What 'Reverse Osmosis Design Criteria' Actually Means
Reverse osmosis design criteria are the locked-in engineering inputs a designer sets before sizing an RO train: feed water SDI <5, free chlorine <0.1 mg/L, recovery 50–95%, flux 10–20 LMH for spiral-wound elements, salt rejection 99.0–99.8%, transmembrane pressure 10–30 bar, specific energy 2–6 kWh/m³, and concentrate disposal plan. Each value is set by feed chemistry, target permeate quality, and concentrate management limits, then validated by pilot testing on the actual site water.
These criteria are not the same as operating parameters. Design criteria are the targets the equipment is selected to meet; operating parameters are the levers a plant engineer adjusts after commissioning to track fouling, temperature, and feed variability. Confusing the two is how a vendor proposal gets accepted with a 30 LMH flux that the membrane will not sustain past month four.
Design criteria are also distinct from the design basis. The basis is the characterization of the feed water (TDS, hardness, SDI, temperature, organics) and the permeate specification (conductivity, silica, boron, TOC). Criteria are the derived targets — recovery, flux, rejection, energy, pretreatment train, concentrate plan — that the basis forces on the system. A 2026 pilot study on flat-sheet polyamide RO treating DAF-pretreated poultry slaughterhouse wastewater ran three recirculation rates in succession, with 0.5 L/min delivering the highest hydraulic stability and flux recovery — confirming that pilot confirmation on real feed, not modelled assumptions, is what freezes the commercial design (Polymers, Aug 2026, 18(16):1974).
Five parameters drive industrial specification more than any others: feed water spec, recovery, flux, rejection, and specific energy. Two more — pretreatment train and concentrate management — decide whether the five core numbers are even achievable on a given site. This article covers all seven in the order a designer would actually work through them.
Core Design Parameters at a Glance
Most specification errors come from values that look reasonable in isolation but were never cross-checked against the other six. The table below consolidates the typical industrial range for each criterion alongside the threshold that forces a design change. Paste it into your basis-of-design document and add site-specific values in the right-hand column.
| Parameter | Industrial Range (Brackish) | Seawater Range | Trigger Threshold (forces a design change) |
|---|---|---|---|
| Feed SDI15 (ASTM D4189) | <5 (target <3) | <5 | >5 → add UF or multi-media pretreatment |
| Free chlorine | <0.1 mg/L | <0.1 mg/L | >0.1 mg/L → carbon filter or sodium bisulfite dosing |
| Langelier Saturation Index (LSI) | Negative preferred | Negative | LSI >0 → acid dosing or antiscalant shift |
| Recovery | 50–95% | 35–50% | Set by concentrate osmotic pressure and scaling indices |
| Flux (25°C, spiral-wound) | 10–20 LMH | 9–14 LMH | >20 LMH accelerates fouling, shortens element life |
| NaCl rejection (new TFC element) | 99.0–99.8% | 99.0–99.8% | Below 99% → element replacement or damage |
| Transmembrane pressure | 10–30 bar | 55–80 bar | Set by feed TDS × recovery + NDP + friction |
| Specific energy | 2–6 kWh/m³ | 4–7 kWh/m³ (2–4 with ERD) | Outside range → audit pump and ERD selection |
| CIP frequency (design assumption) | Every 1–6 months | Every 1–3 months | <30 days → pretreatment upgrade required |
For residential context, the EPA WaterSense specification caps point-of-use RO reject at 2.3:1 (gallons of concentrate per gallon of permeate) — efficient residential units still waste more than 2× the treated volume (EPA WaterSense, Nov 2024 spec; Aug 2026 clarifications). Industrial systems target concentrate-to-permeate ratios of 1:1 to 4:1 depending on feed salinity, which is the same ratio written in reverse. A 75% recovery unit, for example, produces 3 gallons of concentrate for every 1 gallon of permeate that does not make it through the membrane as product — the high end of industrial efficiency.
Feed Water Specification: The Criteria That Drive Everything Else
Feed water quality sets the boundary conditions for every other criterion on the list. A change in feed SDI from 3 to 6 does not just require more pretreatment; it can shift flux by 30%, shorten element life by a year, and force a recovery reduction that increases concentrate volume by 40%.
