基于改进小波神经网络的极谱法多金属离子浓度检测信号的在线解析  

Online analysis on polarographic detection signal of multi-metal ion concentrations based on improved wavelet neural networks

作　　者：王雅琳[1] 黄凯华[1] 黄天红[1] 周晓君[1] 阳春华[1] 

机构地区：[1]中南大学信息科学与工程学院,湖南长沙410083

出　　处：《中南大学学报（自然科学版）》2016年第1期100-107,共8页Journal of Central South University:Science and Technology

基　　金：国家自然科学基金资助项目(61273187);国家科技支撑计划项目(2012BAF03B05);教育部博士点基金(优先发展领域)资助项目(20110162130011)~~

摘　　要：针对极谱法实现锌湿法冶炼过程多金属离子浓度同时检测时所得信号存在重叠峰的问题,提出一种基于改进小波神经网络的多金属离子浓度极谱检测信号在线解析方法。首先,采用离散小波变换求取极谱检测信号的一阶导数,从而提取出极谱信号的特征点作为小波神经网络的输入;然后,提出一种改进的状态转移算法优化小波神经网络参数,实现基于小波神经网络的多金属离子浓度同时测定信号的离线建模和在线解析,提高多金属离子浓度同时测定的检测精度。以实际锌、钴极谱重叠信号为例进行验证。研究结果表明:所提出的方法针对锌质量浓度和钴质量浓度的测定结果优于传统的曲线拟合和基于BP神经网络的方法所得结果。For solving the overlapping peaks problem in multi-component detection of zinc hydrometallurgical process, an online analysis method for polarographic detection signal of multi-metal ion concentrations was proposed based on the improved wavelet neural network. Firstly, the first derivative of polarographic signal was obtained through the discrete wavelet transform, and consequently, the correspond feature points were obtained as the input of wavelet neural network based on the original signal and the first derivative of polarographic signal. Secondly, an improved state transition algorithm was proposed to optimize the parameters of wavelet neural network（WNN）, and then the optimized WNN was adopted to describe the relationship between those feature points and the multi-metal ion concentrations so that it could be used to analyze online the polarographic detection signal of multi-metal ion concentrations. The method was verified by the actual polarographic overlapping peaks signal of zinc and cobalt. The results show that the proposed method is superior to those of the conventional curve fitting and the BP neural network algorithm.

关 键 词：极谱曲线 多金属离子浓度 小波神经网络(WNN) 状态转移算法(STA) 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]