Soft Sensor For Cobalt Ion Concentration Of The Cobalt Removal Purification Process With Arsenic Salt In Direct Leaching For Zinc Smelting

Cobalt ion is the impurity ion of greatest difficulties to be removed in the purification process of direct leaching for zinc smelting, and its concentration is one of the important indexes for the electrolyte. If cobalt ion concentration can’t meet the process requirements of the electrolyte, the cobalt ion will reduce current efficiency of electrolysis process and have negative effects on product quality, still, the unqualified purification solution have to be returned to re-purify. In this thesis, considering that the cobalt ion concentration in the cobalt removal purification process with arsenic salt can’t be detected on-line, we study a soft sensing model for predicting cobalt ion concentration on-line, providing guidance for cobalt removal purification process optimal control.(1). The crafts of cobalt removal purification process with arsenic salt in the direct leaching of zinc smelting is firstly introduced and the mechanism analysis of factors impacting cobalt removal efficiency are studied in this paper. Then the overall framework of the soft sensing model for predicting cobalt ion concentration is proposed.(2). Due to the difficulty in determining the suitable control limits for multivariate outlier detection with traditional principal component analysis method, an online outlier detection method based on principal component analysis and Bayesian classification is presented. Firstly, principal component analysis is used to calculate Q statistics with the training data collected in the normal process. Secondly, combining priori knowledge and sliding window technology to update the sample data, the Q statistic is classified by the Bayesian classification method to determine the current sample is normal signal or disturbance signal. If the current sample is from disturbance signal, it should be further determined that whether the current sample is the case of the abnormal value or caused by the process changes, so as to realize the online outlier detection for the process data.(3). According to the characteristics of cobalt precipitation reaction, the mechanism model of cobalt ion concentration is constructed, and the algorithm of kernel partial least squares is used to identify the unknown parameters of the mechanism model. Hence, a soft sensor model for cobalt ion concentration based on the mechanism and kernel partial least squares is built.(4). According to the characteristics of the soft sensor model, a method for online correction based on bidirectional recursive KPLS model parameters update and filtering correction combined is proposed, so as to improve the precision of soft sensor model.The proposed model is verified using industrial field data, and results have shown that the established soft sensor model for cobalt ion concentration has good tracking performance and high prediction accuracy, as well as effectively identify abnormal values in the process of online prediction. It meets the detection requirement for on-site process and can provide effective operation guidance for control and optimization of the production process. Figures(23), tables(5), references(77).