BP人工神经网络分光光度法同时测定电池厂废水中镉和镍

Simultaneous determination of cadmium and nickel in waste water from battery factory with spectrophotometry coupled with BP neural network

研究BP神经网络和分光光度法结合的镉、镍含量同时测定方法。在Na2B4O7缓冲溶液中(pH=10.2),邻羧基苯基重氮氨基偶氮苯(o-CADD)与镉、镍发生灵敏的显色反应,形成红色配合物,最大吸收峰分别位于525nm和540nm。利用镉和镍配合物吸收光谱上的差异,对16个不同镉、镍离子浓度的混合液组成的校正集进行训练。通过选择测定波长间隔,网络隐含层神经元数,训练函数等,优化了网络。并对3组预测集验证,相对误差的绝对值小于5%。该方法用于电池厂废水中痕量镉镍的同时测定,加标回收率在104%～95.5%之间,表明方法具有较高的准确性。

The spectrophotometry coupled with BP neural network was developed for determination of trace cadmium and nickel in the laboratory.In the presence of borax buffer solution（pH10.2）,cadmium（II）and nickel（II）reacts with o-CDAA to form red complex with maximum absorption at 525 and 540 nm,respectively.Based on the difference in absorption spectrometry of two above complex,sixteen calibration samples which were composed by cadmium and nickel were prepared to train the network.Its parameters of the network were optimized by choosing wavelengths,number of nerve cell in the hidden layers and the training function.By testing prediction samples,the absolute value of the prediction samples＇ relative error was less than 5%.The method has been applied to determine trace cadmium and nickel in waste water from battery factory.The recovery of Cd2＋,Ni2＋was 104%~95.5%,this indicates the method has good accuracy.