Precoat + body-feed cake filtration: Ruth / Darcy law with a Kozeny-Carman filter-aid cake. The 2D twin of the 3D cutaway: same model core, plus the pressure profile across the growing cake.
The science
Filtration cycle. A leaf filter is first precoated: a thin layer of diatomaceous earth (DE, kieselguhr: fossil diatom shells of porous silica) is deposited on the leaf screens. During the run, more DE (body feed) is dosed into the dirty water so the cake stays permeable while the suspended solids are trapped in it. The run ends when pressure drop reaches the limit, when the cake fills the gap between leaves, or (constant-pressure mode) when the flow has fallen too far. Then the cake is washed off and the cycle restarts.
Ruth / Darcy cake filtration (per m2 of leaf, flux u = flow / area, w = cake mass per area, V/A = filtrate per area):
ΔP = μ·u·( Rm + α·w ), w = wprecoat + c·(V/A)
Constant flow rate: u is fixed, so for an incompressible cake ΔP grows linearly with V/A, with slope μ·u·α·c. Constant pressure: dV/dt = A·ΔP / (μ(Rm+α w)), which integrates to the straight Ruth line
t/(V/A) = [ μ·α·c / (2·ΔP) ]·(V/A) + μ·(Rm + α·w0) / ΔP
Kozeny-Carman specific resistance of a bed of porosity ε and effective grain diameter deff (density ρp):
k = ε³·d²eff / ( 180 (1−ε)² ), α = 1 / ( k·ρp·(1−ε) ) = 180 (1−ε) / ( ε³ ρp d²eff )
Compressibility. α depends on the pressure on the cake: α = α₀·(ΔPc/P₀)n, P₀ = 10 kPa, and ΔPc = μ·u·α·w. This is solved numerically at every step by bisection (rigid cake: n = 0, soft cake: n up to 0.7).
Mass balance. Per m2: fed solids (cDE+cs)·V/A = cake growth + solids that leaked into the filtrate. Cake thickness L = w / (ρp(1−ε)).
Particle capture (empirical illustration, not a measured curve). Each 0.25 mm of cake acts as one capture layer with efficiency η₁(d) = 1/(1+(dc/d)³), where the cut size dc is half the hydraulic pore diameter dpore = 2·√(8k/ε). N layers give η = 1−(1−η₁)N, averaged over a log-normal size distribution of the impurities (geometric std 2). So a thicker cake and a finer grade clarify better, while a coarse grade passes more fines and a thin precoat lets more through at the start.
Honest simplifications. (1) deff = 0.2 × the quoted mean particle size, because real DE grains are porous, platy and rod-like and their hydraulic diameter is much smaller than the screen size. This factor is illustrative; with it the permeabilities land in the usual 0.1-50 darcy range for filter aids. (2) Captured impurities have their own resistance αs (slider), because slimy or clay solids are far more resistant than DE. (3) Rm = 1010 m-1, ρp = 2200 kg/m³, cake bridging limit 25 mm per side, filter area 19.4 m² (9 leaves x 2 sides, 0.585 m radius) are typical round numbers. (4) Flow is uniform over all leaves. (5) Cake height in the 3D view is scaled to the bridging limit; the slurry is shown as drifting particles, not simulated fluid dynamics. (6) Temperature only changes water viscosity (Vogel-type fit). Real plants add polymer, pH, filter-aid grade blends and washing steps that are not modelled.
Independent checks. Each check re-runs a separate calculation and compares it with a closed-form result: constant-rate slope vs μ·u·α·c; constant-pressure Ruth line (slope and intercept); compressible cake ΔP vs the closed-form root; solids mass balance and thickness; Kozeny-Carman permeability vs a Darcy pressure-drop and a typical-range test.
Keywords: Diatomite Filtration, Porous Media, Particle Removal, diatomaceous earth, kieselguhr, precoat, body feed, Ruth plot, Kozeny-Carman, cake filtration. 2D art: Blender-rendered diatom frustule sprites and cake texture + Canvas.