神经网络用于光度法同时测定多种重金属离子

采用可见分光光度法,通过构筑16-5-4多目标神经网络模型实现同时测定溶液中Cd2+、Pb2+、Cu2+、As3+的含量。实验以4-(2-吡啶偶氮)-间苯二酚(PAR)作为显色剂,采用"多因素多水平可视化设计法"设计样本,在4种组分可见吸收光谱严重重叠的390~480 nm范围内,选取16个特征波长处的吸光度作为输入信号,应用"留二法"原则训练BP网络。网络准确预测了结果,Cd2+、Pb2+、Cu2+、As3+的平均回收率分别为100.10%、100.03%、100.09%、99.99%,测定结果的相对标准偏差分别为0.18%、0.12%、0.26%、0.13%,达到了4种组分含量同时测定的目的。

Simultaneous Determination of Heavy Metal Ions by Spectrophotometry with Neural Networks

The contents of Cd²⁺, Pb²⁺, Cu²⁺ and As³⁺ in solution were measured simultaneously by visible spectrophotometry through constructing 16-5-4 model of neural network. The experiments used multifactor & multilevel Visual Design to obtain optimal sample, with 4-(2-pyridylazo)-resorcinol(PAR) as color reagent. In the range of 390-480 nm where the visible absorption spectrum of four components was serious overlap, the absorbance at 16 characteristic wavelengths was selected as the input signal to train BP networks in the principle of leave-two out. The trained networks accurately predicted the results and the average recovery rates of Cd²⁺, Pb²⁺, Cu²⁺, As³⁺ were 100.10%, 100.03%, 100.09%, 99.99%, and the relative standard deviations of prediction were 0.18%, 0.12%, 0.26%, 0.13%. The simultaneous determination of four components was well completed.