Reverse Osmosis System Design: A Rigorous Engineering Guide

What data do you need before designing an RO plant?

RO design is a mass-balance and transport problem, so it starts with a full feed-water analysis, not a catalogue flow rate. Assemble:

• Full ionic analysis — Ca²⁺, Mg²⁺, Na⁺, HCO₃⁻, SO₄²⁻, Cl⁻, SiO₂, and total dissolved solids (TDS, mg/L) — this sets osmotic pressure and scaling limits.
• Design permeate flow and required permeate quality (TDS or specific-ion limits).
• Feed temperature range — water viscosity and the membrane permeability coefficient are strongly temperature-dependent.
• Fouling potential — Silt Density Index (SDI₁₅), turbidity, TOC, iron and manganese.
• Target recovery — the fraction of feed converted to permeate, which drives concentrate volume and scaling risk.

These feed the transport equations below. The right RO membrane selection and array design depends on getting this characterisation right first.

How much pressure does osmosis demand? (osmotic pressure)

Osmosis drives water from low to high solute concentration; RO reverses it by applying pressure above the osmotic pressure difference. For dilute solutions the van 't Hoff relation gives a good estimate:

The dissociation factor i accounts for the number of ions released: NaCl gives i ≈ 2, CaCl₂ gives i ≈ 3. Osmotic pressure is colligative — it depends on the count of dissolved species, not their identity — which is why a high-TDS brackish water needs far more feed pressure than the same flow of low-salinity surface water.

How does water actually pass through the membrane? (solution-diffusion)

A dense polyamide RO membrane has no pores in the conventional sense; water and salt dissolve into the film and diffuse across it under their own gradients. This is the solution-diffusion model, which treats water flux and salt flux as independent:

Two consequences matter for design. First, water flux is proportional to net driving pressure, but salt flux is not — so raising pressure or lowering Δπ dilutes the permeate and improves rejection. Second, observed salt rejection R = 1 − C_p/C_f therefore rises with flux, which is why undersized (over-fluxed low-pressure) systems and high-flux fouled elements both pass more salt.

Worked example: osmotic pressure and net driving pressure

Take a brackish groundwater at 3,000 mg/L TDS, 25 °C (298 K), and a plant run at 75% recovery against a feed pressure of 15 bar with a membrane rated A = 3.5 LMH/bar.

• Feed osmotic pressure: π_f ≈ 0.077 × (3,000 / 1,000) = 2.3 bar.
• Concentration factor at 75% recovery: CF = 1 / (1 − r) = 1 / (1 − 0.75) = 4 (with near-complete rejection). Concentrate TDS ≈ 3,000 × 4 = 12,000 mg/L.
• Concentrate osmotic pressure: π_c ≈ 0.077 × 12 = 9.2 bar. A log-mean average across the array is roughly (2.3 + 9.2)/2 ≈ 5.3 bar of osmotic back-pressure to overcome at the tail.
• Net driving pressure (tail element): NDP ≈ ΔP − Δπ = 15 − ~5 (osmotic) − ~2 (pressure drop + permeate back-pressure) ≈ 8 bar.
• Water flux at the tail: J_w = A × NDP = 3.5 × 8 = 28 LMH — comfortably within the ~15–30 LMH design band for brackish water.

The lesson: it is the concentrate-end osmotic pressure, not the feed value, that limits recovery. Push recovery higher and π_c climbs steeply, NDP collapses, and the tail elements stop producing water.

Why does salt pile up at the membrane wall? (concentration polarisation)

Water passes through the membrane but rejected salt does not, so a concentrated boundary layer builds up against the wall. The wall concentration exceeds the bulk by the concentration-polarisation factor β:

Polarisation is harmful twice over: it raises the true osmotic pressure the membrane sees (cutting real NDP) and it pushes the wall concentration of sparingly soluble salts toward saturation, triggering scale. It is controlled by maintaining turbulent cross-flow — hence the minimum concentrate flow and maximum recovery-per-element limits every membrane maker publishes. A 2:1 array (see below) exists partly to keep k high as flow drops along the train.

How is recovery, staging and scaling set?

