Membrane Integrity Testing: Proving MF, UF and MBR Barrier Performance

A membrane plant can produce clear water, hold a stable transmembrane pressure and still be unable to prove that its physical barrier is intact. That distinction matters wherever microfiltration or ultrafiltration is credited with removing pathogens, suspended contaminants or product-damaging particles. It also matters in a membrane bioreactor, where one broken fiber, failed seal or open cross-connection can connect mixed liquor to the permeate side without creating an immediate plant-wide alarm.

The operating question is therefore not simply, “Is the filtrate good today?” It is, “What removal capability was qualified, what part of the installed barrier was tested, how quickly can a loss be detected, and what evidence permits a repaired unit to return to service?” A credible membrane integrity testing program answers all four questions. It combines qualification evidence, a physical test of each isolatable unit, continuous water-quality surveillance, fault localization, controlled repair and a documented release decision.

This guide focuses on low-pressure MF and UF systems used in drinking-water treatment, industrial pretreatment and water reuse, including immersed and sidestream MBRs. The principles can inform other membrane applications, but test physics, regulatory credit and acceptance criteria must be matched to the membrane type and jurisdiction. A pressure-decay method designed for wetted hollow-fiber UF modules should not be copied into an RO specification merely because both assets are called membranes.

The barrier claim has three different questions

Integrity programs become confused when three different forms of evidence are treated as interchangeable. Challenge testing, direct testing and indirect monitoring support one another, but each answers a different question.

Qualification asks what an intact device can remove

A challenge test is conducted on a representative membrane device using the target organism or an accepted surrogate under defined operating conditions. Its purpose is to establish the removal capability of an intact product or product family before routine service. The resulting log-removal value is not an eternal property of the polymer alone. It belongs to a tested configuration, lot-selection rationale, operating envelope, analytical method and data-quality package.

For a challenge organism or surrogate, the basic expression is:

LRV = log10(feed concentration) - log10(filtrate concentration)

That simple equation can be limited by the laboratory reporting limit. If the filtrate result is non-detect, the demonstrated LRV is bounded by the method’s quantitation capability rather than proved to be infinite. Procurement teams should therefore ask how non-detects were handled, whether conservative values were used, and whether the tested module represents the commercial production population.

A direct test asks whether the installed physical barrier is intact now

Direct integrity testing applies a physical challenge to an isolatable membrane unit and measures a response associated with leakage. Pressure decay, vacuum decay, diffusive airflow, water displacement and marker-based methods can all serve this role when their resolution and sensitivity are demonstrated. The test does not merely “check the membrane.” It evaluates the pressure boundary that participates in the test, which may include fibers or sheets, potting, O-rings, module connections, housings and common piping.

A passing result is a time-stamped statement about a defined unit under defined test conditions. It does not guarantee that a break cannot occur five minutes later, and it does not validate components that were bypassed or excluded by the valve lineup.

Indirect monitoring asks what changed between physical tests

Indirect integrity monitoring observes filtrate quality—commonly turbidity or particle counts—as an indicator of particulate removal. It provides the temporal coverage that an offline physical test cannot. However, it is generally less sensitive to small breaches, can be affected by instrument noise and may be weakened by dilution when multiple units discharge to one common header.

The three questions form a proof chain:

1. Capability: What can an intact, qualified device remove?
2. Condition: Is each installed barrier unit physically integral at the time of testing?
3. Continuity: Is there evidence of a performance change between physical tests?

None of the links can replace the others. A strong challenge result cannot prove that an installed seal is correctly seated. A passing physical test cannot establish removal capability for a target that was never challenged. Clear filtrate cannot prove that a high-removal claim remains valid at the detection limit required for regulatory credit.

Draw the complete barrier boundary before choosing a test

The most consequential design step is often a drawing exercise. Mark the untreated side, the treated side and every component that can create a path between them. Then identify which components are stressed and observed during the integrity test. If the marked physical boundary and the tested boundary are different, the gap needs another control.

A complete low-pressure membrane barrier may include:

• hollow fibers, flat sheets or tubular membrane surfaces;
• potting resin, glue lines and module end caps;
• module-to-manifold seals and O-rings;
• pressure housings, cassettes, headers and permeate collectors;
• filtrate, backwash and chemically enhanced backwash valves;
• air-scour or integrity-test connections that can cross-connect streams;
• sample lines, drains, vents and temporary hoses;
• shared filtrate piping before the final compliance sample point; and
• software states that determine valve position during filtration, backwash and testing.

This boundary review changes the failure vocabulary. “Broken fiber” is only one mechanism. A rolled O-ring, cracked potting block, leaking isolation valve, incorrectly installed module, open manual bypass or failed valve feedback can produce the same treated-side contamination. Strong membrane breach detection is therefore designed around pathways rather than around a single favored component.

Freeze the qualification basis before commissioning

The barrier claim should be converted into a controlled basis document before equipment is accepted. It should state the target organisms or particles, credited removal, membrane identity, configuration, production lot strategy, challenge conditions, analytical methods and any safety factor applied by the regulator or owner.

Demand representative challenge evidence

A supplier’s best laboratory module is weak evidence for a fleet unless the relationship to production modules is documented. Ask how modules were selected, whether manufacturing variability was considered, what water temperature and flux were used, and whether the test covered the least favorable credible condition. If multiple membrane formulations, module lengths or potting designs are supplied, determine whether they belong to one defensible product family or require separate qualification.

Also identify what a future change invalidates. A new membrane chemistry, altered module geometry, different potting formulation, new supplier plant or changed integrity-test algorithm may require technical review. Change control should not begin after the first unexplained failure.

Link removal credit to a test that can verify it

An ambitious challenge-test LRV is operationally useful only if the installed test can verify the required level. The maximum qualified removal, the sensitivity of the physical test and the removal credit granted by the applicable authority are related but distinct. The credited value should not exceed what the whole evidence chain can support.

This is where membrane integrity verification becomes an engineering calculation rather than a checkbox. For a pressure- or vacuum-based method, the smallest reliably measurable response must be related to potential water flow through a breach during filtration. For a marker-based method, feed and filtrate marker concentrations and analytical limits define sensitivity. In both cases, instrument resolution, system volume, operating flow, pressure, temperature and concentration assumptions should be traceable.

Design the physical test around four performance properties

A test name is not a test design. Two systems can both use a “pressure hold” sequence yet have radically different abilities to find a meaningful defect. A defensible membrane integrity test procedure defines at least resolution, sensitivity, control limit and frequency.

Resolution: what is the smallest relevant breach?

Resolution is the size of the smallest breach that contributes to a measurable response. In U.S. drinking-water applications receiving Cryptosporidium removal credit under 40 CFR 141.719, the direct method must have a resolution of 3 micrometers or less. That regulatory value is not a generic specification for every industrial application, but it demonstrates the right design logic: the test must respond to a defect relevant to the hazard being controlled.

For a gas-pressure test on a fully wetted porous membrane, applied pressure must overcome the capillary force holding water in the defect, plus relevant backpressure and baseline effects. A test that never reaches the required net pressure may appear beautifully stable because it has not opened the defect it claims to detect.

Sensitivity: can the measured response support the claimed removal?

Sensitivity is the maximum removal level the test can reliably verify. It depends not only on the pressure sensor but also on pressurized volume, test duration, baseline leakage, membrane-unit design flow and the conversion between gas flow during the test and water flow during filtration. Increasing the number of modules in one test unit can make a very small leak harder to distinguish, even though the plant has more membrane area.

Control limit: what separates pass, investigate and fail?

The control limit must sit within demonstrated test capability and correspond to an integral unit able to meet its approved duty. A manufacturer’s default should be reviewed against the actual installed volume, sensor range, piping, elevation, temperature and awarded credit. One number may not be sufficient for every operating state. A program can use a warning zone below the formal failure limit to prompt investigation before compliance is lost.

Frequency: how much unobserved exposure is acceptable?

Test frequency is a risk decision constrained by applicable rules. U.S. systems operating under 40 CFR 141.719 generally conduct a direct test on each operating unit at least daily, unless the state approves a different frequency under the rule’s conditions. Other countries, reuse permits and industrial contracts may set different requirements. More frequent testing reduces the maximum time between proofs but creates downtime, valve cycles and production loss. Continuous indirect signals help manage the interval; they do not automatically authorize a longer one.

Choose a direct method by physics, not familiarity

Pressure or vacuum decay

A pressure decay test isolates a wetted unit, applies gas pressure or vacuum to a defined side and measures change over time. It is widely suited to hollow-fiber MF/UF systems because a sufficiently large defect allows gas to displace water and pass through. The method can be automated and applied rack by rack, but its credibility depends on stable isolation, adequate wetting, known test volume, calibrated pressure measurement and a repeatable stabilization period.

Common false-failure causes include residual temperature equilibration, leaking valves, trapped gas, incomplete wetting and testing too soon after a reverse-flow event. Common false-pass causes include insufficient test pressure, an untested bypass, a sensor with poor resolution, an incorrect system-volume value or software that averages away a short high-decay period.

Diffusive airflow and water displacement

Rather than infer leakage from pressure change, some methods measure gas flow or displaced water. Direct flow measurement can improve diagnostic clarity, but meter range, gas conditions, condensate and background diffusion still require control. A small flow meter installed for a pilot module may not retain adequate resolution when scaled to a large rack.

Marker-based methods

A particulate or molecular marker can be introduced on one side and measured on the other. This can connect the test more directly to separation performance and may suit configurations where a pressure method is difficult. It also introduces marker preparation, mixing, recovery, analytical detection, background concentration, contamination and waste-management questions. A marker method is not inherently superior; it shifts the uncertainty from pressure physics toward dosing and analysis.

Build a trustworthy baseline before writing alarms

Commissioning establishes the reference behavior of intact units. A baseline should not be a single vendor demonstration performed while the project team watches. It should include repeated tests across representative temperature, operating and post-backwash conditions, using calibrated instruments and the final control software.

For each unit, retain:

• asset identity, membrane model, serial or lot data and module position;
• test-state valve lineup and confirmation feedback;
• initial, minimum and final pressure or vacuum;
• stabilization time, measurement duration and sampling rate;
• water temperature and applicable correction method;
• tested gas or liquid volume and calculation version;
• raw sensor series, not only the final pass/fail bit;
• baseline response, warning threshold and formal control limit; and
• the software, firmware and configuration revision used.

Baseline data should be traceable and comparable. If an algorithm changes after a SCADA upgrade, the historical series may no longer mean the same thing. Version the calculation and preserve enough raw data to reconstruct a result. A stable green icon is not an auditable record when no one can explain how it was calculated.

Use continuous monitoring as a sentinel, not a substitute

Between physical tests, unit-specific filtrate monitoring can detect a gross breach or changing removal performance. For drinking-water low-pressure membranes, current EPA optimization guidance discusses continuous individual-filter-effluent turbidity or particle-count goals that are more stringent than the federal rule’s basic trigger framework. Those values are operational optimization goals, not permission to overwrite a plant’s approved regulatory limits. Owners should preserve the distinction between regulation, permit, optimization target and internal early-warning value.

Turbidity

Online turbidity is familiar, continuous and useful for large particulate excursions. At very low values, sample bubbles, fouling, flow variation, stray light and instrument maintenance can dominate the reading. The sample must represent one membrane unit if it is expected to localize an event. A common-header turbidimeter may show an acceptable average while one train is releasing off-spec filtrate.

Particle counting

Particle counters can be more responsive to discrete particle passage and size ranges, but coincidence error, sample tubing, bubbles and counting statistics matter. Alarm design should use verified instrument performance at the low concentrations expected from intact membranes.

Hydraulic indicators

Permeability, flux, TMP and pressure drop are essential for fouling and operability management, but they are not direct proof of barrier condition. A single fiber break can have negligible effect on total hydraulic resistance. Conversely, a fouled but physically intact unit can have poor permeability and still retain particles. Integrity and fouling dashboards should be connected for diagnosis without being conflated.

This is particularly important when interpreting ultrafiltration membrane integrity. A UF rack can meet a permeability target after cleaning yet fail the physical test because a seal was disturbed. It can also show declining permeability while continuing to pass the physical test because the problem is deposition rather than breach.

Turn an alarm into a controlled state transition

An alarm is useful only if the plant has predetermined what happens to water, equipment and evidence. The response should be designed as a state model rather than left as an operator memory test.

State 1: Normal production

The unit has a current passing direct test, indirect signals remain inside approved limits, instruments are healthy and no unreviewed maintenance has opened the barrier boundary.

State 2: Warning or suspect

A trend, single reading, test-quality flag or maintenance event suggests uncertainty. The control system preserves the data, checks instrument health and increases scrutiny. Depending on risk, filtrate may continue temporarily under an approved response or be diverted immediately.

State 3: Isolate and contain

A failed direct test, sustained indirect excursion or confirmed cross-connection removes the affected unit from service. Automatic valves prevent suspect filtrate from entering finished-water storage or product distribution. The plant identifies the earliest possible event time and assesses water produced since the last credible proof.

State 4: Diagnose and localize

The team distinguishes a real barrier defect from a test artifact, then narrows the problem from train to rack, cassette, module, seal, fiber or piping path. Evidence is collected before repair erases the failure signature.

State 5: Repair, challenge and release

The repair is documented, the unit is flushed as required, and the direct test is repeated. Return to service occurs only after the result meets the approved control limit and any quality-hold samples or operational checks are cleared. A manual “looks good” reset is not release authority.

The state sequence makes responsibilities explicit. Operations controls isolation and water disposition; maintenance performs work; laboratory staff confirm sample validity; engineering owns test calculations and change control; quality or the designated authority approves release where required.

Localize the fault without destroying the evidence

A unit-level failure creates a localization problem. Repeatedly running the same full-rack test may confirm that the signal is reproducible, but it does not reveal which component failed. A practical hierarchy moves from large to small while maintaining chain of custody.

1. Validate the test. Confirm sensor calibration, wetting, temperature stability, gas supply, valve positions and raw data.
2. Repeat under controlled conditions. Use the approved repeat protocol; do not alter pressure or duration informally until the baseline is no longer comparable.
3. Segment the unit. Isolate racks, cassettes or module groups if the design permits.
4. Apply a diagnostic method. Bubble observation, sonic monitoring, conductivity profiling, local pressure testing or manufacturer-specific tools may identify the leak path.
5. Inspect boundary components. Check O-rings, couplings, end caps, potting, headers and valves, not just membrane fibers.
6. Preserve findings. Photograph defects, record module position and retain removed components when a root-cause review is warranted.

Localization capability should be procured with the plant. A design that can isolate only an entire 50-module train may turn one damaged module into a major production outage. Test valves, sample taps and module-position records are inexpensive compared with emergency disassembly conducted without a map.

Repair decisions must protect both quality and capacity

Broken hollow fibers may be pinned or sealed using an approved method. Damaged modules can be replaced, seals reseated and faulty valves repaired. Every intervention changes the asset population. Pinning fibers reduces effective area; mixing old and new modules can alter hydraulic distribution; replacement modules can have different permeability; repeated potting damage may indicate a systemic pressure, chemical or handling problem.

A repair record should include the defect type, exact location, suspected cause, repair method, materials used, number of fibers disabled, module identity, technician, post-repair test result and release approver. Trend repair density by module lot and position. A growing number of pins can remain hidden inside passing unit-level tests while capacity and reliability erode.

Direct testing should also follow events that can disturb the boundary, such as module replacement, seal work, unusual pressure exposure or chemical cleaning when required by the approved program. The point is not to test after every operator touch by habit; it is to identify events that can change integrity and define proof before returning to production.

MBR integrity requires a reuse-quality view

In an MBR, the membrane operates inside or alongside concentrated biological solids. The process can produce low-turbidity permeate, but membrane separation does not automatically prove every reuse requirement. Nutrients, dissolved salts, trace organics and disinfection needs depend on the destination. The earlier guide to membrane bioreactor design for water reuse explains why the reuse duty, biology, hydraulics and residuals must be designed as one system.

MBR membrane integrity adds several operating complications:

• mixed liquor creates a severe consequence if a fiber, seal or header opens directly to permeate;
• common permeate headers can dilute the signal from one damaged cassette;
• air-scour, relaxation, backpulse and chemically enhanced backwash repeatedly change pressure states;
• cleaning or maintenance can disturb seals even when the membrane surface is restored;
• permeate turbidity may signal a gross solids path but cannot prove removal of every pathogen or dissolved constituent; and
• downstream disinfection or advanced treatment remains a separate barrier whose capability must be verified.

The MBR program should define cassette- or train-level isolation, off-spec permeate diversion, sample locations, response to a turbidity step change and the test method used after maintenance. If an integrity test cannot be performed at full operating solids conditions, the transition into the test state and the relevance of that state to real filtration should be documented.

Industrial UF still needs a defined barrier claim

Industrial UF used ahead of RO may not receive pathogen-removal credit, but integrity still has commercial value. A breach can send suspended solids, colloids or microorganisms to cartridge filters and RO feed spacers, shorten cleaning intervals and destabilize production. In food, electronics or pharmaceutical utilities, the consequence can include product-quality risk rather than only membrane maintenance.

The owner should define the controlled hazard, acceptable particle or microbial passage, monitoring location and response. The industrial membrane filtration selection framework can help distinguish the separation duty of MF, UF, NF and RO. Integrity proof should then be written for the chosen barrier rather than inherited from a different technology.

For industrial service, membrane filtration compliance may mean a customer specification, validated utility standard, discharge permit or internal quality system rather than a drinking-water regulation. The evidence architecture remains useful, but numbers such as test frequency, breach resolution and turbidity triggers must be derived from the applicable duty. Copying a U.S. drinking-water limit into a factory SOP without its regulatory context creates the appearance of rigor without a defensible basis.

Make the data package capable of surviving an investigation

Integrity data often becomes important only after an excursion, when teams need to reconstruct what happened. A pass/fail historian tag alone is inadequate. The plant should retain raw test curves, calculated values, test-quality flags, valve feedback, instrument status, relevant operating conditions and the identity of the unit tested.

Synchronize clocks and asset names

A filtrate turbidity spike at 10:04 cannot be correlated with a backwash at 10:03 if the analyzer, PLC and historian clocks differ. Asset naming must also be consistent: “Train 2” on the HMI must correspond to the same rack, sample line and work-order location in every system.

Version the calculations

Record the formula, constants, system volume, sensor range, correction factors and control-limit revision. If a module population changes, confirm whether tested volume and design flow must be updated. A calculation that was valid for six modules may not retain the same sensitivity after an expansion to eight.

Record overrides and water disposition

Every alarm bypass, manual valve movement and delayed test should be time-limited, authorized and visible. When suspect filtrate is diverted, the record should show where it went, how much was produced and how the affected tank or product was released or disposed. Integrity management is incomplete if the mechanical fault is repaired but the potentially affected water has no disposition record.

Write procurement specifications around evidence and response

A strong specification does not ask only for “an automatic integrity test.” It describes the proof that must be delivered and how the plant will respond when the proof fails.

Qualification deliverables

• product-specific challenge report and independent review status;
• tested module identity, lot-selection method and worst-case rationale;
• target or surrogate, analytical method, reporting limits and LRV calculation;
• operating envelope and change-control triggers; and
• relationship between challenge result, requested credit and direct-test sensitivity.

Installed test deliverables

• test boundary drawing and isolatable membrane-unit definition;
• method resolution, sensitivity calculation, baseline and control limit;
• pressure, temperature, level, flow and valve instrumentation requirements;
• raw-data storage, calculation transparency and historian export;
• automatic failure response and permissive logic;
• diagnostic tools, module-localization method and repair kit; and
• site acceptance test using the final hardware and software.

Operational acceptance deliverables

• operator and maintenance procedures for normal, suspect and failed states;
• instrument calibration and functional-test intervals;
• post-maintenance and post-cleaning test requirements;
• repair limits, module replacement rules and residual-capacity tracking;
• indirect-monitoring sample design and alarm validation; and
• training based on real failure scenarios rather than only routine screens.

The commercial evaluation should compare outage duration, localization capability, false-alarm risk, data transparency and service response—not only membrane price. A cheaper rack that cannot isolate a fault may impose a higher cost per reliable operating day.

Four incident files show why one alarm has several meanings

Incident A: failure immediately after backwash

A unit fails its first physical test after a reverse-flow backwash but passes after a controlled wetting and stabilization sequence. The investigation shows that the original test began before the membrane and test piping returned to the validated state. Corrective action is not to widen the failure limit. It is to correct the sequence, prove wetting and retain a quality flag when test preconditions are not met.

Incident B: a broken fiber diluted by a common header

Finished-water turbidity remains stable, but one train’s particle count rises and its next physical test fails. The common header had diluted the release below the plant-wide analyzer’s ability to localize it. The event supports train-specific monitoring and automatic isolation, not a conclusion that turbidity is useless.

Incident C: seal leak after module replacement

New modules pass supplier documentation review, yet the rack fails on recommissioning. Segment testing localizes the response to one housing; inspection finds a rolled O-ring. Replacing more membranes would have added cost without closing the bypass path. The root cause belongs to assembly control and post-maintenance proof.

