Upward Flow Filter

An upward flow filter is a pressure or gravity vessel where the fluid enters at the bottom, rises through a graded bed of filter media, and exits cleaned at the top. Municipal water treatment plants and hydraulic flushing rigs rely on this configuration to polish particulate down to 10-20 µm. The upflow direction lifts and lightly fluidises the bed during operation, so the coarsest grains sit at the bottom and the finest at the top — opposite of a downflow filter. The result is longer run times between backwashes and lower clean-bed pressure drop at equivalent loading rates of 5-15 m³/m²·h.

Upward Flow Filter Cross Section A static engineering diagram showing a vertical cross-section of an upward flow filter vessel with graded media layers, underdrain distributor, and flow direction indicators. Clean fluid outlet Dirty fluid inlet Collector screen Freeboard (≥30%) Fine sand (0.4-0.6 mm) Medium media Coarse garnet (1.5-2.5 mm) Underdrain nozzles 5-10% bed expansion Deep-bed capture: Large particles stop in coarse layer Flow direction: Bottom to top Media grading: Coarse at bottom, fine at top (opposite of downflow filters)
Upward Flow Filter Cross Section.

Operating Principle of the Upward Flow Filter

The mechanism is simple in principle but unforgiving in execution. Dirty fluid enters through an underdrain distributor at the base of the vessel, passes upward through a stratified bed of media — typically silica sand on top of garnet or anthracite at the bottom — and exits via a top collector. As fluid rises, particulate gets trapped in the deeper, coarser layers first, then the finer top layers polish what remains. This is the inverse of a conventional downflow filter, where the finest grains sit at the surface and clog quickly.

The critical design parameter is the fluidisation velocity. The upward flow must be fast enough to keep the bed slightly expanded — usually 5-10% above the static height — so that particulate works deep into the bed instead of forming a surface cake, but it cannot exceed the terminal settling velocity of the smallest media grains, or you wash media straight out the top. For 0.5-0.8 mm silica sand at 20 °C, that working window sits between roughly 8 and 18 m/h. Run too slow and you get channelling — fluid finds preferential paths, the bed pressure drop reads low but turbidity at the outlet climbs. Run too fast and the freeboard fills with mobile media, the underdrain laterals erode, and you start finding sand grains in your downstream pump suction.

Freeboard expansion needs at least 30% of the static bed height above the top of the bed, otherwise the backwash cycle throws media into the outlet collector. The underdrain distributor itself must achieve flux uniformity within ±5% across the cross-section — anything worse and you get dead zones where media compacts, channels open, and effective filter area drops by 20-40% inside the first month of service.

Key Components

  • Underdrain distributor: A perforated lateral or nozzle plate at the vessel base that introduces flow uniformly across the full cross-section. Open area typically 0.2-0.4% of the bed area, with nozzle slot widths of 0.2-0.3 mm — narrow enough to retain the support gravel but wide enough that head loss stays under 0.3 bar at design flow.
  • Graded media bed: Layered filter media with the coarsest grains (1.5-2.5 mm garnet or gravel) at the bottom and finest (0.4-0.6 mm sand or anthracite) at the top. Total bed depth runs 600-1200 mm depending on duty. The grading must hold under fluidisation, so each adjacent layer's d10 ratio should not exceed 4:1 to prevent intermixing.
  • Freeboard volume: Empty vessel space above the bed, sized for at least 30% bed expansion during backwash. Skimp on freeboard and you carry media into the upper collector — common failure on retrofits where someone added a deeper bed without raising the dome.
  • Top outlet collector: Fine-slot screen or nozzle array at the vessel top that captures cleaned fluid while excluding media. Slot width must be smaller than the d10 of the finest media layer — typically 0.15-0.25 mm for a 0.5 mm sand top layer.
  • Backwash inlet and waste outlet: Reverse-flow connections used to fluidise and rinse the bed once differential pressure exceeds 0.8-1.0 bar. Backwash flow rate runs roughly 2-3× service flow to lift the bed cleanly without ejecting media.
  • Differential pressure gauge: Reads ΔP across the bed. Clean-bed ΔP sits at 0.1-0.3 bar; backwash trigger point is typically 1.0 bar. If ΔP climbs faster than expected, you have either inlet solids loading above design or a partially blocked underdrain.

Where the Upward Flow Filter Is Used

Upward flow filters show up wherever long run times, low clean-bed pressure drop, and gentle handling of fragile floc matter more than absolute filtration sharpness. You see them in municipal water treatment, hydraulic flushing, sugar refining, and hatchery recirculation systems. The same configuration also handles ion exchange polishing because upflow keeps the resin bed compact against the top retention screen.

