Intelligent predictive control of cleaning flotation process based on froth texture features

Cleaning process is the key stage affecting the final concentrate grade of mineral flotation. The operation parameters are normally adjusted manually by operators to achieve the optimal control of the concentrate grade. However, this method is of great subjectivity, randomness and uncertainty. Therefore, an intelligent predictive control method for the cleaning process based on the froth texture features is proposed in this paper. Firstly, the features of cleaner flotation froth are analyzed, and the significant features are expressed with texture features by using the color co-occurrence matrix. Then an improved online prediction model is constructed and optimized by rolling optimization with the differential evolution algorithm, in order to achieve the optimal control of the pulp level. The validation of industrial data shows the effectiveness of the proposed method in bauxite cleaning flotation process, which is applied to stabilize the flotation process and concentrate grade.