Soft-Sensor Modeling for Separation Performance of Dense-Medium Cyclone by Field Data

Abstract

Four measurable parameters, that is, density of dense-medium suspension (d), inlet pressure of dense-medium suspension (p), content of magnetic substance (c), and coal feed rate (r) were adopted to build a soft-sensor model for calculating the two performance index of a dense-medium cyclone in a Taixi plant. Uniform design was adopted to reduce the number of experiments. The models of actual separation density (δp) and probable error (Ep) obtained by genetic algorithm and regression were proved to be basically right by the 12 training records and another test result. The accuracy of the δp model was 0.7% for the training set and 0.63% for the test data while that of the Ep model was 9.41% and 13.54%, respectively. The behavior of the models were in accordance with field experiences, which showed that p had the most significant effect on Ep and c affected δp most prominently in daily operation.