  • Municipal water treatment: Pre-treatment upflow roughing filters at the Phnom Penh Water Supply Authority's Chroy Changvar plant, used to drop Mekong River turbidity before downstream rapid sand filters.
  • Hydraulic system flushing: Portable upflow flushing skids from Hy-Pro Filtration used to clean ISO VG 46 hydraulic oil to ISO 4406 17/15/12 cleanliness on injection-moulding machine commissioning.
  • Sugar refining: Sweetwater polishing filters at Tate & Lyle's Thames refinery, removing fine bone-char fines from melt liquor without crushing the carbon.
  • Aquaculture: Upflow biofilters in recirculating salmon hatcheries at Mowi's Rosyth facility, where the upward flow keeps the nitrifying biofilm aerated without abrading it.
  • Power generation condensate polishing: Upflow ion exchange beds at the Drax power station condensate polishing plant, where cation and anion resin layers stay compacted against the top screen during service.
  • Industrial process water: Upflow multimedia filters from Pentair's structural composite vessel range used for pre-RO filtration on dairy CIP rinse-water recovery at Fonterra Te Rapa.

The Formula Behind the Upward Flow Filter

The single most useful calculation for sizing an upward flow filter is the filtration loading rate — the volumetric flow per unit bed area. At the low end of the typical 5-15 m³/m²·h range, you get excellent polishing but the vessel is oversized and capital cost suffers. At the high end you get compact equipment but run-time between backwashes drops and turbidity breakthrough risk climbs. The sweet spot for most multimedia duties sits around 10 m³/m²·h, where bed pressure drop, run length, and capital cost balance out. The formula tells you what diameter vessel you need for a given service flow.

Abed = Q / vload and Dvessel = √(4 × Abed / π)

Variables

Symbol Meaning Unit (SI) Unit (Imperial)
Abed Required filter bed cross-sectional area ft²
Q Service flow rate m³/h gpm
vload Filtration loading rate (superficial upflow velocity) m/h (m³/m²·h) gpm/ft²
Dvessel Internal vessel diameter m ft

Worked Example: Upward Flow Filter in a craft brewery liquor pre-filter

You are sizing an upflow multimedia filter to polish municipal feed water before the reverse osmosis skid at Brewdog's Ellon brewery in Aberdeenshire. The feed water enters at 14 m³/h with average turbidity of 2.5 NTU, peaking to 6 NTU during winter run-off events. The downstream RO membranes need feed below 0.5 NTU and SDI<sub>15</sub> under 4. You need to pick a vessel diameter and check the operating window.

Given

  • Q = 14 m³/h
  • vload,nom = 10 m/h
  • vload,low = 5 m/h
  • vload,high = 15 m/h

Solution

Step 1 — at the nominal loading rate of 10 m/h, calculate the required bed area:

Abed = 14 / 10 = 1.40 m²

Step 2 — convert that to a vessel diameter:

Dvessel = √(4 × 1.40 / π) = √1.783 = 1.335 m

You'd round up to a standard 1.4 m diameter Pentair-style composite vessel. That gives you the design sweet spot — long run times of 18-30 hours between backwashes, clean-bed ΔP around 0.2 bar, and turbidity output reliably under 0.3 NTU.

Step 3 — check the low end of the operating range, 5 m/h, which is what you'd run during low-demand mash-tun cycles:

Abed,low = 14 / 5 = 2.80 m² → D = 1.89 m

At 5 m/h the same 1.4 m vessel is operating at half its design loading. Polishing is excellent — turbidity drops below 0.15 NTU — but the bed barely expands, channelling risk goes up, and you spend capital on a vessel twice the size needed.

Step 4 — check the high end, 15 m/h, which is what happens if you push the same 1.4 m vessel during a peak brew-day surge:

vactual = 14 / 1.40 = 10 m/h (nominal); but at 21 m³/h surge: vactual = 21 / 1.40 = 15 m/h

At 15 m/h the bed runs near the upper edge of the working window. Run time between backwashes drops to 6-10 hours, ΔP climbs faster, and turbidity breakthrough during the last hour of a run becomes a real risk — you'll see the SDI15 on the RO inlet creep from 2.5 toward 4.5.

Result

The nominal answer is a 1. 4 m internal-diameter vessel running at 10 m/h loading rate. That gives you a 24-hour run length on average Aberdeenshire feed water and clean-bed ΔP around 0.2 bar — exactly the sweet spot. At 5 m/h the same vessel polishes harder but wastes capital; at 15 m/h surge you lose roughly 70% of your run length and risk turbidity breakthrough hitting the RO membranes. If the measured outlet turbidity is higher than predicted, the most common causes are: (1) underdrain lateral cracking causing flow maldistribution and channelling, (2) media intermixing from a backwash that was run above 25 m/h and disrupted the grading, or (3) a partially blocked top collector screen forcing localised high-velocity exit zones that drag fines through the bed.

Upward Flow Filter vs Alternatives

Upward flow filters compete against conventional downflow rapid sand filters and cartridge filters for the same duty points. Each has a defensible niche — the comparison below uses the engineering attributes you'd actually search on when specifying.

