Installing a nitrogen generator oxygen monitoring system is more than adding an analyser. Where will the sample be taken? Where will off-specification gas go during start-up? What happens when the analyser fails? How will production and quality teams use the record? This guide develops a specification for a Thai manufacturing site, from residual oxygen measurement and PLC changeover to acceptance testing and maintenance. It separates oxygen remaining inside the nitrogen product line from oxygen in the air people breathe. These are different hazards, instruments and decisions.
Two oxygen concentrations, two decisions
Product-side oxygen is the residual oxygen measured in a nitrogen pipe at the generator outlet, receiver or point of use. It helps determine whether gas can be supplied to packaging, purging, oxidation control or another process. An oxygen analyser measures oxygen, not every component of the gas. A display that estimates “nitrogen purity” as 100% minus oxygen must state its assumptions: argon, water vapour, carbon dioxide and other impurities do not disappear from the mass balance.
Ambient oxygen is measured in the workplace air, including rooms, pits and areas influenced by vents. Its purpose is to address oxygen displacement caused by nitrogen leakage or discharge. An OSHA interpretation describes nitrogen as a simple asphyxiant and discusses a potentially hazardous atmosphere below 19.5% oxygen in its US context. That number is not a declaration of the legal alarm setting for a Thai factory. Local requirements, the site’s risk assessment, ventilation and instrument specifications determine the response. A product-gas analyser cannot replace an ambient oxygen deficiency system. Product residual oxygen rising is commonly a quality problem; workplace oxygen falling is a personnel hazard.
| Measurement | Location | Decision | Typical owners |
|---|---|---|---|
| Residual oxygen in nitrogen | Product-gas sampling point | Supply, divert or hold gas | Production, quality, engineering |
| Oxygen in workplace air | Room, pit or vent-influenced area | Alarm, evacuate, ventilate and control access | EHS, engineering, operators |
Why one outlet measurement may not be enough
The gap between a catalogue purity value and gas arriving at the process is where failures occur. Start-up can include a stabilisation period. Feed-air temperature, pressure, flow, dew point and contamination can alter separation performance. A Parker membrane-generator manual states that its published flow and purity data are based on 20°C inlet air; it also notes that pressure and flow disturbances affect outlet purity. These are conditions for the named equipment, not universal guarantees for every nitrogen generator.
A receiver and long distribution pipe create a delay between the generator and point of use. Air left in a pipe after service, backflow, leaks or switching between lines may affect the point of use even if the generator outlet appears normal. Conversely, an analyser only at the point of use can detect an upset too late. Choose sample points by asking which failure must be detected and how much response time the process allows.
The system’s response includes sample-pipe transport, filters, analyser response, PLC scan and valve motion. Do not use only the sensor’s catalogue response time in the protective calculation. Measure the end-to-end time from an off-specification change at the sample point to confirmed valve changeover during commissioning.
Basic system architecture
The usual elements are compressed-air pretreatment, nitrogen generator, oxygen analyser, changeover valves, receiver, point of use, PLC and historian. Logging feed-air pressure, temperature and dew point alongside product oxygen helps identify whether the generator or its utilities caused an upset. Parker’s food-gas white paper treats air pretreatment, gas quality and food-safety assessment together. A stand-alone oxygen meter provides less diagnostic value when filter blockage or condensate cannot be traced.

The P&ID should show the gas sample point, analyser exhaust, calibration connection, isolation and bypass valves, drains, and destination for off-specification gas. An outlet analyser shows generator performance; a downstream analyser shows what is delivered to the process. Whether both are needed depends on product risk, receiver volume, pipework and allowed response time. Avoid dead legs, excessive sample-line length and liquid traps that prevent a representative reading.
Specify measurement principle, range, accuracy, response, sample flow and pressure, temperature compensation, zero/span checks, consumable replacement, output signals and fault diagnostics. Parker’s HFX documentation distinguishes models with integrated oxygen analysers from those without. Therefore, verify the exact model and installed pipework rather than assuming that all generators from a manufacturer have the same monitoring features.
Define the required quality at the process
Start with the user’s process, not the generator brochure. Packaging depends on the food and pack design; purging depends on vessel volume and allowed oxidation; furnace atmosphere depends on the product and temperature. “99.9% nitrogen” alone leaves the permitted residual oxygen, moisture, oil, particles, sampling condition and excursion duration undefined. Quality and production should agree what is measured, what constitutes acceptance and what to do with gas that fails.
