Membrane Fouling and Cleaning: Diagnosis, CIP Chemistry and Protocol Design

What are the four membrane fouling mechanisms?

Membrane fouling is not one phenomenon but four, each with different chemistry, different diagnostic signatures and — critically — different cleaning agents. Treating them as interchangeable is the most common and most expensive mistake in membrane operation. The four recognised mechanisms are:

• Particulate / colloidal (cake) fouling. Suspended solids, colloids and flocs are convected to the membrane and deposit as a cake or gel layer that adds hydraulic resistance in series with the membrane itself. This is largely a hydrodynamic problem governed by the balance of permeation drag against back-transport (shear-induced diffusion, inertial lift and Brownian diffusion).
• Organic fouling. Natural organic matter (NOM), humic and fulvic acids, proteins, polysaccharides and other extracellular polymeric substances (EPS) adsorb onto and into the membrane. Adsorption is thermodynamically driven and can occur even at zero flux, which is why organics foul membranes that particulates never reach — inside the pore walls.
• Inorganic scaling (precipitative fouling). As permeate is withdrawn, sparingly soluble salts concentrate in the boundary layer beyond their solubility limit and crystallise. Calcium carbonate, calcium sulphate, barium and strontium sulphate, calcium phosphate and silica are the usual culprits. Scaling is a concentration-polarisation-driven supersaturation problem, most severe on the last elements of an RO/NF train.
• Biofouling. Viable microorganisms attach, grow and secrete EPS to form a hydrated biofilm. Unlike the other three, biofouling is self-replicating — a 99.9% kill still leaves enough survivors to regrow. It is the hardest fouling to control because you cannot filter your way out of a living, reproducing deposit.

In real plants these overlap: a biofilm traps colloids, scale nucleates on organic conditioning films, and cake layers create the stagnant micro-environment that biofilms exploit. Diagnosis therefore means identifying the dominant mechanism, because that is what dictates the cleaning chemistry.

Reversible vs irreversible fouling and the resistance-in-series model

The cleanest way to frame fouling quantitatively is the resistance-in-series model, which treats every layer opposing flow as a hydraulic resistance added to the intrinsic membrane resistance. Flux is then simply the transmembrane pressure driving force divided by the product of permeate viscosity and total resistance.

The value of the model is the distinction it forces. Reversible fouling (R_c, and the polarisation term R_p) is removed by physical action — relaxation, backwash, air scour — because it sits loosely on the surface. Irreversible fouling (R_f) is chemically bound or lodged inside pores and only chemical cleaning removes it. What backwash cannot recover, CIP must; what CIP cannot recover has become irrecoverable fouling, and that fraction sets the membrane's ultimate service life.

Operationally you track this by measuring flux (or permeability) immediately before and after each backwash. The ratio recovered by physical cleaning is the reversible fraction; the residual, slowly climbing baseline is the irreversible accumulation that eventually triggers a CIP. When even CIP fails to return permeability to within roughly 85–90% of the original clean-water value, the membrane is approaching end of life. Sound membrane system design and integration builds this monitoring in from day one rather than bolting it on after fouling appears.

How do you read flux decline and TMP rise? (normalisation)

Raw flux and TMP readings lie, because they drift with feed temperature, pressure and recovery even when the membrane is unchanged. Viscosity alone changes flux roughly 2–3% per °C. To separate genuine fouling from operating-condition noise you must normalise — and the standard tool is the temperature-corrected specific flux, or permeability.

Two failure modes then read differently. On a constant-flux plant (the usual UF/MBR mode), fouling shows as a rising TMP at fixed permeate rate — the pump works harder to push the same water through a more resistive membrane. On a constant-pressure plant, fouling shows as declining flux at fixed TMP. In both cases the normalised permeability K₂₀ falls, and it is the single most reliable fouling indicator you can trend. A useful decision rule: a sustained 15% drop in normalised permeability, or a 15% rise in normalised TMP, is a common trigger to schedule a CIP.

