The collector fan may be running while dust escapes at a distant machine hood. A duct can begin to accumulate material even though filter differential pressure appears normal. These are different failure modes, and a fan run signal does not tell the maintenance team what is happening in every branch. A duct airflow monitoring system gives the team a continuous view of selected points and helps it spot changes from a properly commissioned baseline. One velocity reading cannot certify every branch, hood capture or worker exposure.
This guide focuses on airflow between the hoods and the main duct, rather than the collector filter. For the latter, see our dust collector differential pressure monitoring guide. The discipline of defining measurement points, data ownership and maintenance responsibilities also applies to our steam flow meter selection guide, although a steam metering method cannot simply be transferred to dusty air. The steps below cover a Thailand factory retrofit from the request for proposal through FAT, SAT and ongoing operation.
Define the decision before buying a sensor
Start with the abnormal condition the plant needs to detect, who will act, and how quickly. A remote branch that frequently accumulates dust may deserve a dedicated observation point. A main duct point may reveal a change in total system flow, especially when combined with fan speed and the set of active machines. If the central issue is dust escaping a hood, the investigation must also examine the hood’s position, enclosure, cross-draughts and the actual capture of material. A duct sensor alone cannot close that question.
HSE’s HSG258 recommends that an LEV specification describe each hood and its position, face velocity and static pressure, duct dimensions, transport velocity and volume flow, branch balancing, air cleaner and fan conditions, and indicators and alarms. That makes airflow monitoring one part of the ventilation specification. Write down whether the intended use is to flag transport-condition drift, branch imbalance, a change after cleaning or filter replacement, or a documented maintenance trend. Rank these purposes if more than one matters.
| Decision | Useful observation | Additional evidence | Main limitation |
|---|---|---|---|
| Is one branch declining? | Straight section of that branch | Hood, damper and machine status | A stable main duct can hide this loss |
| Is the whole system changing? | Appropriate main duct section | Fan speed, active branches, filter differential pressure | A main duct point does not prove every hood works |
| Is material accumulating? | Vulnerable branch and inspection access | Cleaning history and internal inspection | A sensor cannot locate or quantify deposits alone |
| Is exposure controlled? | Source, hood and work position | Capture observation and appropriate exposure assessment | Duct velocity is not an exposure measurement |
Keep velocity, volume flow and capture separate
For average duct velocity v and cross-sectional area A, volume flow Q is approximately v × A. The word average matters. A single probe in a non-uniform velocity profile does not directly produce the cross-sectional mean. Bends, junctions, dampers and fans create swirl and uneven flow; deposits may change the effective area. A fixed sensor should therefore be correlated with reference measurements during commissioning and used to compare like-for-like operating modes, with its limitations recorded.
Transport velocity describes air carrying particles through a duct. Hood capture describes whether the contaminant enters the hood before spreading into the workspace. HSE defines duct velocity as the cross-sectional average and treats hood condition, hood face velocity and static pressure, branch duct velocity, fan flow and filter differential pressure as distinct checks. OSHA’s ventilation investigation guidance describes a multi-point duct traverse for a more accurate measurement. These UK and US technical documents provide methods and concepts; they do not establish legal compliance in Thailand.
Donaldson explains that the needed conveying velocity depends on the dust. Too little can allow dropout and accumulation; too much can increase abrasion and energy demand. Do not copy an example value from a supplier article or HSE’s indicative table into a universal alarm setpoint. Dust density, size, moisture and agglomeration, duct dimensions, flow, material, number of open hoods and the fan’s available pressure all matter. Combustible or toxic dust and recirculated air require their own hazard assessment.

Select points for representativeness and access
Overlay the actual plant onto a system drawing: every hood, branch, junction, damper, collector, fan and discharge. Add which machines operate together, complaints, cleaning frequency and previous blockages. Choose points that change a decision. A vulnerable branch, a remote hood and the main duct have different roles. The number of permanent points should follow the abnormalities the plant needs to distinguish, not a target sensor count.
HSE advises measuring a duct serving each hood, where practicable, in a straight section well downstream of bends and turbulence. A retrofit may not offer an ideal run. In that case, do not promise ideal accuracy. Compare the available point with a reference traverse or other accepted method across relevant modes, document reproducibility and state the interpretation limit. Provide safe access for cleaning, calibration and replacement; the inspection opening and platform are part of the project, not an afterthought.
A branch point is sensitive to its own branch, but opening another branch can change its reading. A main duct point sees total change but can miss a decrease in one branch offset by an increase elsewhere. If the proposal contains only one main duct sensor, list the branch faults it cannot identify. Add permanent points to critical branches or use periodic traverses and hood checks where appropriate.
