A Concentration Interval Identification Method of Cobalt Ions Based on Optimal Spectral Bands Selection and Naive Bayes Classification

The concentration of cobalt ions in the solution after zinc smelting and purification is an important indicator for evaluating the qualification of zinc solution. This paper mainly realizes the online detection of the concentration of cobalt ions in the purified solution of zinc smelting based on spectrophotometry. Due to the complex composition, small concentration change, and small spectral characteristic difference of the liquid, most spectral identification methods are prone to confusion and misjudgment for some samples with similar spectral characteristics. In this paper, the concentration interval identification method based on band optimization selection combined with the Naive Bayesian classification was adopted. Firstly, according to the "fingerprint characteristics" of the molecular spectrum, the characteristic spectral bands of cobalt were selected by the moving window method. The spectral bands containing the most cobalt ion information were selected according to the best prediction performance. Then apply the Gaussian Naive Bayes model to classify and predict the quality level of the liquid after cobalt removal. Comparing the proposed method with other common classification methods, the results show that this method can better capture the subtle differences between spectral features and identify the characteristic properties of cobalt ions. The concentration interval of cobalt can be distinguished with higher recognition accuracy and less computation time. The concentration interval recognition of purified liquid can also be accurately achieved in a relatively small number of samples.