Madaline网络分离重叠光谱峰在灰色体系分析中的应用  

Resolution of Overlapped Spectra with Madaline and Its Application to Gray System

作　　者：方慧生[1] 吴玉田[1] 陈斌[1] 

机构地区：[1]第二军医大学药学院药物分析教研室,上海200433

出　　处：《分析化学》1999年第8期875-880,共6页Chinese Journal of Analytical Chemistry

摘　　要：将Madaline网络分离重叠光谱峰法应用于组成范围已知的灰色体系进行分析.模拟数据研究表明:当测量误差在[-0.005,+0.005]范围内,且所含组分峰形相同但峰位不同时,最低检测限是峰位差不小于1.5nm.15个实际样品分析表明:本法能给出比较满意的分析结果.它为组成范围已知的灰色体系分析提供了一种新途径.The method of resolution overlapped spectra with Madaline network is used to determine gray system whose range of possible coexisting species has been known, but the exact number and identities are unknown. The computational process of simulation data shows that the more overlapped spectra cost the more time of training procedure. Also, the results giving the condition of this method are that the minimum difference of peaks' location whose shapes are same, is 1.5 nm while the range of measuring error is from -0.005 to 0.005. The determination of 15 samples shows that it can give satisfactory results. It provides a new method to determine the above gray system. If all of the known compounds spectra in the world have been determined and their own output layer defined, the training procedure of Madaline will give the connection weight (W) . However, a part of mixture with unknown interference may be transformed into the above mentioned gray system and W will give their exact components. Therefore, the method probably can solve some mixture with unknown interference.

关 键 词：多重自适应 线性 神经网络 灰色体系 

分 类 号：O651[理学—分析化学]