System recovery r = Q_permeate / Q_feed is the master economic variable: higher recovery means less feed, less pretreatment chemical and a smaller concentrate to dispose of — but a higher concentration factor CF = 1/(1−r) and steeper scaling risk. Because each 8-inch element recovers only ~10–15%, achieving 75% overall needs elements in series (6–7 per pressure vessel) and vessels in stages.

As permeate is removed, concentrate flow falls, so later stages use fewer parallel vessels to hold cross-flow velocity up — the classic 2:1 array (e.g. two first-stage vessels feeding one second-stage vessel). Scaling is screened with saturation indices before committing to a recovery:

• Langelier Saturation Index (LSI) for CaCO₃ in brackish water; the Stiff & Davis Index (S&DSI) for higher-ionic-strength/seawater concentrate. Positive values indicate a scaling tendency.
• CaSO₄, BaSO₄, SrSO₄ and CaF₂ checked against their solubility products at the concentrate-side ionic strength.
• Silica limited to roughly 120–150 mg/L as SiO₂ in the concentrate (temperature- and pH-dependent), or higher with antiscalant.

Antiscalant dosing and acid addition raise the allowable recovery by suppressing nucleation; a full projection program models each element to keep every stage below saturation.

What sets the energy demand?

RO energy is dominated by the high-pressure feed pump. The specific energy consumption (SEC, kWh/m³ of permeate) is:

There is a hard thermodynamic minimum energy of separation — the least work to produce permeate equals the osmotic pressure of the concentrate at the exit. For seawater at 35,000 mg/L and 50% recovery this floor is about 1.06 kWh/m³; real plants run 2–3× that because of over-pressure, polarisation and inefficiency. Isobaric energy-recovery devices (pressure exchangers) recovering >95% of concentrate pressure are what bring seawater SEC close to the limit. Where the concentrate itself has value or a disposal cost, pairing RO with an industrial water reuse scheme can shift the economics of running at higher recovery.

RO vs NF vs UF: which membrane process?

RO is one point on a pressure-driven membrane spectrum. Selecting the right one is a function of the smallest species you must remove and the osmotic penalty you are willing to pay:

UF is a common pretreatment ahead of RO because it delivers a low, stable SDI. NF is chosen when only hardness or organics must go and the full desalination energy of RO is unnecessary. If the duty is a biological effluent rather than a raw water, compare the upstream options in our MBR vs MBBR guide before specifying the membrane polish.

Why does pretreatment decide RO life?

Membrane life and cleaning frequency are governed by what reaches the element, quantified by the Silt Density Index. Design targets are SDI₁₅ < 3 for spiral-wound RO (some warranties require < 5). Pretreatment typically stacks:

• Removal of particulates/colloids — coagulation plus media or UF membranes to drop SDI and turbidity.
• Scale control — antiscalant and/or acid to manage LSI and sulphate/silica saturation at the concentrate end.
• Oxidant management — dechlorination (SMBS) upstream of polyamide, which is destroyed by free chlorine, then a non-oxidising biocide programme for biofouling.
• Iron and manganese removal — below ~0.05 mg/L to avoid metal-oxide fouling.

Fouling and scaling are the two failure modes an RO process design of the pretreatment train must pre-empt; almost every premature membrane replacement traces back to a pretreatment shortfall rather than the membrane itself.

RO system design sequence

1. Characterise the feed. Obtain a full ionic analysis, temperature range, SDI15 and target permeate quality. Compute feed osmotic pressure from TDS.
2. Set target recovery. Choose recovery from the concentration factor CF = 1/(1-r) and screen scaling with LSI/S&DSI, sulphate and silica saturation at the concentrate.
3. Select flux and membrane area. Pick a conservative average flux (15-30 LMH brackish, 12-17 LMH seawater) to fix element count; derate for temperature and fouling.
4. Configure the array. Set elements per vessel (6-7) and stage ratio (e.g. 2:1) to hold concentrate cross-flow up and keep the polarisation factor beta below 1.2.
5. Size the pumps and energy recovery. Compute feed pressure from net driving pressure at the tail, then SEC; add a pressure exchanger for seawater duty.
6. Design pretreatment and validate. Specify coagulation/UF, antiscalant, dechlorination and biocide to hit SDI15 < 3, then confirm the whole design in a projection program.