析方可确定,实为分析化学实践中常见的一类分析体系。基于分枝定界原理,本文提出一种可直接对其进行定性定量解析的光谱多元分析算法,可显著减少计算量,却无丢失真实信息的风险。对实际紫外可见及荧光混合物光谱进行解析,均获满意结果。

For analysis of multicomponent mixtures using linear regression algorithms, a qualitative detection of the analytical sample must be made to ascertain the exact nature of coexisting species in the mixture. The most frequently occurring cases in analytical practice is that the range of possible coexisting species is known and a decision has to be made which species actually coexist in the sample. This paper describes a new chemometric method based on bound-branch approach to solve this problem which has been turned to the problem of selecting the optimal subset for regression. The computation time is much less than the enumeration procedure and there is no risk of missing the real combination of the coexisting components in the sample. A program written in True BASIC has been testified for UV and fluorescence spectrophotometric analyses of mixtures containing various possible coexisting organic compounds with satisfactory analytical results.