A novel CFD-DEM methodology for simulating particle deposition and post-deposition mobility on filter fibers

Modeling air filtration of solid particles in fibrous filter media commonly assumes that the particles remain immobile once they are deposited on the fibers, despite experimental evidence indicating otherwise. Using a new simulation approach, this work investigates how post-deposition particle displacement affects deposit formation and, consequently, overall filtration performance. The new approach models the filtration process as a sequence of particle deposition events. Each event combines a spatially resolved steady-state IBM simulation with a one-way-coupled Computational Fluid Dynamics (CFD)-Discrete Element Method (DEM) simulation. The former determines the quasi-steady flow field around fibers and deposited particles, while the latter captures particle dynamics within this frozen flow field. Compared to temporally and spatially resolved four-way-coupled CFD-DEM-Immersed Boundary Method (IBM) simulations of the deposition process, the new approach achieves similar accuracy while being orders of magnitude more efficient. Simulation results for deposit morphologies of polystyrene particles on a single steel fiber agree well with experimental images under the same conditions. Post-deposition particle displacement is identified as an important factor determining the final deposit morphology. Loose dendritic structures, characteristic for low Stokes numbers, remain highly mobile, undergoing mostly gradual relocation via dendrite rotation driven by aerodynamic forces or triggered by particle impacts. With increasing particle inertia, collision-induced interactions dominate, leading to short-range particle rearrangements and progressively more compact deposits. Similar particle displacement patterns are also observed for a multi-fiber arrangement, where particle relocation leads to interaction of particle deposits pertaining to different fibers, which results in increased flow blockage, and earlier clogging than predicted by a ‘Caught on First Touch’ collision model.