Worked example — normalised specific flux. A UF train runs at J = 60 LMH and TMP = 0.9 bar with feed at 12 °C. Normalising: TCF = 1.03^(20−12) = 1.03⁸ = 1.267. Raw specific flux = 60 / 0.9 = 66.7 LMH/bar. So K₂₀ = 66.7 × 1.267 = 84.5 LMH/bar. Three months later the same train reads J = 60 LMH but TMP = 1.35 bar at 20 °C: K₂₀ = (60/1.35) × 1.0 = 44.4 LMH/bar. Permeability has fallen from 84.5 to 44.4 — a 47% loss that the raw flux (unchanged at 60 LMH) completely concealed. That is why you never trend raw numbers.

What do the pore-blocking (Hermia) models tell you?

To go beyond “the membrane is fouling” and identify how, engineers fit dead-end flux-decline data to the classical Hermia blocking models. All four derive from a single differential form — the rate of TMP or resistance change scaling with a power n of the flow — but each corresponds to a distinct physical picture of where the foulant goes.

The practical payoff is diagnostic and it is real: internal (standard) blocking implies foulant inside the pores that backwash cannot reach and that will demand chemical cleaning, whereas cake filtration implies a surface layer that physical cleaning should largely recover. Fouling commonly evolves through the sequence — pore blocking first, then cake growth once the surface is covered — so a log-log plot of the decline data that changes slope is itself telling you the mechanism has shifted. That shift is a signal to change either the pretreatment or the cleaning strategy, not merely to clean harder.

Critical and sustainable flux: designing to foul slowly

The single most powerful fouling-control idea is that fouling is a function of how hard you drive the membrane. The critical flux, defined by Field and co-workers in 1995, is the flux below which permeability stays essentially stable over time; above it, fouling accelerates sharply. It is the flux at which permeation drag toward the membrane just balances the back-transport of particles away from it.

Critical flux is measured by the flux-stepping method: hold flux constant for a fixed interval, record the TMP slope, step up, repeat. The flux at which the TMP slope stops being flat and starts climbing is J_crit. Everything else about anti-fouling operation — crossflow velocity, air scour in an MBR, backwash frequency — is really an effort to raise the critical flux by strengthening back-transport, so you can run more productively before fouling takes hold. The same logic underpins the flux ceilings used in membrane plant process design and it is why over-fluxing to hit a production target almost always costs more in cleaning and membrane replacement than it saves.

Cleaning hierarchy: backwash, CEB and CIP

Cleaning is a staged escalation, cheapest and gentlest first. You do not jump to a full CIP for reversible cake fouling any more than you backwash a scaled-up RO element and expect a result.

• Physical cleaning — relaxation and backwash. In UF/MF and MBRs, permeation is paused (relaxation) or reversed (backwash), often combined with air scour, every few minutes to tens of minutes. This lifts the reversible cake (R_c). It uses no chemicals and takes seconds to minutes, so it runs continuously as the first line of defence. RO/NF spiral elements cannot be backwashed — their only physical tool is a low-pressure forward flush.
• Chemically enhanced backwash (CEB). A backwash dosed with a low concentration of chemical (typically hypochlorite, acid or caustic at tens to a few hundred mg/L) performed every few hours to daily. CEB attacks the incipient irreversible layer before it consolidates, dramatically stretching the interval between full CIPs. It is short (minutes) and largely automated.
• Clean-in-place (CIP). The heavy chemical clean, run when normalised permeability has fallen past the trigger despite CEB — typically weeks to months apart. Concentrated cleaning solution is recirculated, often warm, through the modules for 30–120 minutes with a soak, then rinsed. CIP is where matching cleaner to foulant is decisive.

The economic logic is that each tier defends the next: good hydraulics keep the cake reversible so backwash copes; frequent CEB keeps the irreversible layer thin so CIPs are rare; effective CIP keeps the irrecoverable fraction low so membranes last their full 5–10 year life. Neglect any tier and the cost cascades down to premature replacement.

CIP chemistry: matching the cleaner to the foulant

This is the crux of the whole subject. The wrong cleaner does nothing at best and, at worst, sets fouling permanently — caustic will precipitate calcium and fix a scale you meant to dissolve. The rule is simple: acids for minerals, alkalis and oxidants for the organic and the living.

