Vibrating screens are widely used in industry for classifying particulate materials according to size. However, due to complicated particle-particle and particle-screen interactions, the design and control of vibrating screens still lack effective process models for guidance. DEM enables particle-scale analysis of screening processes but how to link DEM results to macroscopic models remains challenging. Recently, our studies proposed to combine DEM, machine learning and physics modeling to develop process models for single layer vibrating screens. In this paper, we review and generalize this methodology and extend it to multi-layer screens. Based on the assumption that particles passing through a certain section of a screen are dependent on local screen and flow conditions, the relationship between the passing rate and local condition is established through machine learning of data generated by a series of controlled DEM simulations. Two local passing functions are developed for top and bottom layer screens due to different aperture shapes. The local passing functions can predict the local flow of different size particles on a screen segment, according to the local vibration conditions, inclination angle, and the inlet flow. Then, the process model for the whole screen is developed by connecting different segments based on mass continuity, which can predict the overflow partition curve under various conditions more efficiently than DEM. The methodology can be used for not only multi-layer vibrating screens but also other complicated industry screening processes.
 

granular material,DEM,screening,machine learning,process model,local packing funciton