The Silt Density Index at 15 minutes (SDI15, ASTM D4189) is the universal gate. Any feed measuring above SDI 5 will foul spiral-wound elements with colloids faster than CIP can keep up. Below SDI 3, the membrane runs at design flux for years. Multi-media filter vessels that drop SDI below 5 are the workhorse for surface water and clarified secondary effluent, and they sit immediately upstream of the cartridge filter and high-pressure pump.
Free chlorine must stay below 0.1 mg/L at the membrane inlet. Thin-film polyamide composite membranes — the industry default — oxidize above this threshold; rejection drops permanently within hours of exposure, and the element is not recoverable. Two options handle this: granular activated carbon (GAC) filtration for feeds with low chlorine demand, or continuous sodium bisulfite (SBS) dosing controlled by an inline ORP probe. Both require an automatic chemical dosing system sized to the peak chlorine load, not the average.
Hardness, silica, barium, and strontium do not damage the membrane directly but set the upper recovery limit through scaling indices. The standard practice is to run the feed composition through projection software (e.g., Proton, ROSA) that iteratively raises recovery until the concentrate saturation index for CaCO3, BaSO4, SrSO4, or SiO2 crosses the antiscalant-modified threshold. That recovery becomes the design ceiling — pushing past it is what causes irreversible scaling that CIP cannot remove.
Iron and manganese limits are tight: <0.05 mg/L Fe and <0.02 mg/L Mn at the membrane inlet. Both catalyze oxidative fouling on the polyamide surface, accelerating flux decline and creating a rough layer that harbours bacteria. Temperature affects viscosity, and design flux is corrected to 25°C with roughly 3% flux change per degree — a 10°C winter drop on a once-through cooling-water feed reduces design flux by about 25% if uncorrected.
Recovery, Flux, and Rejection: The Three Numbers Vendors Quote
Every RO datasheet leads with recovery, flux, and rejection. Every RO warranty denial starts with the same three numbers being misread. The definitions are not interchangeable, and the engineering coupling between them is what most proposals hide.
Recovery is permeate flow divided by feed flow, expressed as a percentage. The brackish industrial default is 75%, set by the balance between permeate volume and concentrate scaling indices. Seawater runs 35–50% because osmotic pressure at the concentrate end rises faster than the pump can compensate. For high-purity duty — pharmaceutical water-for-injection loops, semiconductor rinse, boiler feed >6 bar — a double-pass configuration pushes system recovery to 90–95%, achievable with industrial RO systems delivering up to 95% recovery on brackish feeds.
Flux is permeate flow per unit membrane area, measured in LMH (litres per square metre per hour). Brackish RO is designed at 10–20 LMH; seawater at 9–14 LMH, both corrected to 25°C. Pushing flux higher shortens element life and increases CIP frequency — a 25 LMH design might look 25% more productive on paper, but the operator will pay for it in 18-month element replacement instead of 5 years, plus 3× the cleaning chemical cost.
Rejection is element-specific. New thin-film composite elements reject 99.0–99.8% of NaCl under standard test conditions (2,000 mg/L, 25°C, 15% recovery, 150 psi). Vendor proposals sometimes cite 85% nitrate rejection, copied from residential point-of-use literature — that figure comes from a household undersink unit, not an industrial spiral-wound element, and using it to size a process train will undersize the rejection stage by an order of magnitude. Stick to 99%+ as the industrial reference and verify against the manufacturer's certificate of analysis for the actual element.
The coupling is the part vendors leave out: raising recovery raises concentrate osmotic pressure, which raises the feed pressure the pump must deliver, which raises specific energy. You cannot pick 85% recovery, 25 LMH flux, 99.8% rejection, and 2 kWh/m³ on the same membrane — at least two of those numbers have to give. The Polymers 2026 study made this coupling explicit in its hydraulic evaluation: the 0.5 L/min recirculation rate that produced the best flux recovery also bounded the recovery the system could sustain before fouling accelerated (Polymers, Aug 2026). Pilot work on the actual site water is what determines which of the four numbers gets relaxed on a given project.
