UF, RO and polishing for recycled cooling water: a practical selection guide
Contents
Direct answer
Ultrafiltration, reverse osmosis and final polishing can convert variable municipal recycled water into reliable cooling make-up, but only when each process has a defined purpose. Ultrafiltration controls particles and protects downstream equipment. Reverse osmosis removes dissolved salts. Polishing stabilises the final chemistry or disinfectant condition. None of them resolves supply interruption, tower contamination or concentrate disposal by itself.
The best treatment train is not necessarily the one with the highest recovery or the most barriers. It is the arrangement that meets the cooling-water design basis across the feed envelope, remains operable at part load and leaves a manageable set of residual streams.
Within a standardised programme, these processes can be configured as a DC-MU make-up module; where the source is treated on site, coordinate them with the DC-RW wastewater-reuse system.
1. Decide whether membrane treatment is necessary
Start with the feed-water envelope, target cooling-tower cycles, metallurgy and discharge limits. If filtered recycled water can support acceptable operation with appropriate chemistry, full-flow reverse osmosis may add unnecessary energy, consumables and concentrate. If chloride, sulphate, silica or total dissolved solids controls the tower, membrane treatment may be essential.
Compare at least three concepts: non-membrane conditioning, partial membrane treatment with blending and full membrane treatment. Calculate the resulting recirculating-water chemistry and residuals for each design case. This prevents a familiar process train from becoming the default before the actual problem is defined.
The treatment boundary also matters. Airborne dust and biological activity enter the open cooling system after make-up treatment. A high-quality make-up supply does not remove the need for side-stream filtration, cooling chemistry and water-safety controls.
2. Give equalisation and pretreatment real design duties
Equalisation should moderate changes in flow or quality, provide stable feed to downstream equipment and create time for off-spec decisions. Its volume must be linked to event duration, plant demand and backup strategy. Poorly mixed storage can instead create sediment, biological growth and delayed quality changes.
Pretreatment may include screening, coagulation, clarification, media filtration, oxidation, activated carbon, softening or targeted removal of iron, manganese or organics. Each stage should have a stated inlet condition, removal objective and waste stream.
Coagulation can improve particle and organic removal but increases sludge and demands careful dose control. Softening can protect membranes and cooling assets, although it creates brine or sludge depending on the process. Activated carbon may reduce particular organics or disinfectant residual, but it can become biologically active if not managed.
Chemical compatibility should be reviewed across the complete train. Residual coagulant, oxidants, antiscalants and cleaning chemicals can affect membranes, product water and discharge.
3. Apply ultrafiltration within its limits
Ultrafiltration is a physical barrier to suspended particles and many microorganisms. It can produce consistently low turbidity and silt density, improving reverse-osmosis reliability. It does not materially remove dissolved salts, hardness, alkalinity or most small dissolved organics.
Membrane material, pore rating and module configuration should be chosen for the source and cleaning regime. Design flux must account for temperature, fouling tendency, backwash frequency and one-train-out operation. A flux demonstrated with potable water is not a reliable basis for municipal recycled water.
Specify backwash, chemically enhanced backwash and recovery cleaning. Identify the water source for each, required tank capacity and resulting waste. Air scour and integrity testing may be appropriate. The design should state how a failed integrity test is isolated and how product quality is protected.
Key trends include transmembrane pressure, normalised permeability, filtrate turbidity, backwash interval and chemical consumption. These reveal gradual fouling earlier than a final high-pressure alarm.
4. Set reverse-osmosis recovery from chemistry
Reverse osmosis separates water from much of the dissolved load. Recovery determines how much feed becomes permeate and how concentrated the reject becomes. Raising recovery reduces feed volume but increases concentration at the membrane surface, scaling potential, osmotic pressure and concentrate salinity.
Use the worst credible combinations of calcium, sulphate, alkalinity, silica and other scale-forming constituents. Apply temperature correction and realistic membrane age. Antiscalant is not a substitute for a mass balance or a reliable concentrate route.
Two-pass reverse osmosis is rarely justified for ordinary cooling-tower make-up unless a particular product requirement demands it. A second stage or concentrate recycle may improve recovery but increases complexity and exposure to fouling. The design should compare energy, cleaning, availability and residuals rather than recovery alone.
Monitor feed pressure, differential pressure, permeate flow, permeate conductivity, salt passage, recovery and normalised performance. Trending distinguishes scaling, particulate fouling, organic fouling and membrane damage more effectively than a single product-conductivity limit.
5. Use polishing to finish a defined product
Polishing is the final adjustment between membrane or filtered water and the cooling make-up specification. It may include pH or alkalinity correction, remineralisation, decarbonation, disinfection, dechlorination or a targeted adsorption process.
Very low-mineral reverse-osmosis permeate can be aggressive to certain materials. Blending or controlled remineralisation may produce a more stable cooling make-up than untreated permeate. The final chemistry should suit the cooling programme and materials, not simply achieve the lowest conductivity.
Disinfection must be coordinated with storage residence time and cooling-tower treatment. A residual that protects a treated-water tank may interact with downstream corrosion inhibitors or membranes. The design needs defined sample points and separate limits for product water, stored water and recirculating cooling water.
6. Consider partial treatment and blending
Partial-flow reverse osmosis can remove enough dissolved load to support target tower cycles while avoiding full-flow membrane treatment. Permeate is blended with ultrafiltered recycled water or an approved backup source. The required fraction is calculated against the controlling constituent.
For a conservative constituent, blend concentration is the flow-weighted average of the component concentrations. In practice, multiple limits must be met simultaneously, and analyser accuracy, valve range and source variation must be included. The allowable blend window should be calculated for each design case.
