Performance optimization of banana vibrating screens based on PSO-SVR under DEM simulations

This paper carried out the numerical simulation about the movement of non-spherical particles on banana vibrating screen using direct element method (DEM) considering the complexity of particle collision and avoiding obtaining motion information with difficulty. Experimental prototype of banana vibrating screen under variable parameters was manufactured to verify the feasibility of simulations. Because the complex non-linear mathematical model is the basis of optimization. Based on the simulation data this paper applied the least squares support vector machines (LS-SVM) to establish relationships between vibrating parameters of banana screen and screening performance. LS-SVM based on statistical theory can effectively solve the mapping problem of small sample. At same time, in order to improving the quality of modeling, the kernel parameters of SVM were optimized by particle swarm optimization (PSO). Considering multi-extremum, large-scale, and non-differentiable of this computational model, the artificial fish-swarm algorithm (AFSA) with strong robustness and global convergence was applied to vibration parameters optimization. Finally, the optimal vibration parameters were: vibration amplitude 2.4 mm, vibration frequency 21 Hz, vibration direction angle 40 degrees