摘要：

将Kohonen神经网络与RBF网络相结合，对废水中吸收光谱严重重叠的钴、镍、钒三组分体系进行解析。利用Kohonen神经网络选择全谱特征波长，优化确定了RBF网络的结构和参数，使光度分析计算的校正模型的优化问题得到有效解决。分析结果表明，经Kohonen神经网络方法进行波长选择后，优化了RBF网络的输入并提高了其预测能力。

关键词: Kohonen神经网络, RBF网络, 钴, 镍, 钒

Abstract:

By combining Kohonen neural network with radial basis function(RBF) network,the seriously overlapping spectra of the three components of cobalt,nickel and vanadium in wastewater has been analyzed. The most informative wavelengths are selected from the full spectra,and the structure and parameters of RBF network are defined by optimization. As a result,the optimization problems in calibration model for the calculation of photometric analysis are solved effectively. The results prove that after using Kohonen network method for selecting the most informative wavelengths,the input of RBF network is optimized and the prediction ability is improved.

Key words: Kohonen artificial neural network, radial basis function network, cobalt, nickel, vanadium

中图分类号: 

• X832