Blending also creates operational choices. During favourable feed conditions, the RO fraction may fall; during high salinity it may rise. The control system should enforce maximum cooling make-up limits and fail safely if a source analyser becomes unavailable. More detail is provided in Storage, blending and monitoring.
7. Design redundancy, cleaning and turndown
AI data centres develop in phases, while water-treatment equipment has minimum stable flows. Modular trains can match early demand and preserve efficiency as the campus expands. The duty/standby philosophy should cover pumps, membranes, chemical dosing, analysers and control power—not only headline treatment capacity.
Define clean-in-place systems, chemical storage, heating if required, neutralisation, transfer and disposal. A common cleaning skid may be acceptable if simultaneous demand is not credible. Critical spares should be based on lead time and failure consequence.
At low load, stagnant headers and oversized tanks may be a greater risk than hydraulic capacity. Automatic flushing, recirculation and minimum-turnover logic should be included in the operating sequence.
8. Control biological stability
Recycled water can contain nutrients and biodegradable organics even when it is clear. Warm tanks, membranes and dead legs create opportunities for growth. Biological fouling reduces membrane performance and can compromise stored-water quality.
Control measures may include upstream disinfection, low-residence-time design, periodic cleaning and compatible biocides. Oxidant exposure must respect membrane limits. Dechlorination upstream of RO may be necessary for some membrane materials, increasing the importance of downstream hygiene.
Cooling-tower public-health management remains separate. The make-up treatment plant can reduce loading, but it does not replace the site’s competent Legionella risk controls, tower inspection, drift management, monitoring and disinfection.
9. Account for concentrate and backwash
List every residual by normal and peak flow, composition, frequency and duration: clarifier sludge, media-filter backwash, UF backwash, chemically enhanced wash, RO concentrate, cleaning solution, tank washdown and off-spec product.
Do not assume these streams can be combined. Acidic and alkaline cleaning wastes may require neutralisation; oxidants may affect biological sewer treatment; concentrate may govern salinity. Equalisation can reduce instantaneous load but does not reduce total salt mass.
Where blowdown recovery is considered, evaluate it with the RO concentrate route. Two independent recovery targets can produce an unmanageable final brine if designed separately.
10. Specify and test performance
The supplier specification should include feed design cases, product limits, net capacity, recovery, availability, turndown, energy, chemicals, waste flows, membrane life assumptions and cleaning frequency. State which values are guarantees and how they will be corrected for temperature and feed quality.
Factory testing should prove controls and hydraulic functions. Site testing should use representative water and include normal, limiting and fault cases. Verify automatic diversion, standby start, recovery calculation, analyser failure, clean-in-place sequences and restoration after an upset.
A reliability run should demonstrate stable normalised performance and acceptable waste production over an agreed period. Handover requires baseline membrane data, calibration records, operating limits, consumables, spares and trained operators.
11. Compare whole-life efficiency rather than membrane recovery
Energy and chemical consumption should be normalised to net compliant make-up delivered, not raw feed treated. A high-recovery plant may appear efficient by intake volume while requiring higher pressure, more antiscalant and more frequent cleaning. A lower-recovery concept may consume more feed but operate more reliably and leave a residual the site can discharge.
Build an annual operating model across source temperature, salinity and campus load. Include feed and transfer pumping, air scour, backwash, RO pressure, cleaning, remineralisation and standby operation. Early phases can be inefficient if large pumps and membranes operate below their intended range.
Consumables deserve the same scrutiny. Estimate cartridge filters, UF cleaning chemicals, RO antiscalant, clean-in-place reagents, membranes, neutralisation and laboratory analysis. Confirm storage life and local availability. A process that depends on an imported proprietary chemical with a long lead time needs suitable stock and an approved alternative.
Availability calculations should include cleaning duration and common utilities. Nominal N+1 membrane trains do not provide N+1 performance if they share one feed pump, chemical system or cleaning skid that can disable both. Review failure modes at the process boundary.
The selected concept should show a clear performance margin without excessive conservatism. Document why each stage exists, which constituent controls it and what would allow it to be bypassed or reduced in a future optimisation.
12. Plan safe operation and maintenance
Operator tasks should be designed into the layout. Provide safe chemical delivery, bunding, ventilation, lifting access, sample sinks, membrane removal routes and drainage. Locate instruments where they can be isolated and calibrated without interrupting every treatment train.
Operating procedures should cover start-up, shutdown, extended standby, clean-in-place, membrane preservation, off-spec diversion and backup transition. Training should use the actual cause-and-effect logic and include abnormal scenarios.
Maintain baseline normalised membrane data and review trends at a defined frequency. Trigger investigation before performance reaches the cleaning limit. This supports planned intervention and avoids a sudden loss of capacity during peak cooling demand.
13. How Crescent Engineering can support the project
Crescent Engineering can compare treatment concepts, model recovery and concentrate, define modular packaged equipment, prepare controls and commissioning plans, and provide lifecycle service support. The scope can begin with bench or pilot testing where the feed envelope contains material uncertainty.
14. References and further reading
Final membrane design, chemical programme and residual disposal must be validated against representative water and local requirements.
Related Crescent articles
Continue exploring this topic:
Reclaimed Water for AI Data-Centre Cooling in the GCC: An Engineering Guide
Municipal recycled water for AI data-centre cooling: a Middle East & UAE design guide
Feed-water variability: designing a treatment envelope for recycled cooling water
Storage, blending and monitoring for reliable recycled cooling water
Protecting cooling assets from fouling and corrosion with recycled water
Recycled-water utility interfaces and compliance for GCC data centres
Reclaimed-water applications for data-centre cooling in the UAE
Comments