Pretreatment Train: How Each Design Criterion Cascades Upstream
Pretreatment is not a separate workstream to be sized after the RO is selected. Each piece of upstream equipment exists to make the feed meet one of the criteria defined in the previous sections. Mapping the cascade is the most efficient way to avoid the common failure mode of buying an RO skid and then discovering the site feed will not let it run.
| Feed Condition | Trigger | Pretreatment Unit | Design Effect |
|---|---|---|---|
| High TSS, oil & grease | TSS >100 mg/L, O&G >50 mg/L | DAF as the first stage of an RO pretreatment train | Removes floatables and settleables before further solids handling |
| High turbidity or moderate SDI | Turbidity >50 NTU, SDI 3–5 | Multi-media filter (sand + anthracite + garnet) | Drops SDI to <5, protects downstream cartridge filter |
| High colloids, bacteria, or municipal effluent | SDI >5, variable feed, biological fouling risk | 0.03 µm UF pretreatment ahead of RO | UF effluent typically SDI <2; absolute barrier to bacteria and turbidity |
| Free chlorine, scale-forming ions | Cl2 >0.1 mg/L, LSI >0 | Automatic antiscalant / acid / SBS dosing | Protects membrane from oxidation and scaling |
| Final particulate guard | Always | 5 µm cartridge filter | Catches pretreatment media fines before high-pressure pump |
The full train reads, in order: raw water → DAF (if O&G or floatables present) → equalization → multi-media filter (or UF) → cartridge filter (5 µm) → RO → permeate polishing or concentrate management. Two items that look optional on a P&ID but never are: a 5 µm cartridge guard immediately before the high-pressure pump, and a flush port that lets the operator bypass the RO during pretreatment transients without dumping raw water at membranes.
For sites scaling up an existing RO train with tighter feed (e.g., adding a boron or silica limit), a nanofiltration retrofit and upgrade guide covers the option of inserting NF ahead of the RO to split the rejection load. For semiconductor rinse applications where the permeate specification is sub-ppb, the upstream criteria tighten to resistivity >18.2 MΩ·cm and TOC <10 ppb — see the semiconductor ultrapure water process design reference for the full polishing train.
Energy and Concentrate Management Criteria
Two criteria decide whether an RO project clears a P&L review: the specific energy in kWh per cubic metre of permeate, and the cost — environmental and financial — of the concentrate stream. Both are locked in at design, not at commissioning, and both are coupled to recovery and feed TDS.
Specific energy for brackish RO runs 2–6 kWh/m³ across most industrial plants. Seawater RO without energy recovery sits at 4–7 kWh/m³; adding an isobaric pressure exchanger or a Pelton turbine drops that to 2–4 kWh/m³, which is why every modern SWRO plant over 5,000 m³/d includes ERD as a default. The energy number is set by the feed pressure equation: feed pressure = osmotic pressure of the concentrate + net driving pressure + friction losses. On a 2,000 mg/L TDS feed at 75% recovery, the osmotic pressure at the tail element is roughly 10–12 bar, so the pump has to deliver 15–20 bar at the membrane inlet after friction and NDP are netted in.
Concentrate disposal is the criterion most often treated as an afterthought and most often the reason a permit is held up. Five options, in order of increasing cost: surface water discharge (regulated for TDS, temperature, and sometimes specific metals — NPDES permit required in the US), sewer disposal (requires POTW approval and a hydraulic load study), deep well injection (Class I well, geologically screened), evaporation ponds (climate- and footprint-limited), and zero-liquid-discharge crystallization (brine concentrator + crystallizer, capex-heavy but the only option where water rights are binding). The EPA explicitly notes that RO is water-intensive by nature, and the concentrate stream is a real environmental compliance item, not a footnote in the O&M manual (EPA WaterSense).
Energy and concentrate are linked: pushing recovery from 75% to 85% cuts the concentrate stream by a third but raises specific energy by 15–20% as osmotic pressure at the tail climbs. The optimization point is site-specific and should be modelled in the design phase, not after the pump is ordered. RO membrane elements and pressure vessels should be specified after the recovery-vs-energy trade-off is resolved, not before.