Sequence matters as much as selection. For mixed fouling the near-universal order is alkaline/oxidant first, acid second: the caustic-hypochlorite step degrades the organic and biofilm matrix that binds the deposit together, exposing the mineral scale beneath so the subsequent acid can reach and dissolve it. Reverse the order and the acid merely bounces off an organic conditioning film. Always check the membrane manufacturer's chemical-compatibility envelope first — polyamide RO membranes are destroyed by free chlorine (their tolerance is measured in a few hundred to ~1000 ppm-hours over the membrane life), so hypochlorite that is routine on a PVDF UF membrane is forbidden on a thin-film RO element.

Scaling indices and antiscalant: preventing what you can

The cheapest cleaning is the clean you never have to do, and for scale that means predicting supersaturation before commissioning. The classic screen for calcium carbonate is the Langelier Saturation Index (LSI); for high-TDS RO concentrates the Stiff–Davis index is used instead, and specific salts (CaSO₄, BaSO₄, silica) are checked against their own solubility products at the concentrate condition.

Because RO concentrates the feed, you must evaluate the index at the concentrate composition, not the feed. At 75% recovery the concentration factor is CF = 1/(1−r) = 1/(1−0.75) = 4, so calcium and alkalinity in the reject are roughly four times feed values — and it is the reject-side LSI that governs scaling on the tail elements. Antiscalants (phosphonates, polyacrylates) do not stop supersaturation; they work sub-stoichiometrically by adsorbing onto crystal nuclei to poison crystal growth and disperse micro-crystals, allowing the plant to run safely at a positive LSI it could otherwise never tolerate. This is the same recovery-vs-fouling trade central to reverse-osmosis train design, and it is explored further for RO trains in our companion articles.

Worked example — CIP dose and volume. Size an alkaline CIP for a UF skid holding 1.5 m³ of module and pipework volume, targeting a 0.2% w/v NaOH solution. Mass of NaOH = 0.002 × 1500 L × 1000 g/L × 0.01 … taken directly: 0.2% w/v = 2 g/L, so 2 g/L × 1500 L = 3.0 kg NaOH (or 7.5 L of a 40% w/w, ~1.43 kg/L caustic stock, i.e. 3.0/(0.40×1.43) ≈ 5.2 L). Add hypochlorite to 500 mg/L free chlorine: 0.5 g/L × 1500 L = 750 g available chlorine, delivered from a 12.5% w/v NaOCl stock as 750/125 = 6.0 L of stock. Recirculate at high pH and 30–35 °C for 45–60 min, soak, then flush to drain within the site's trade effluent consent limits before returning to service.

Cleaning protocol design and membrane autopsy

A CIP protocol is an engineered procedure, not a bucket of chemicals. Six variables set its effectiveness, and each has a defensible optimum: chemistry (matched to the diagnosed foulant), concentration (enough to react, within the membrane compatibility limit), temperature (reaction rates roughly double per 10 °C, so warm cleaning is faster — but capped by the membrane's thermal limit, often 35–40 °C), crossflow / mechanical energy (to shear loosened foulant away), contact time (recirculate-and-soak cycles), and pH control during the clean. Optimising these is applied reaction engineering: you are running a heterogeneous reaction at a solid–liquid interface and every lever that raises reaction rate or removes product helps.

When cleaning stops recovering permeability, the definitive diagnostic is a membrane autopsy — a sacrificial element is cut open and the foulant characterised. A typical autopsy runs: visual and dye inspection; loss-on-ignition (organic vs inorganic split by combustion); SEM-EDX for elemental mapping of the deposit (calcium and carbon point to carbonate, sulphur to sulphate scale, silicon to silica); FTIR for organic functional groups; and ATP or culture assays for viable biomass. The autopsy tells you which of the four mechanisms actually dominated, closing the loop: it validates — or corrects — the cleaning chemistry, the pretreatment, and the flux ceiling for the next design cycle.

Fouling control is ultimately a plant-wide discipline, not a membrane-module concern. It reaches back into pretreatment selection — whether an ultrafiltration stage should protect a downstream reverse osmosis train, or whether an MBR versus MBBR choice better suits a fouling-prone wastewater — and forward into the CIP waste handling and neutralisation that keep the site compliant. Get the diagnosis right and the cleaning follows; get it wrong and no amount of chemical fixes it.