Membrane cleaning: the chemicals, and how frequency is set

Membrane cleaning is scheduled by pressure, not by the calendar

MBR membrane cleaning maintains a hydraulic property. A membrane is not dirty in the sense that a filter is dirty; it has accumulated resistance, and cleaning is how that resistance is removed. The interval between cleans is therefore not a date but a response to the transmembrane pressure trend, the same record that names the fouling mechanism in the companion article on the four mechanisms of membrane fouling.

A calendar regime is convenient and wrong in both directions. Clean too often and the plant spends membrane life, because every chemical clean consumes some of the tolerance the module has; clean too rarely and a deposit a light clean would have lifted has compacted into one that only a recovery clean can reach, which costs more life still. Neither error appears in the cleaning log. Both appear in the replacement date.

Two kinds of cleaning: maintenance and recovery

A membrane cleaning programme has two chemical layers, and they do different jobs. Physical cleaning runs between them and is not a chemical step at all.

• Maintenance cleaning — a light clean at low chemical strength, run frequently and kept short. It holds the fouling rate down and postpones the heavier clean rather than replacing it.
• Recovery cleaning — a heavier clean at a higher concentration, run when the hydraulic performance has fallen far enough to justify it. Its job is to bring permeability back towards the clean baseline, and it is the clean whose result is measured.
• Physical cleaning in between — relaxation, backwash or air scour. In a submerged membrane train the routine step is usually a relaxation, where permeate production stops while the air scour continues, rather than a true reverse backwash. Where a backwash is used it runs for a shorter time at a modest multiple of the filtration flux. Which of the two the module takes is a property of the module, not a preference.

The two layers are often confused, and the confusion is expensive. A maintenance clean that is run at recovery strength burns oxidant exposure for a result the maintenance interval did not need; a recovery clean that is run at maintenance strength returns little permeability and is then recorded as a failed clean, when what failed was the dose. The maintenance layer exists to protect the recovery layer, and the recovery layer exists to reset the baseline the maintenance layer works from.

The membrane cleaning chemicals, in three families

Almost every regime is built from three families, and most plants use more than one.

• Oxidants — sodium hypochlorite is the usual choice. It attacks organic and biological deposits: it oxidises the extracellular polymers that hold a biofilm together and disrupts the cells inside it. It is also the family with the most consequence, because the oxidant attacks the membrane as well as the deposit, and every module carries a cumulative exposure it is rated for.
• Acids — citric acid is the common choice, with hydrochloric or phosphoric acid where the scale is heavy. They dissolve carbonate scale, hydroxide deposits and the oxides of iron and manganese. They do not touch calcium sulphate or silica, which need a sequestrant rather than an acid.
• Alkalis and surfactants — caustic soda, often with a surfactant or a chelating agent. An alkali clean saponifies fats, disperses the organic gel and lifts what an oxidant has already broken up. It is a separate step from the oxidant clean rather than a substitute for it.

The values usually quoted for these cleans — of the order of 0.05 to 0.5 per cent active chlorine for a maintenance clean and up to about 1 per cent for a recovery clean, of the order of one per cent for a citric acid clean, contact times of the order of an hour or more, and solutions warmed to somewhere around 30 to 40 °C — should be read as a starting method rather than as a specification. The useful band depends on the deposit, on the module and on the water, and the concentrations and contact times that suit your membrane module are confirmed with the factory for your membrane module and your deposit.

Two rules sit outside that flexibility. An oxidant and an acid are never mixed in the same tank, because the mixture releases chlorine gas; and no clean exceeds the chemical exposure the module is rated for, because that exposure is consumed permanently whether or not the clean works.

Cleaning in place: what the sequence does

A cleaning in place sequence on a membrane train has the same shape whatever the chemistry, and the steps that are skipped are usually the ones that decide the result.