CIP, Shutdown, and Long-Term Operating Criteria
Most RO designs over-specify day-one performance and under-specify what the system needs in years 2 through 10. The clean-in-place parameters, shutdown flush protocol, and element replacement schedule are part of the design criteria, not commissioning afterthoughts.
CIP frequency is a design input, not a guess. The typical brackish RO design assumes CIP every 1–6 months; seawater every 1–3 months. Anything more frequent than monthly means the pretreatment train has failed one of the upstream criteria, and the fix is upstream, not in the cleaning chemistry. The CIP system itself includes a heated tank (30–35°C), a circulation pump sized at the same flow as the RO feed, pH-adjusted cleaners (typically pH 2 alkaline foam clean, pH 12 acid clean, in sequence), and a neutralization step before discharge.
The triggers to call a CIP are standardized: 10–15% decline in normalized flux at constant temperature and pressure, or 15% increase in differential pressure across the train. Both data points come from the RO SCADA trending — the instrumentation to capture them (flow, pressure, conductivity, temperature on each stage) is part of the design, not an option. The 2026 Polymers poultry-wastewater study found that even with optimized chemical cleaning, irreversible fouling accumulated across successive filtration cycles — meaning each CIP recovers less than the last, and the design life of 3–5 years for brackish elements and 5–7 years for seawater assumes the operator starts CIP before the irreversible fraction takes hold.
Shutdown criteria are short and non-negotiable: flush the train with permeate within 5 minutes of a stop, regardless of cause. A high-pressure pump that trips and sits stagnant for 30 minutes will foul the tail elements with concentrate that the next start cannot push out. For long-term storage (more than 72 hours), circulate and store in 1% sodium metabisulfite solution to prevent biological growth, and verify pH weekly. RO valves, instruments, and consumable media should be specified with documented shelf life and CIP compatibility so they do not become the weak link in year five.
One design choice that pays back across the asset life: oversize the CIP tank and heater by 20% above the minimum calculated volume. The extra hour of cleaning time is the difference between restoring 95% of original flux and restoring 80%, and that 15% delta is the difference between year 6 and year 4 element replacement.
Frequently Asked Questions
What SDI is required for RO feed?
SDI15 below 5 (ASTM D4189) is the universal gate for any spiral-wound RO element, with below 3 the practical target for trouble-free operation. Sites with municipal secondary effluent, high-algae surface water, or industrial wastewater with significant colloidal load need UF pretreatment; a well-designed UF stage delivers SDI below 2 consistently.
What is the typical recovery for an industrial RO system?
Brackish industrial RO defaults to 75% recovery, with 50–95% as the design range depending on feed scaling indices. Seawater systems run 35–50% because osmotic pressure at the concentrate end constrains recovery. Double-pass configurations on brackish feed for high-purity duty reach 90–95% system recovery.
How much energy does an RO system use per cubic metre?
Specific energy for brackish RO is 2–6 kWh/m³ of permeate across most industrial plants. Seawater RO without energy recovery runs 4–7 kWh/m³; adding an isobaric pressure exchanger or Pelton turbine drops seawater energy to 2–4 kWh/m³, which is the standard for any modern SWRO plant above 5,000 m³/d.
What flux should I design for?
Design flux for brackish spiral-wound RO is 10–20 LMH corrected to 25°C, with 14–17 LMH the typical operating window. Seawater RO is designed at 9–14 LMH. Pushing flux above 20 LMH on brackish accelerates colloidal and biological fouling and shortens element life from 5 years to 18–24 months.
When is UF required ahead of RO?
UF pretreatment is required whenever feed SDI exceeds 5 and cannot be brought down reliably with multi-media filtration, or whenever the feed is municipal secondary effluent, surface water with seasonal algae blooms, or industrial wastewater with significant colloidal and organic load. UF at 0.03 µm nominal delivers RO feed at SDI below 2 and removes the biological-fouling risk that multi-media filtration cannot address.
Related Equipment
- multi-media filter vessels that drop SDI below 5 — specifications, capacity range, and technical data