Choose technology for the actual air and maintenance work
Options include a pitot or differential-pressure arrangement, a thermal velocity probe, duct static pressure as a proxy, and a combination of sensors with fan and damper data. A display in m/s does not make these instruments interchangeable. Check the maker’s stated range and accuracy conditions, temperature and humidity limits, tolerance of dust deposition and pulsation, cleaning method, insertion depth, spare parts, electrical output and calibration process. Trial the candidate at a representative point if the dust is likely to foul it.
The TSI TA465 manual illustrates the operating and maintenance conditions that apply to a portable air-velocity instrument. Portable reference instruments and fixed monitors serve different purposes. The former can be used for a traverse or comparison; the latter follows one point over time. When readings disagree, examine probe position, operating mode, corrections, contamination and instrument condition before calling one device defective. Label every tag as local velocity, estimated cross-sectional mean or a static-pressure proxy.
An inserted probe can obstruct the flow or collect material. The installation plan should include the opening, support, air tightness, electrical grounding where relevant, isolation before maintenance and closure when a probe is removed. Combustible dust needs a specific ignition and electrostatic hazard review by competent specialists. This article does not select an explosion-protection method or certify regulatory compliance.
Build baselines by operating mode
HSE describes commissioning as the step that demonstrates adequate control from day one and provides a benchmark for later examinations. Establish the baseline while the process operates under realistic dust-generating conditions. A fan-only test with machines idle cannot substitute for the actual damper combinations and filter loading encountered in production.
For every point, record the ID and location, duct size, instrument and method, date, process and material, active hoods, damper positions, fan speed or drive frequency, filter condition and relevant air conditions. Keep the reference cross-sectional measurement or the designer-approved method alongside the fixed sensor’s local reading, hood checks, static pressure and filter differential pressure. These details explain what a later decline means.
Set alarm levels from the approved design range, instrument uncertainty, normal variation across modes, expected filter loading and response time. Consider a warning for a developing trend and a separate level requiring immediate investigation, with suitable delays and hysteresis to prevent transient nuisance alarms. Excessive delay can also hide a real decline. There is no defensible universal m/s threshold for every factory; the SAT evidence must support the chosen values.

Preserve the meaning of each PLC and IoT tag
A sensor-to-PLC-to-dashboard data path is incomplete unless each value carries its unit, point ID, measurement principle, interpretation, operating mode, limits, alarm rule and last calibration date. Display a failed sensor or lost communication separately from a genuine low-flow reading and from a stopped fan. These states call for different actions.
Put fan running state and drive frequency, active machines, damper positions and filter differential pressure on the same time axis. A single branch decline with a stable main duct suggests a local investigation; a system-wide decline with reduced fan speed suggests a control-side check. Filter differential pressure and flow may together point to filter or cleaning-system investigation. These correlations help triage; a dashboard cannot establish the cause by itself.
Donaldson’s connected-monitoring literature shows a practical example of combining collector data, a dashboard, thresholds and notifications. For the plant’s own installation, define data ownership and retention, network connection, a local indication during communication loss, alert recipients and after-hours response. Specify who decides whether a process should stop; an email alone is not an operating procedure.
Make proposals comparable
Before issuing the RFP, photograph the hood inlet, duct bends and junctions, inspection ports, dampers, collector, fan and existing control panel under consistent equipment IDs. Photographs cannot establish airflow on their own; use them with drawings and operating records to focus the site survey. Ask each supplier to explain how it will confirm measurement positions, perform reference tests, cover actual operating modes during commissioning and hand over an alarm-response procedure. This distinguishes a hardware-only quotation from a proposal that designs the operating process.
Require suppliers to state excluded or limiting conditions. High temperature or humidity, oil mist, sticky dust, water washing, frequent damper switching and insufficient straight run can affect service life and reading stability. A candid field-test plan and maintenance method is more useful than an unsupported accuracy promise. Specify in the contract that final alarm values will be approved from SAT measurements rather than copied unchanged from a design table.
The RFP should state processes and materials, operating combinations, current drawings, past problems, the decisions the system must support, candidate points, installation constraints, PLC interfaces, alarms, history and reporting, acceptance tests and maintenance ownership. If the supplier chooses final positions, require a site survey and an explicit statement of straight-run conditions and measurement limitations.