Food applications require particular care. EIGA Doc 194/24 addresses on-site nitrogen generation for food use, including processing-aid and food-additive contexts, PSA and membrane hazards, HACCP and construction materials. EIGA Doc 125/20 says on-site equipment should produce gas suitable for its intended use and that on-site analytical equipment may be needed when gas quality must be controlled and tracked. These European documents do not themselves certify compliance with Thai law or a customer’s specification. An audit file should identify the applicable requirement, its version, components to be controlled, calibration record and procedure for affected product lots.
Parker’s HFX document describes a 99% to 99.5% product-purity range for certain membrane models. It is a model-specific statement under specified conditions. It is not a general purity target and does not prove that a particular food, electronics or chemical process is adequately protected. Separate the generator’s demonstrated capability from the process-specific alarm threshold.
Keep oxygen and purity labels unambiguous
An HMI that alternates among oxygen concentration, nitrogen purity and oxygen content invites an operator to apply the wrong alarm direction. Oxygen usually rises as product gas deteriorates; an estimated nitrogen purity falls. Display the units, such as vol% or ppm, and document any conversion and its assumptions. Verify PLC data type, scaling, limits, rounding and missing-data display, especially when a ppm signal is compared with a percentage limit.
If purity is calculated from oxygen, label it as an oxygen-based estimate. Oxygen is only one possible impurity. Moisture, oil and particles may need separate measurement or testing for the target process. Consider analyser uncertainty and the calibration range when selecting a limit; a control threshold set at the instrument’s detection limit may not be defensible.
Sampling position and sample handling
Sampling at the generator outlet can reveal an upset early, but cannot by itself prove the quality beyond a receiver or distribution header. Sampling at a point of use is closer to the product yet adds transport delay. If both points are monitored, identify one as generator control and the other as process-supply verification. Do not automatically apply the same control threshold to signals with different purposes.
Sample exhaust is a safety design item. Do not release a nitrogen-rich sample into a poorly ventilated instrument enclosure or room. The reading on the gas analyser says nothing about the air around the cabinet. Include exhaust routing, calibration-gas discharge and technician exposure in the site risk assessment. Select tubing and fittings for gas quality, pressure and temperature. Protect the sensor against condensate, oil and particles in accordance with its manufacturer’s instructions.
Calculate a preliminary sample transit time from line volume and actual flow, then measure the installed system with a step change. Diffusion, stagnant sections and filter volume can make a theoretical estimate optimistic. Record the time from a simulated off-specification event at the sample point to confirmed safe valve position. That record is also useful when the plant modifies the line later.
Changeover logic for off-specification gas
A sensible control principle is to open the process path only after a valid measurement has remained within specification for the required stabilisation time. Start-up, power restoration, analyser warm-up, calibration, communications failure, sensor fault and low sample flow all require defined handling. A momentary plausible number in any of those states is not proof of compliant gas. The P&ID, control narrative, PLC program and on-site test must agree on the relationship between the process valve and the divert or vent valve.

An immediate trip reduces exposure but may increase nuisance stops; a delay reduces nuisance stops but permits some off-specification gas to pass. Hysteresis can prevent valve chatter, but the recovery condition must not undermine quality. No universal threshold or delay is given here. Define them from allowable off-specification gas quantity, downstream volume, analyser accuracy and system response.
Specify the valve position after loss of power or PLC output. A fail-closed process path is often appropriate, but pipe pressure, backflow and divert capacity still require a site-specific engineering review. Valve-position feedback allows the PLC to compare the command with the real position. A command-only log could incorrectly show a successful changeover when a valve has stuck.
Record useful data in PLC, SCADA and MES
Store measured residual oxygen, units, sample location, analyser status, sample flow, generator state, valve command, valve feedback and alarm timestamps as distinct tags. Add feed-air pressure, temperature and dew point plus nitrogen flow and pressure where they help diagnose an excursion. Agree on time zone and clock synchronisation across PLC, SCADA, MES and quality systems.
For lot decisions, an average alone may hide a short excursion. Retain peaks, time above limit, changeover delay, affected production order and lot. Each MES record should identify when and where a value was sampled and whether the analyser was valid. Mark readings obtained during calibration or a fault so that they cannot silently enter a normal quality average. A CSV containing only numeric values will not reproduce a lot-disposition decision.
Set retention according to customer contracts, product requirements and internal quality rules. Trend-display retention and audit-record retention may differ. Log the operator and reason for alarm-limit edits, calibration-factor changes, manual bypasses and forced valve operation. The value of OT and IoT integration is an explainable record of which gas reached which operation, not simply a larger dashboard.
Implementation sequence: survey to acceptance
First, survey the generator model and technology, rated flow, feed-air quality, receiver and pipe capacity, use points, installed valves, power and communications. Trend oxygen, pressure, flow and start/stop events. Verify the range, calibration, signal outputs and sensor condition of any existing analyser. Check actual equipment against drawings before choosing a sampling point.