• Isolate and drain the module or the tank, so the cleaning solution is not diluted by the process water around it.
• Prepare the solution at the concentration and temperature the module allows, in water that will not itself scale the membrane.
• Soak or recirculate for the contact time, keeping the solution in contact with the whole membrane area; some designs recirculate through the module, others fill the tank and soak.
• Drain and rinse to the point where the permeate is chemically neutral and free of residue, because residual oxidant keeps working on the membrane after the clean is over.
• Return to service and measure the permeability recovered, which is the number that decides whether the clean is repeated, changed, or recorded as having failed.

Some designs take a recovery clean out of place instead, removing the cassette and cleaning it in a vessel the plant can heat and dose properly. Which route the module takes is a property of the module, and the recovery clean is the one worth doing properly, because it is the only one that resets the baseline.

How membrane cleaning frequency is set from the TMP trend

The frequency comes out of the pressure record, and setting it is a sequence rather than a guess.

• Establish a clean baseline. After the first recovery clean, record the permeability the membrane reaches and use that as the reference; a baseline taken on a new membrane that has never been cleaned is optimistic.
• Choose a trigger as a departure from the baseline. A fall of the order of 20 to 40 per cent in permeability, or a rise to the pressure ceiling the design permits, is the band commonly used to call a recovery clean.
• Trigger on the slope as well as the level. An accelerating fouling rate calls for a look upstream rather than simply an earlier clean.
• Let the maintenance interval follow the same record: when the fouling rate between recovery cleans rises, shorten the interval rather than adding a second chemistry.
• Require a result from every clean. A clean that returns no permeability has consumed exposure and taught the plant only that the chemistry was wrong.

A submerged membrane train is normally designed to run within a few tenths of a bar of transmembrane pressure, and that ceiling is a design decision rather than a universal figure. The permeability that counts as clean, the ceiling that applies and the interval that follows from them are confirmed with the factory for your membrane module and your water.

What the cleaning chemistry will not tolerate

Every membrane module is rated for a pH band, a maximum temperature and a cumulative oxidant exposure, usually expressed as concentration multiplied by time. The first two are recoverable if they are exceeded briefly; the third is not. Oxidant exposure is consumed permanently, which is why a cleaning procedure that works is not automatically a procedure to repeat indefinitely, and why a clean that has to be run at a higher concentration every time is telling the plant something about the deposit rather than about the dose.

The pH band, the temperature limit and the cumulative exposure are published for each module, and it is worth reading them together with the cleaning record rather than once at commissioning, because the record is what says how much of the allowance has been spent.

Common mistakes

• Cleaning on a calendar while the pressure record is uncorrected for temperature. A winter rise that a viscosity correction explains is not a reason for a chemical clean.
• Choosing the chemistry by habit rather than by the deposit. An acid clean on an organic gel, or an oxidant clean on a carbonate scale, both leave the deposit where it was and charge the membrane for the attempt.
• Mixing an oxidant and an acid in one tank, or in one sequence without an intermediate rinse.
• Answering a clean that did not work with a stronger version of the same clean, rather than with a different family.
• Waiting for the pressure ceiling before the first recovery clean, by which point the deposit has had months to compact.
• Judging a clean by the pressure read the moment it finishes. Permeability needs a day of operation to settle, and a clean assessed too early is usually judged to have failed.

Self-check before the cleaning regime is fixed

• Is permeability calculated from pressure and flow together, and corrected for temperature?
• Is there a clean baseline recorded after a recovery clean, rather than after commissioning?
• Is the recovery trigger a departure from that baseline, and is it written down?
• Is the deposit identified — organic, biological, carbonate, sulphate — before the chemistry is chosen?
• Is the membrane cleaning procedure recorded with concentration, temperature and contact time, so that a change of operator does not change the dose?
• Is the oxidant exposure spent so far being counted against what the module is rated for?

From a cleaning regime to a membrane duty

A cleaning regime is the operating half of a membrane plant, and the cleaning half of its running cost. The companion article on the four mechanisms of membrane fouling explains what the pressure trend is showing and which mechanism each shape belongs to. The integrated MBR equipment page sets out how the membrane, the biology and the air scour are packaged together, and the sequence, the concentration and the frequency that suit your membrane module are confirmed with the factory for your membrane module and your water rather than copied from a general table.