Advancing Dynamic Process Modeling of Comminution and Classification Circuits: A Paradigm Shift with GPU-Enabled DEM Solver

                                
==Abstract==

The minerals processing and aggregate industry have relied on steady-state
population and mass balance simulators for decades. However, accurately modeling new
processes remains a critical challenge that hinders innovation and decision-making in the
industry. In recent years, time-dynamic simulators have been developed, which offer
more accurate predictions of process variability and performance, as well as the ability
to introduce regulators and control algorithms. Yet, these still require simplified process
models of each unit in the system. The development of high-performance discrete element
method (DEM) solvers with advanced particle physics models presents a new opportunity
to model complete comminution and classification processes.
In this paper, we discuss the potential, challenges, and current limitations of using
DEM for advanced dynamic process and equipment evaluation, exemplified by a coarse
comminution crushing and screening case. We demonstrate the methodology using a
GPU polyhedral DEM implementation with a boundary-volume hierarchy (BVH) collision
search algorithm. The results show that the scale of a full-scale two-stage crushing process
is possible to simulate. The transition from algebraic process models to DEM would make
a significant advancement, bridging the current gap between overly simplified generalized
process models and specific equipment design. This approach offers exciting opportunities
for the mineral processing and aggregate industry to develop more innovative and efficient
circuits.