Second, write a one-page requirements matrix: residual-oxygen limit by application, relevant flow, permitted excursion, divert destination, acceptable downtime and required audit record. Engineering owns instrumentation and maintenance; quality owns acceptance and lot disposition; production owns shutdown and restart; EHS owns nitrogen exhaust and ambient oxygen risk. Assign who receives each alarm and who authorises restart.
Third, design and install the sample point, sample treatment, valves, exhaust and workplace detectors on a P&ID. Provide access for calibration, replacement, drainage and isolation. EIGA Doc 125/20 addresses installation conditions, maintenance access and protection of feed-air intakes from contamination. Food-use modifications should also preserve hygienic maintenance and foreign-body controls.
Fourth, perform FAT and SAT. Apply in-specification and off-specification test gases, then verify display, alarm, divert, logging and upstream notification as one chain. Test communications loss, sensor disconnection, zero sample flow, power loss and return, calibration mode and manual bypass. FAT cannot prove the installed sampling delay or real exhaust routing; measure those during SAT. Keep the settings and evidence of actual valve position.

Finally, transfer the system into routine operation with instructions for daily checks, calibration, calibration-gas control, consumables, drains, alarm response, isolation, quality decisions and restart. Parker’s HFX manual describes weekly checks and oxygen-analyser calibration for its specific equipment. Follow the selected analyser and generator manuals and the site’s risk-based schedule rather than copying that interval to all models.
Compare the complete cost
Compare quotations with the same scope: analyser, sample treatment, tubing, valves, PLC I/O, HMI, historian, panel, installation, calibration gas, consumables, commissioning and training. Spare PLC capacity does not remove programming, validation or shutdown work. Where audit evidence is required, include record storage, access rights and change logs.
Build the business case from site data: scrap avoided through earlier detection, less manual sampling, shorter investigation and fewer quality complaints. Over-sensitive alarms can also create costly stops. Review the first weeks of alarm history and adjust sampling or hysteresis within the approved risk envelope; do not relax a quality requirement to make a dashboard look quiet.
Considerations for a Thai factory
Measure feed-air temperature, dew point and drainage under actual seasonal and operating conditions. A tropical location does not automatically mean lower nitrogen purity; installation, air conditioning and compressor operation matter. Include panel environment, power variation, restart practice, local calibration-gas availability and maintenance response in the specification.
Use short alarm text in the language operators work in. “High oxygen”, “Analyser fault”, “Low sample flow” and “Valve mismatch” call for different actions. Explain where diverted gas goes, who may authorise a bypass and how the first product after restart is assessed. A Japanese or English design document should be accompanied by a usable Thai operating procedure where that is the site language.
Food use: link HACCP and traceability
For modified-atmosphere packaging or other food contact, analyse the generator, distribution line and packaging operation as one chain. EIGA Docs 194/24 and 125/20 address on-site food-gas generation and supply, but the plant still needs its own process-specific HACCP assessment and applicable local/customer compliance review.
Relate residual-oxygen records to lot numbers using gas transit time and packaging changeover times. An alarm timestamp alone may not mark the first and last affected pack because gas remains in piping and receivers. Log the valve changeover, receiver capacity and process consumption so quality can justify the hold range. Traceability succeeds when the affected product can be identified and explained.
Maintain confidence in the measurement
Sensors drift and consumables age. Track calibration due date, zero/span results, sample flow and pressure alongside the measurement. Parker’s HFX manual lists annual replacement of a galvanic cell for identified models; this is not a universal sensor-life rule. Use the installed instrument’s manual and actual service conditions to plan replacements.
Tell the PLC and historian when the analyser is in calibration mode so its readings cannot be used for release decisions. Control calibration-gas certificates, concentration, expiry and connection procedure. On an unusual reading, distinguish real gas deterioration from blocked sampling or calibration drift while retaining the previously specified safe state. Investigation is not a reason to keep feeding unverified gas to the process.
Common mistakes
Displaying oxygen in SCADA without defining a divert action leaves off-specification gas flowing to the process. Treating generator-outlet oxygen as the delivered value ignores receivers and branch lines. Holding the last good value through a sensor failure hides missing evidence; retain a separate invalid-data flag. Treating the product analyser as ambient oxygen protection confuses quality with personnel safety. Replacing the analyser without logging feed-air conditions prevents proper diagnosis of generator variation.
The upstream utility deserves attention. Related guides on compressed-air dew-point monitoring and compressed-air flow-meter selection can help identify changes before they appear as nitrogen-quality excursions.
FAQ: nitrogen generator oxygen monitoring system
Is one analyser at the generator outlet enough?
