Presets
Playback
Parameters
Camera
Layers
Quality
Model check
The science
What a cyclone does. Dusty gas enters a cylinder tangentially and spins. Every particle feels a centrifugal push outwards and a Stokes drag from the gas that drifts inwards towards the exit tube (the vortex finder). Heavy particles win and are thrown to the wall, where they slide down the cone into the bin; light ones are dragged into the core and leave with the clean gas.
Gas field. The tangential speed at the wall follows angular momentum from the inlet centre line, vtw = vin(R - b/2)/R. Inside the barrel it is a free-vortex profile vt = vtw(R/r)n down to the finder radius rx, and solid-body rotation inside it. The exponent n is Alexander's (1949) correlation n = 1 - (1 - 0.67 D0.14)(T/283)0.3, about 0.56 for D = 0.3 m. The radial gas speed comes from continuity with a control cylinder of height H - S under the finder: ur(rx) = -Q/(2 pi rx (H - S)), falling to zero at the wall.
Particle. Stokes relaxation time tau = (rhop - rhog) d2 / (18 mu); the radial drift is ur + tau vt2/r, plus a random walk with diffusivity Dt = I vin 0.02 D that stands in for turbulence. A particle is collected when it reaches the wall and lost when it reaches rx below the finder mouth. The 3D dust you see is moved by exactly these equations (one particle = one random draw from the log-normal mass distribution; each draw carries equal mass).
Cut sizes. Barth's force balance at the finder radius gives d502 = 18 mu Q / (2 pi (H - S) (rhop - rhog) vt(rx)2). Lapple's classic correlation is d50 = sqrt(9 mu b / (2 pi N vin (rhop - rhog))) with N effective turns, and its grade curve eta = 1/(1 + (d50/d)2). Pressure drop follows Shepherd and Lapple, dP = (1/2) rhog vin2 16ab/De2.
Simplifications, honestly. Real cyclones also have wall friction (which lowers vt by 10 to 30 %), re-entrainment from the dust outlet, particle-particle collisions at high loads, a short-circuit flow along the roof and finder, and an inner vortex that is not perfectly axisymmetric; none of them is modelled, so predicted efficiencies are optimistic for fine dust and the fines floor (about 20 to 30 % of 0.5 um particles caught by wall diffusion) depends on the assumed turbulence. Slip correction is ignored (it matters below about 1 um). Particles do not affect the gas. The scene runs in slow motion (default 1/30 of real time), geometry follows the Stairmand high-efficiency and Lapple conventional proportions. Models of this kind agree with measurements only to about a factor of 1.5 in d50. The dust colours are particle sizes, blue small to red large; the sorting you see is the model result, not a scripted outcome.