Modeling pressure drop coefficient for cyclone separators: A support vector machine approach

As one of performance critics for cyclone separators, pressure drop is an important parameter to evaluate and design cyclone separators. In order to accurately predict the complexly nonlinear relationships between pressure drop coefficient (PDC) and geometrical dimensions, a support vector machine (SVM) model is developed and employed to model PDC for cyclone separators. Based on the normalization method and the random sampling technique for the experimental sample dataset, a dynamically optimized search technique with cross validation is introduced to determine optimal algorithm parameters in the model. Then the optimized SVM model is trained and tested by the simulation results. According to the predicted accuracy of PDC for cyclone separators, the SVM model performance is compared and evaluated. It is found that the SVM model provides the higher generalization performance than the conventional models including the theoretical and statistical models as well as the artificial neural network model, with the mean squared error of 3.64×10−4 and the correlation coefficient of 0.9974. The result also demonstrates that SVM can offer an alternative and powerful approach to model cyclone pressure drop.