It may be enough to monitor generator performance. If a receiver, long header or several use points lie downstream, assess whether point-of-use verification is needed for a product-quality decision. The answer depends on pipework and the time available to detect a fault.
What oxygen limit should trigger a valve changeover?
There is no universal number. Set the limit and delay from process requirements, measurement uncertainty, sampling delay and downstream hold-up. Obtain quality and production approval and prove the response in FAT and SAT.
Can oxygen be converted into nitrogen purity?
It can provide a simplified estimate if other constituents are explicitly ignored, but an oxygen analyser does not directly measure nitrogen. Moisture or other impurities may need their own controls. Label a calculated value as an oxygen-based estimate.
Is an online oxygen analyser sufficient for food-grade nitrogen?
Not necessarily. Review intended use, required components, feed air, materials, hygiene, lot traceability, calibration and deviation handling against applicable requirements. EIGA guidance is a useful source, not a substitute for local law or customer specifications.
Can the same analyser protect workers from oxygen deficiency?
No. Product residual oxygen is measured inside piping. Personnel protection monitors the surrounding air and needs separate placement, alarm and response design.
What determines the installation cost?
Sample-point count, analyser performance, sample treatment, valves, PLC changes, piping, commissioning, maintenance and records all matter. Compare both capital work and operating cost at an equal scope.
Put decision tables and exceptions in the specification
“Monitor oxygen concentration” is not a complete control specification. Map each input to a decision, action and audit record. During normal operation, for example, the analyser must be valid, sample flow adequate, oxygen within the approved limit and equipment and valve states consistent. Only then may the process path open. If a condition fails, define whether the process path closes and where the gas is safely diverted. This exposes differences between an instrumentation contractor’s idea of “normal” and the quality team’s definition of “acceptable” before programming begins.
At start-up, distinguish the generator’s run signal from the point when gas has stabilised within specification. If gas left in the receiver or pipes after shutdown is unknown, define a confirmation or purge step. After power restoration, do not treat the PLC’s retained value or last valve command as proof of current quality. Recheck analyser readiness, sample flow and actual valve position. A maintenance bypass needs an entry condition, authorised person, time limit and exit check. Flag bypass data so that it cannot be counted as normal product-quality evidence.
At minimum, examine high product oxygen, internal analyser failure, communication loss, low sample flow, calibration mode, low feed-air pressure, valve-position mismatch and power loss. For each, decide whether to alarm, stop the process supply or stop equipment. A low oxygen number can be stale when sample flow has stopped. A valve-close command does not prove closure when feedback disagrees. The value, the validity of the value and proof of the control action are three different decisions.
Acceptance-test records should include the applied input, expected valve/alarm/upstream notification, actual action, timestamp and approver. Agree on pass conditions before testing. Test normal, fault and recovery paths so production, quality and maintenance can follow the same timeline when a real deviation occurs. A large collection of tags is of little use if no one can reconstruct these decisions.
Rehearse the quality decision for a deviation
Walk through a hypothetical event on paper: when did each value change, when did each valve move and which lot was in production? If feed-air dew point worsens before residual oxygen rises, align generator-outlet and point-of-use timelines to see the delay between possible cause and effect. Correlation alone does not prove the cause; check drains, filters, pressure and analyser calibration. Link maintenance recovery, quality hold or retest and production restart to one event identifier.
Recovery requires more than clearing an alarm. Confirm that the analyser is valid again, gas has stabilised within specification, actual valve position is correct and affected pipe volume has been displaced. Decide in advance whether the first product after restart requires extra quality checks. If an ambient oxygen-deficiency alarm is involved, personnel evacuation, ventilation and controlled re-entry take priority under a separate safety procedure. Put the rehearsal findings into the specification and training material so the monitoring screen supports real decisions.
Conclusion
A nitrogen generator oxygen monitoring system begins with a process decision: which gas quality may reach which operation, under which verified state. Treat product residual oxygen and workplace oxygen deficiency separately. Specify sampling, total response time, divert logic, instrument fault state and lot record together. Test the installed chain before relying on an alarm to protect production quality.
If you are assessing an analyser retrofit or designing sampling and PLC changeover for a new nitrogen generator, contact TOMAS TECH. We can review the process requirement and available pipework drawings at an early stage.
Primary sources
- Parker, Nitrogen Generator IOM, N2-120/160/240.
- Parker, Balston HFX Nitrogen Generators IOM.
- Parker, NITROSource Food Grade Gas White Paper.
- EIGA, Doc 194/24: On-site Nitrogen Generators for Food Use.
- EIGA, Doc 125/20: Guide to the Supply of Gases for Use in Foods.
- OSHA, interpretation on inert gases and oxygen displacement.