The Method of Reagent Control Based on Time Series Distribution of Bubble Size in a Gold-Antimony Flotation Process

Due to the distribution of bubble size changes with the reagent dosage in a flotation process, a dosage control method based on time series distribution of bubble size during the gold-antimony flotation process is proposed. Firstly, since the flotation conditions cannot be described fully by the features of a single froth image, the concept of cumulative distribution function (CDF) is presented for describing the time series distribution of consecutive multi-frame bubble sizes, and approximated by a third order B-spline function. After that, a method that combines an radial basis function (RBF) neural network with the genetic algorithm (GA-RBF) is used to obtain the optimal CDF of the bubble size, then, a projection pursuit method is employed to reduce the multi-dimensional weights. Finally, a nonlinear prediction model combining reagent dosage and the projection vector is constructed based on partial least square through spline transformation (PLSS). The error between the output and the given CDF can be optimized by a differential evolution algorithm. The industrial experiment demonstrates the effectiveness of the proposed method.