**Zhaohui Tang 1, Liyong Tang 1, Guoyong Zhang 1,\*, Yongfang Xie <sup>1</sup> and Jinping Liu <sup>2</sup>**


Received: 29 March 2020; Accepted: 30 April 2020; Published: 3 May 2020

**Abstract:** It is well known that the change of the reagent dosage during the flotation process will cause the froth image to change continuously with time. Therefore, an intelligent setting method based on the time series froth image in the zinc flotation process is proposed. Firstly, the sigmoid kernel function is used to estimate the cumulative distribution function of bubble size, and the cumulative distribution function shape is characterized by sigmoid kernel function parameters. Since the reagent will affect the froth image over a period of time, the time series of bubble size cumulative distribution function is processed by the ELMo model and the dynamic feature vectors are output. Finally, XGBoost is used to establish the nonlinear relationship modeling between reagent dosage and dynamic feature vectors. Industrial experiments have proved the effectiveness of the proposed method.

**Keywords:** flotation process; reagent dosage; time series froth image; cumulative distribution function
