Integrated estimation model of clean coal ash content for froth flotation based on model updating and multiple LS-SVMs

Clean coal ash content, a prominent product index describing coal froth flotation, is difficult to be measured online. This constraint leads to a lack of timely guidance during operation and impedes the optimal operation of the coal flotation process. To solve this problem, considering the fluctuation of working conditions, the heterogeneity of raw coal and the variation of feed coal classes, an integrated estimation model of clean coal ash content for coal flotation based on model updating and multiple least squares support vector machines (LS-SVMs) is proposed. First, a single estimation model for a single class of coal based on LS-SVM is built, and the internal parameters are optimized by gravitational search algorithm (GSA). Second, the model updating strategy is designed to solve the problem of the decline in single model accuracy. Furthermore, a multiple LS-SVMs model formed by several single models for different classes of coal is studied along with the model switching mechanism to address the problem of model mismatch. Finally, an industrial experiment and evaluation are conducted. The mean relative error between the estimated and actual values is 3.32%, and the correlation coefficient is 0.9331. The estimation accuracy and adaptability of the integrated model can meet the industrial requirements.