Property Upward Flow Filter Downflow Rapid Sand Filter Cartridge Filter
Typical loading rate 5-15 m³/m²·h 5-12 m³/m²·h 10-30 m³/m²·h face velocity
Run length between cleaning 18-30 hours typical 8-24 hours typical 1-6 months (then replace)
Clean-bed pressure drop 0.1-0.3 bar 0.2-0.5 bar 0.1-0.2 bar new, up to 2 bar fouled
Filtration cut point 10-20 µm nominal 15-30 µm nominal 0.5-50 µm rated
Capital cost (per m³/h capacity) Medium Medium-high Low (housing) but high lifetime cartridge cost
Maintenance interval Backwash every 18-30 h, media change 5-7 years Backwash every 8-24 h, media change 3-5 years Cartridge change 1-6 months
Best application fit Pre-RO, hydraulic oil polish, fragile-floc duties Municipal potable water, high-solids raw water Low-flow polishing, point-of-use, batch process
Risk if oversized Channelling, poor effluent quality Surface caking, short run length Wasted spend on housings

Frequently Asked Questions About Upward Flow Filter

This is media restratification lag. Right after backwash the bed is fluidised and grains haven't fully resettled into their graded layers — fines that should sit at the top are still distributed through the bed depth, so the polishing layer is effectively missing for the first 5-15 minutes of forward service flow.

The fix is a settle period. After backwash, hold the vessel static for 60-90 seconds before opening the service valve, then ramp service flow up over another 2-3 minutes rather than slamming it open. If you still see post-backwash turbidity spikes after that, your backwash flow rate is probably too high and is throwing the d10 fines into the freeboard, where they re-enter the bed at random depths.

Single-media sand is fine for service flows below about 8 m/h and feed turbidity below 5 NTU — the bed depth does the work and capital cost is lower. Above those numbers you want multimedia because it gives you a deeper effective filtration zone without proportionally raising pressure drop.

Rule of thumb: if your design loading is above 10 m/h or your feed solids loading exceeds 30 mg/L, go multimedia. If you're polishing already-clarified water at 5-8 m/h, sand alone is the cheaper, simpler answer.

Low clean-bed ΔP almost always means flow is bypassing the bed rather than passing through it. The two usual culprits are channelling — preferential flow paths through compacted dead zones — and underdrain cracks letting fluid skip the lower media layers entirely.

Diagnostic: pull a turbidity sample at the outlet. If ΔP is low AND turbidity is high, you have channelling or an underdrain breach. If ΔP is low AND turbidity is also good, you may simply have a feed that's cleaner than design assumed, which is fine but means you can push loading rate higher and shrink the vessel on the next install.

You can push to 18-20 m/h on a multimedia bed with coarse top media (anthracite at 0.8-1.0 mm), but you'll cross the fluidisation threshold for finer top layers and start losing media out the top collector. Once you hit roughly 22-25 m/h on standard 0.5 mm sand the top inch of the bed is in continuous mild fluidisation during service flow, not just during backwash, and grains migrate into the freeboard.

Practical limit: stay below 80% of the calculated minimum fluidisation velocity of your finest media layer at operating temperature. For 0.5 mm silica sand at 20 °C that's around 14-15 m/h. Higher temperatures lower viscosity and let you push harder, which is why hot-process applications (sugar liquor at 80 °C) routinely run 18-22 m/h.

If media is escaping during forward service flow rather than backwash, the top outlet collector slot width is too wide for your finest media layer — or the screen has cracked. Slot width must be smaller than the d10 of the smallest media grains, typically 0.15-0.25 mm.

Pull and inspect the top collector. A cracked nozzle or a corroded wedge-wire screen lets the d10 fraction through, and once even 5% of the fines escape, the bed grading is permanently disturbed and effluent quality keeps degrading. Replacement collectors are cheap; running a damaged one for a month and contaminating a downstream RO membrane stack is not.

Yes — viscosity drops roughly 50% from 5 °C to 25 °C, which means the same loading rate produces a different bed expansion and a different effective filtration depth. A filter sized at 10 m/h on 20 °C summer feed will see noticeably less bed expansion on 4 °C winter feed, which can flip a marginal design from 'lightly fluidised' into 'static and channelling'.

If your feed temperature swings more than 15 °C across the year, size the bed for the cold-weather case and verify the warm-weather backwash flow doesn't fluidise the bed past the freeboard limit. The Brewdog Ellon example earlier sees 4-18 °C feed across the year, which is why the design holds up at both extremes — but it was checked, not assumed.

References & Further Reading

  • Wikipedia contributors. Rapid sand filter. Wikipedia

Building or designing a mechanism like this?

Explore the precision-engineered motion control hardware used by mechanical engineers, makers, and product designers.

← Back to Mechanisms Index
Share This Article
Tags: