When a Thai factory reports falling reverse osmosis (RO) permeate production or rising conductivity, an immediate chemical clean is an expensive guess. An industrial RO membrane performance monitoring system should record feed, concentrate and permeate pressure, flow, conductivity and temperature at the same time; compare performance under equivalent conditions; and show the operator what to check next. This guide translates that requirement into measurement points, normalization, cleaning decisions, a request for proposal (RFP), and factory and site acceptance tests (FAT/SAT).
Define the decision before adding instruments
A dashboard full of live numbers is not, by itself, membrane performance monitoring. A drop in permeate flow could reflect colder feedwater, a change in salinity or recovery, a different pump setpoint, a failing instrument, or fouling. A rise in permeate conductivity could reflect feedwater changes, increasing salt passage, an O-ring leak, bypass mixing, or measurement error. Manufacturers therefore recommend reliable operational records and comparison of normalized performance with a stable initial baseline, rather than a judgment from one live reading. DuPont’s technical manual and Toray’s operating manual explain the approach.
The useful outcomes are fourfold: detect a change with a shared definition; avoid unnecessary cleaning without delaying a necessary one; measure performance before and after cleaning or membrane replacement; and give the factory, skid supplier and membrane maker a common evidence set. The software should support a documented human decision process. It should not silently choose a chemical and cleaning procedure from an alarm alone.
Survey the actual Thai factory installation
For each skid, list the number of trains and stages, pressure vessels, membrane make and model, design flow and recovery, source water, pretreatment, cleaning skid, existing PLC/SCADA, and laboratory tests. Record source changes, demand-driven production changes, filter replacement, dosing adjustments, membrane replacements and cleaning as timestamped events. A factory alternating estate water and well water cannot interpret a single unlabelled trend as membrane deterioration.
Include cartridge filter differential pressure, relevant pretreatment status, and oxidant checks upstream of the RO. If pretreatment has failed, cleaning the membrane without correcting the upstream cause merely starts the fouling cycle again. For the wider water balance and meter locations, see our water usage monitoring guide for Thai factories.
RO differential pressure: two pressure concepts that must stay separate
Stage feed-to-concentrate pressure drop is the pressure at that stage’s feed inlet minus its concentrate outlet pressure. It indicates resistance along the feed channel and can rise as that path fouls or plugs. The normalized pressure drop in DuPont’s FilmTec cleaning guidance refers to this feed-to-concentrate difference after accounting for operating conditions.
The driving force that moves water through the membrane is a different concept. It involves the pressure on the feed/concentrate side relative to the permeate side and the osmotic pressure difference. Calling feed minus concentrate pressure “transmembrane pressure” and using it directly to explain permeate production confuses flow-channel loss with membrane transport. Use distinct tag names and definitions in the P&ID, calculation sheet and HMI. Do not treat pump-discharge pressure as stage-inlet pressure without checking pipe losses and measurement location.

For a one-stage train, measure feed-inlet, concentrate-outlet and permeate-header pressures. For a multistage train, independently measure each stage inlet and outlet if stage-level diagnosis is required. Overall pressure readings alone may not distinguish a first-stage deposit from later-stage scaling. Toray’s manual calls for checking individual stage differential pressures and the vessel pressure-drop limits in the applicable product sheets. Mark startup, shutdown, flushing, cleaning, pump speed and control mode in the historian so transient readings do not contaminate the steady-state baseline.
Permeate conductivity is a quality signal, not a complete diagnosis
Measure feed and permeate conductivity using documented sample points and temperature compensation; add concentrate conductivity where the process and diagnostic plan require it. Trend salt passage or rejection with the changing feed concentration, rather than treating a permeate conductivity alarm as proof of membrane damage. A fixed conductivity-to-TDS conversion factor can be misleading if the ion mix changes. Compare it with laboratory analysis and document any chosen factor.
The US EPA Membrane Filtration Guidance Manual discusses conductivity as an indirect NF/RO monitoring method and conductivity profiling as a way to locate leaks along pressure vessels. It also cautions that recovery, flow, pH and temperature can change the interpretation. That US guidance is a technical reference here; it is not a statement of Thai legal requirements.
For a sudden permeate conductivity increase, first check analyzer calibration, sample flow, air bubbles, temperature compensation, feed conductivity, recovery and bypass valves. If needed, take stage or vessel samples and discuss profiling with the membrane supplier. For a gradual trend, compare normalized salt passage and the operating history. Neither pattern justifies automatic membrane replacement.
Use normalized permeate flow to compare like with like
Normalized permeate flow converts the present operating point to an agreed reference condition, including relevant effects of feed temperature, salinity and pressure. Cooler water may produce less permeate without fouling. Conversely, a constant-flow plant may hide deterioration by increasing feed pressure. TorayTrak is Toray’s tool for comparing normalized RO performance with startup conditions; DuPont describes FTNORM for FilmTec systems. Select the method for the membrane and configuration actually installed.
Choose the baseline after commissioning has stabilized, instruments are calibrated, the membrane layout is confirmed, and representative water quality has been recorded. Do not use the first startup reading automatically. Keep the old baseline when a membrane, stage arrangement, instrument or source-water regime changes; record the change date and approve a new baseline. Otherwise a later operator cannot reconstruct whether an apparent improvement came from maintenance, new elements or a revised reference point.

If normalization is embedded in a dashboard, document its calculation method, input units, temperature and salinity assumptions, treatment of missing data, and software version. Feed identical test records into the application and the relevant manufacturer tool; agree a tolerance and compare outputs during FAT. Distinguish a rough trend indicator from a formal normalized figure. Show raw and normalized permeate flow, stage feed-to-concentrate pressure drop, feed and permeate conductivity, salt-passage trend, temperature, recovery and pump state on the same time axis. Flag startup, cleaning, instrument service and invalid records instead of replacing them with zero.
Data quality comes first. Specify whether pressure is gauge or absolute; whether a flow value is instantaneous or derived from a totalizer; and whether conductivity is temperature compensated. Synchronize timestamps at the PLC and historian. Check the feed flow against permeate plus concentrate flow, while accounting for flush, recycle, blowdown and actual meter locations. Set an acceptable balance error from instrument uncertainty and piping design, not from a universal rule. Toray recommends daily reliable records and regular calibration so that the historical trend can support troubleshooting.
Make CIP a reviewed decision, not a single threshold
DuPont’s FilmTec cleaning procedure lists cleaning considerations including a 10% fall in normalized permeate flow, a 5–10% rise in normalized salt passage, or a 10–15% rise in normalized feed-to-concentrate pressure drop. These are FilmTec guidance values, not universal RO standards or Thai regulatory limits. The same document advises checking pretreatment, pressure control and recovery before concluding that the elements need cleaning. Hydranautics TSB107.30 distinguishes typical and high-fouling conditions, reinforcing the need to use the current instructions and warranty terms for the installed product.
A useful workflow has three gates. Trend review: verify that the change exceeds normal variation and check the instruments. Cause diagnosis: compare flow, stage pressure drop, salt passage, source-water analysis, dosing, filters and valve history. CIP planning: confirm the membrane maker’s compatible chemicals and procedures, cleaning conditions, wastewater handling, lockout and required downtime. An alert opens the investigation; it does not automatically choose a cleaning chemical.

If permeate production falls while feed temperature falls, recalculate the normalized flow. If first-stage pressure drop rises together with a pretreatment filter problem, investigate pretreatment and foulant type. If later stages change, review recovery and scaling potential. If conductivity jumps while flow and stage pressure drop stay nearly steady, inspect the sampling system, seals and bypasses. These are diagnostic paths, not definitive rules. Use samples and manufacturer support when the evidence is ambiguous.
Keep a CIP record with the reason, stable pre- and post-cleaning data, chemicals and concentrations, pH, temperature, circulation flow, pressure, contact time, waste handling and approver. Compare normalized results after the plant has stabilized. Poor recovery should trigger a new diagnosis before repeating the same treatment. For adjacent maintenance processes, see our guides to filter differential pressure monitoring and CIP conductivity monitoring. Use the actual RO pretreatment filter’s specifications for its replacement limit.
System architecture and operating records
A practical arrangement has five layers: field instruments and analyzers; PLC or acquisition equipment; a timestamped historian; monitoring and calculation screens; and reports and notifications. Keep pump protection and safety shutdown behavior in the local control system. If the network fails, a local buffer should retain timestamped data and upload it without duplicates when service returns. The upper layer can support review and reporting without becoming a single point of failure for safe operation.
Place instruments where they can be calibrated and serviced. A conductivity analyzer needs a suitable sample flow. A pressure instrument needs consideration of vibration and pressure transients. A flowmeter needs the full-pipe and straight-run conditions required by its manufacturer. Walk the pipework with the P&ID before approving the design. Store raw measurements, quality flags, normalized outputs, calculation version, alarms, acknowledgments, approvals and maintenance events. Separate permission to view trends from permission to change thresholds or edit events; audit every change.
Alarm logic for a real two-stage skid
Suppose a two-stage skid shows lower permeate flow and higher feed pressure after a shift change. The first display should show temperature, feed conductivity, recovery, pump speed and valve position before and after that event, alongside stage pressure drop and normalized flow. If temperature changed while normalized flow stayed stable, the likely action is to document the operating change. If first-stage drop and normalized flow both worsened, examine pretreatment and feed quality. If the second stage changed most, investigate recovery and concentrate-side conditions. A screen that merely says “CIP required” obscures this reasoning.
Give protection alarms, water-quality excursions, instrument failures and gradual performance alerts different priorities and response times. Keep equipment protection within the approved local control design. A failed sensor should produce “decision unavailable,” not a false membrane-cleaning diagnosis. Specify alarm persistence, reset conditions, notification routing and holiday coverage, then test each rule. Daily reports should show raw and normalized flow, stage drops, conductivity, recovery and events. Monthly review should classify what each alert turned out to be: instrument, pretreatment, operation, membrane condition or unresolved. That record improves both alarm settings and capital decisions.
What to request in the RFP
Start with the skid inventory, current P&ID and PLC, installed membranes, present operating problem, and the decisions the system must support. Ask suppliers to separate a survey-stage estimate from a price confirmed after checking pipework, panel capacity, communication and water quality. Compare delivered data quality and operating documents, not simply sensor count.
Measurement specification: state location, units, range, accuracy, compensation, calibration method and maintenance access for each feed, concentrate and permeate pressure, flow, conductivity and temperature point. List the stage pressure points needed for stage-level diagnosis. If concentrate flow is inferred, require the calculation and expected uncertainty. Define which pretreatment and source-water data are included.
Data specification: state sampling and storage intervals, clock synchronization, offline buffer, missing-value flags, raw-data export, retention, backup and network boundaries. Require a tag list with data type, unit, polling interval and ownership for integration with existing SCADA. Choose the sampling interval from the diagnostic purpose; do not retain only a coarse daily average if simultaneous values are needed for normalization.
Decision specification: identify the membrane make and model, baseline approval conditions, normalization method and version, candidate cleaning alerts, operator checks and approval rights for threshold changes. Require the supplier to cite the specific manufacturer source and applicability of each preset threshold. Explainability is a better acceptance criterion than an unsupported promise of automatic AI diagnosis.
Deliverables and support: request the updated P&ID, instrument register, PLC tags, calculation specification, screens, backup and restore procedure, calibration records, operator instructions in the languages used on site, maintenance contacts and a regression check after replacing an instrument. Separate instrumentation, panel work, wiring, software, survey, commissioning, training and maintenance in the price. Check downtime and warranty implications with the equipment supplier.
FAT and SAT: accept an actionable system
At FAT, use simulated input records to test tags, units, screens, alarms, calculations, missing data and roles. Include a temperature-only change, a conductivity spike, a first-stage pressure-drop rise, an invalid sensor, and a communication interruption. Compare the coded pressure-drop and normalization logic against its specification. If the baseline is not approved, the display should not label a trend as a confirmed CIP decision.
At SAT, check instrument location against the physical P&ID, reading direction and units, comparison with field gauges, clock synchronization, steady-state flow balance, offline buffering, alarm delivery, report exports and data restoration. Dangerous pressure or poor water quality need not be created to test the system; use simulated signals and safe operating ranges. Compare several stable days of actual data with an approved calculation example or manufacturer tool and record tolerance, exceptions, calibration status and signoff. If no real CIP occurs before handover, test a before/after report with approved historical or simulated data and schedule a follow-up at the next actual clean.
Test the awkward operating states as well: power recovery, source-tank switching, weekend shutdown, low-demand running and analyzer calibration. Preserve raw values while flagging records excluded from the baseline. If a network link is down, the server must not calculate a fresh normalized result from stale values. Distinguish instrument time, PLC time and server receipt time. On recovery, import buffered values in order without duplicate alarms. If source-water chemistry changes beyond the calculation’s stated assumptions, show that the result requires water analysis and supplier review.
A realistic 90-day example begins with a site and data survey, then design and FAT cases, installation during an agreed outage, SAT and baseline approval using stable operating data. Procurement lead time, production windows and quality approval may change this schedule. In the first month of live operation, tune nuisance alerts and review each finding with operations. The project succeeds when the team can explain the evidence behind a maintenance decision and later reconstruct it.
FAQ
Where should RO differential pressure be measured?
Measure the inlet and concentrate outlet of the same stage to obtain its feed-channel pressure drop; do this for each stage if stage diagnosis is required. Measure permeate pressure separately when evaluating membrane driving force or backpressure. Confirm allowable limits in the installed membrane and vessel specifications.
Does a rise in permeate conductivity mean membrane replacement?
No. Check calibration, sampling, feed conductivity, temperature, recovery and bypasses first. Use stage or vessel sampling, potentially conductivity profiling, if a local leak is suspected. Only then assess cleaning or replacement with the relevant manufacturer.
Can normalized permeate flow be calculated in a spreadsheet?
Yes. Toray provides TorayTrak, for example. Choose the version for the installed train, control units and baseline inputs, and verify any dashboard implementation against the approved calculation in FAT/SAT.
Is a 10% flow reduction a universal CIP trigger?
No. The DuPont value refers to normalized permeate flow in FilmTec cleaning guidance, not a 10% reduction in uncorrected flow and not a rule for every membrane. Apply the current membrane-specific guidance, warranty and site diagnosis.
Can an existing PLC and SCADA be reused?
Often, if available I/O, communication, timestamping, history, access control and backup meet the requirements. Survey the tags and P&ID first. Our PLC and SCADA guide for water treatment discusses the local control boundary.
Summary
Effective RO performance monitoring begins with properly located pressure, flow, conductivity and temperature measurements on the feed, concentrate and permeate paths. Keep feed-to-concentrate stage pressure drop separate from membrane driving pressure. Compare normalized values against an approved baseline, and use membrane-specific instructions plus checks of pretreatment, control and instruments before planning CIP. Put calculation traceability in the RFP and diagnostic scenarios in FAT/SAT so the delivered system supports decisions rather than merely displaying numbers.
If your Thai factory is still deciding which existing instruments to reuse or how to specify stage monitoring, normalization and CIP review, contact TOMAS TECH. A P&ID, membrane model and sample operating log are enough to begin identifying the measurement and acceptance questions.
Technical sources
- DuPont, FilmTec RO/NF Technical Manual, Rev. 19, February 2026.
- DuPont, FilmTec Cleaning Procedures Manual, Rev. 13, February 2026.
- Toray, RO Operation, Maintenance and Handling Manual, April 2026.
- Toray, TORAY TRAK Normalization Software.
- Hydranautics, TSB107.30 Foulants and Cleaning Procedures, April 2026.
- US EPA, Membrane Filtration Guidance Manual, 2005. Technical background, not Thai regulation.