Incident D: pressure drift mistaken for damage

Several racks begin failing within the same shift. Raw curves show a similar offset unrelated to module age or position. A reference pressure check identifies sensor drift after maintenance on the common test-air system. The plant keeps affected water isolated until test credibility is restored, recalibrates the instruments, reruns the tests and documents why the original results were invalid.

These cases demonstrate the value of evidence hierarchy. The program must be sensitive enough to find a true breach and disciplined enough not to “repair” an instrument problem by disabling healthy membranes.

Focused FAQ

What is the difference between a challenge test and an integrity test?

A challenge test demonstrates the removal capability of an intact membrane product or representative device using a target organism or surrogate. An integrity test evaluates whether an installed, isolatable unit currently has a physical breach. Challenge evidence supports the capability claim; routine direct testing supports the condition of the installed barrier.

Is low filtrate turbidity proof that an MF or UF membrane is intact?

No. Low turbidity is useful indirect evidence, especially for gross failures, but a small defect can be diluted, the source water may contain few particles, and analyzer limitations can mask change. Unit-specific turbidity or particle monitoring should be paired with an approved physical test.

How often should a direct test be performed?

Frequency depends on the applicable regulation, permit and risk basis. Under the U.S. federal membrane-filtration provisions in 40 CFR 141.719, each operating membrane unit is generally tested at least daily, subject to state-approved alternatives described in the rule. Reuse and industrial systems should establish a justified frequency rather than copying that value without context.

Why can a healthy membrane unit fail a pressure-based test?

Incomplete wetting, unstable temperature, trapped gas, a leaking isolation valve, wrong system-volume data, inadequate stabilization or pressure-sensor drift can create an apparent failure. Test-quality checks should be reviewed before physical repairs begin, while affected water remains controlled.

Can a fouled membrane pass an integrity test?

Yes. Fouling raises hydraulic resistance and may reduce permeability without opening a particle pathway. Integrity and cleanability are separate properties. A plant should trend permeability, TMP and pressure drop for fouling while using the validated direct method for barrier condition.

Does a passing test prove every seal and valve is sound?

Only if those components are inside the tested boundary and the test can respond to their leakage path. Review the valve lineup and boundary drawing. Components bypassed during the test require another functional check or a redesigned test configuration.

What should happen after a failed direct test?

The affected unit should be removed from service according to the approved response, suspect filtrate controlled, the test validated, the fault localized and the repair documented. The unit should return only after it passes the approved direct test and any additional quality-release requirements are satisfied.

Should a plant replace an entire train after one failure?

Not automatically. Use localization evidence to distinguish a single fiber, module, seal, header, valve or test-system problem. Replacement scope should follow the defect and root cause. Fleet replacement is justified by broader evidence such as systemic aging, widespread damage or an obsolete product—not by one unresolved alarm.

How should integrity be specified for an MBR?

Define the reuse-quality duty, isolatable cassette or train, direct-test method, permeate monitoring point, diversion route, maintenance triggers and downstream barriers. Account for common-header dilution, mixed-liquor consequence, air-scour and cleaning states. Do not rely on low turbidity alone.

Can the same procedure be used for RO?

Not by default. RO and NF use different membrane structures, operating pressures and failure diagnostics. Conductivity profiling, salt passage, vacuum testing and mechanical inspection may be relevant, but the method must be validated for the specific technology. Low-pressure hollow-fiber criteria should not be transferred without engineering justification.

The management principle: no barrier credit without living proof

A membrane is valuable because it creates a physical separation, but that value cannot be managed by appearance or automation status. The defensible program starts with a qualified removal claim, draws the complete installed boundary, selects a physical test with adequate resolution and sensitivity, watches unit-specific filtrate quality between tests and controls every transition after an alarm.

The best integrity system is not the one that produces the most pass results. It is the one that makes uncertainty visible, isolates suspect water quickly, preserves diagnostic evidence, limits repair scope and requires objective proof before release. When challenge data, direct testing, continuous monitoring, maintenance records and water disposition agree, the plant has more than a membrane skid. It has an auditable barrier.

#MembraneIntegrity #MembraneTesting #Ultrafiltration #Microfiltration #MBR #WaterTreatment #BarrierVerification #WaterReuse #DrinkingWater #MembraneMonitoring