Abstract: The influent ionic concentration of the second stage is an important factor for the operation optimization of cobalt removal purification process in zinc hydrometallurgy.With the difficulty in online measurement of the influent ionic concentration，a soft sensing model of ionic concentration is proposed based on the time series analysis of manual measurement results，using the Wavelet Analysis（WA） and Support Vector Machine（SVM）.The time series of ionic concentration are decomposed into multiple subsequences by using wavelet decomposition and these subsequences are reconstructed into each subspace in phase space.Then，the SVM model is built in every subspace and the parameters in SVM model are optimized by using the chaotic Particle Swarm（PSO） algorithm.Finally，the outputs of each subsequence model are synthesized as the soft sensing result of influent ionic concentration.The verified results of the production data show that the proposed model has a high precision and the samples with relative error less than 10% are up to 97.5%.It indicates that the online prediction precision can meet the technological requirement of practical production.

Key words: support vector machine, wavelet analysis, chaotic Particle Swarm Optimization（PSO）, ionic concentration, prediction online

摘要： 二段入口离子浓度是锌湿法冶炼二段净化除钴过程优化控制的一个关键因素。针对二段入口离子浓度在线检测困难的问题，提出了一种结合小波分析和支持向量机的离子浓度参数软测量方法，直接采用人工检测的历史数据直接建立离子浓度软测量模型。该方法采用小波分析方法将时间序列分解成具有不同频率特征的子序列。在相空间重构的基础上，利用最小二乘支持向量机建立各子序列估计模型，其中模型中的参数采用混沌粒子群算法进行优化选择。对各子序列输出重构合成得到最终的在线估计结果。应用工业现场数据的验证结果表明，所提模型具有较高的精度，相对误差小于10%的样本达97.5%，在线估计精度能够满足现场实际生产工艺要求。

关键词: 支持向量机, 小波分析, 混沌粒子群算法, 离子浓度, 在线估计