Parallel AHU Filters: Calculate Flow Redistribution When Bank Resistances Differ

Original illustrative engineering schematic. The worked calculation below uses explicit example inputs.
A total-airflow reading can conceal a significant imbalance inside an AHU filter bank. When several modules connect the same upstream and downstream plenums, a lower-resistance module normally takes more air. If some modules load faster than others, dividing the measured bank flow by the module count can therefore misstate the duty of every filter.
This matters during replacement planning and fault diagnosis. A bank may meet its scheduled 4.0 m³/s while two modules each handle 21% more than the equal-share design flow. The useful check is a small parallel-flow network: preserve the total flow, require the same pressure difference across each path, and use the appropriate resistance curve for each module.
Start with the pressure boundary
The model below places four equal-size modules side by side between two large plenums. Upstream pressure is taken as uniform, downstream pressure is taken as uniform, and local frame or approach-flow differences are neglected. Every path therefore has the same pressure drop. Flow is allowed to redistribute until all four resistance relationships and the total-flow balance agree.
Do not impose this simplified boundary on a bank directly behind an elbow or partly obstructed by internal hardware without checking the pressure field. Different local approach conditions can add unequal losses. That is a distribution problem requiring additional measurements or analysis, not a reason to force all module flows equal.
Use curves rather than a universal square law
Air-filter pressure drop should be taken from applicable product data or defensible measurements. A pure square-law assumption is not automatically valid for fibrous media. For this original worked example, use two explicitly hypothetical curves containing both linear and quadratic terms. They demonstrate the method without pretending to reproduce a particular product.
Clean module: Δp = 40q + 60q².
Loaded module: Δp = 80q + 120q².
Here Δp is in Pa and q is the actual module flow in m³/s; the coefficients carry the corresponding units.
At 1.0 m³/s, the clean curve gives 100 Pa and the loaded curve gives 200 Pa. Doubling the resistance curve is a teaching assumption for this snapshot. It is not a universal description of dust loading, and it says nothing by itself about a filter’s particle-removal efficiency or remaining useful life.
Real selections need the exact module size, media configuration, rated airflow and allowable operating range. Camfil’s Hi-Flo data, for example, list different dimensions and rated flows and state a maximum flow for that product family. Those product-specific limits illustrate why one nominal bank-face velocity is insufficient; they are not the source of the hypothetical curves used here.

Illustrative four-module parallel bank: two lower-resistance modules each pass 1.214 m³/s and two higher-resistance modules each pass 0.786 m³/s at a common 137 Pa pressure drop; total flow is 4.000 m³/s. Rounded values; see the stated resistance assumptions.
Original example: hold total flow at 4.0 m³/s
Two clean modules follow the first curve; two loaded modules follow the second.
A functioning airflow-control loop maintains total flow at 4.0 m³/s within fan capability.
The same dry air state is assumed across the bank for the volumetric balance.
No bypass leakage, missing filter, damaged pocket or local obstruction is included.
Let x be the flow through each clean module and y the flow through each loaded module. Symmetry within each pair makes those pair members equal, but it does not make the clean and loaded flows equal. Continuity gives 2x + 2y = 4, so y = 2 − x.
Equal pressure: 40x + 60x² = 80(2 − x) + 120(2 − x)².
Rearranging gives 60x² − 600x + 640 = 0.
The physically admissible root is x = 1.214061 m³/s; y = 0.785939 m³/s.
Common bank pressure = 40(1.214061) + 60(1.214061)² = 136.999 Pa.
Substituting 0.785939 into the loaded curve gives the same 136.999 Pa.
Total flow = 2(1.214061) + 2(0.785939) = 4.000000 m³/s.
Each clean module carries 121.4% of the equal-share design flow, while each loaded module carries 78.6%. The two clean modules together carry 60.7% of the total through half the available module positions. Check 1.214 m³/s against the actual selected module’s operating limits before deciding whether the arrangement is acceptable.
Why averaging 100 Pa and 200 Pa gives the wrong answer
An arithmetic average of the two pressures at 1.0 m³/s would suggest 150 Pa. But 1.0 m³/s through every module requires two different pressure differences between the same plenums. That is inconsistent with the model. The common pressure has to be solved after the flows redistribute, and the result here is about 137 Pa.
The same caution applies to spreadsheet schedules. A row labelled “average filter pressure drop” may be useful for a procurement summary, but it is not automatically a network calculation. Retain the individual curves and multiplicities so a reviewer can reproduce the equilibrium rather than accepting an unexplained average.
Change the operating boundary and the result changes
Now consider a separate hypothetical case where the pressure across this bank is held at 100 Pa instead of total flow being maintained. Each clean module passes 1.0 m³/s. Solving the loaded curve at 100 Pa gives only 0.638492 m³/s per loaded module.
Loaded module at 100 Pa: 120y² + 80y − 100 = 0, giving y = 0.638492 m³/s.
Total bank flow = 2(1.0) + 2(0.638492) = 3.276984 m³/s.
This is an 18.1% shortfall from the 4.0 m³/s target. It is not a prediction for a constant-speed fan: that fan would settle where its complete fan curve intersects the complete system resistance curve. Likewise, a duct-static-pressure controller does not necessarily hold filter-bank pressure constant. Identify the controlled variable before choosing either simplified case.
Turn the calculation into a field decision
Confirm the module schedule, installation direction, seating, seals and replacement history before interpreting differential pressure.
Record total AHU flow and bank pressure at the same operating condition, plus fan-control mode and any limiting override.
Compare each module group with suitable resistance data; do not infer all four flows from one pressure reading without those curves.
Where needed, use a safe, appropriate measurement method and representative locations to investigate distribution. Do not probe moving machinery or disturb filters in operation.
After corrective work, repeat the same operating-point check and retain the before-and-after evidence.
For a bank containing different sizes, include each size separately and sum its predicted flow at a trial common pressure. Iterate that pressure until the flow sum matches the target. This pressure-based method also works with tabulated manufacturer curves and avoids requiring all curves to share one algebraic form.
First establish that the bank has enough correctly selected modules, then investigate how flow divides between them. The related filter-bank sizing review addresses dimensions and rated flow; the bypass inspection review addresses air that avoids the media altogether. Neither issue should be hidden inside a resistance-adjustment factor.
Build a more complete component-review and fault-diagnosis workflow with the AHU Selection & Fault Diagnosis Handbook. It brings filtration, fan duty, installation checks and commissioning into one engineering reference, so a bank-pressure reading can be reviewed alongside the rest of the air handler.
Continue the calculation review
Sources and calculation basis
Primary references checked on 23 October 2026. The worked examples and diagrams are original illustrative calculations, not manufacturer selections or project approvals. Apply the stated limitations and the requirements governing the actual project.


Comments