Compare complete installed costs: duct work, safe access, isolation, cabling, panel changes, PLC tags, networking, screens, reference measurements, commissioning, training, cleaning, calibration, spares and warranty. Compare alternatives such as every branch monitored, critical branches plus main duct, or main duct with periodic manual surveys by the faults each can detect. The cheapest sensor count may miss the very branch the factory wants to protect.
| RFP item | Supplier evidence | Acceptance evidence |
|---|---|---|
| Points | Drawing, position, straight run and rationale | Installed position and access |
| Instrument | Principle, range, air conditions, cleaning and calibration | Model, settings and test record |
| Baseline | Plan by operating mode and calculation method | Reference data and baseline table |
| Alarms | Warning, action and fault states, recipients | Simulated low flow and communication loss |
| Handover | Drawings, tags, training, maintenance and spares | Procedure that site staff can repeat |
FAT and SAT must test more than a number on screen
At FAT, check sensor or simulated input, scaling, units, tag names, display, history, alarm logic and communication-loss indication. A simulated signal verifies the signal chain, not real duct accuracy or hood capture. Match physical labels, drawings and screens, and record probe orientation.
At SAT, reproduce the operating modes and log reference and fixed readings at the same time. Use a documented traverse or another accepted method under suitable straight-run conditions with a calibrated instrument. Document points, method and required corrections. Inspect hood operation, branch conditions, fan flow and filter pressure separately, as HSE’s guidance does. A green dashboard indicator must not override visible dust escape. Test disconnection, power loss, fan stop, damper change, a specified low-flow simulation and the rules for acknowledgement and recovery.

Connect trend review to physical maintenance
Use the same equipment IDs in patrol sheets and digital history
An abnormal dust-extraction system leaves evidence beyond a sensor trace: dust around the hood, leakage at a duct joint, deposits at an inspection opening, a changed damper position, fan vibration or a different sound from the filter-cleaning system. Paper patrol sheets can still work if point ID, branch ID, machine ID and inspection time match the dashboard. A vague name such as “line 2 duct” makes it hard to connect an alarm with a field action.
Link each automatic alarm to the operator’s inspection under one case number. A useful record might read: low-flow alert, dust escaping at the hood, damper found partly closed, position restored, and reference check completed. This sequence shows whether the alarm caught a real problem, whether the response worked and whether it recurs. An alert count alone is not a maintenance outcome.
Reassess thresholds and inspection intervals with evidence
The initial threshold is a hypothesis supported by commissioning data. Once operation is stable, review the normal distribution across seasons, product changes and filter loading with the designer. Do not relax the threshold merely because alerts are frequent. For each event, compare the operating mode, reference reading, visible dust, probe fouling and communication state. Classify the cause before changing a setting, and retain the old and new value, reason, approving person and effective date.
Cleaning intervals need not depend only on the calendar. Consider dust loading, sensor fouling, trend changes and inspection findings. During the early period with limited evidence, do not extend inspections too quickly. A system with no alerts may simply have a failed instrument or notification path. Test the signal path and indication periodically, and record probe cleaning and calibration. HSE notes that alarms can fail without warning and recommends stating test intervals in the user manual.
Judge energy savings and availability separately
Once monitoring exists, a supplier may propose using a variable-speed drive to reduce fan energy. Lower velocity is not automatically a good outcome: adequate hood capture and particle transport remain constraints. Donaldson explains both the risk of deposition at low conveying velocity and the abrasion and energy cost of excessive velocity. Before changing fan control, check the worst-positioned branch and operating combination. Use branch measurements, hood observations and cleaning history along with drive frequency.
An energy comparison needs similar production volume, operating hours, dust generation, filter condition, cleaning work and downtime before and after. A lower electricity bill under a lower production load does not prove an improvement. Conversely, detecting a blockage before an unplanned shutdown has value even if energy consumption barely changes. Choose business indicators tied to the purpose: time to notice a genuine alarm, time to identify its cause, recurrence and completion of safe inspections.
Write an equipment-specific response, not one rule for the whole plant
Neither stopping every process for every low reading nor allowing every process to continue is a sound default. The action depends on the dust hazard, generation rate, visible hood escape, other controls and worker position. Put clear steps on the screen: notify the responsible person, check the hood and damper, and escalate a decision on safely stopping the affected machine. Define a substitute check when the sensor itself fails. For a safety-related automatic stop, assess the sensor’s reliability and the control system’s failure modes; an ordinary IoT alert is not a safety function.
During operation, staff can observe dust escape and unusual noise, but must not open a running duct casually. Maintenance should check fan operation, filter differential pressure, damper position, cleaning equipment and likely blockage in the affected branch. If cleaning restores the reading, investigate why material accumulated: source conditions, duct slope, bends, junctions and flow balance. If only the probe was fouled, change its cleaning interval or mounting position to prevent repeating false alarms.
Evaluate data by operating mode
Day and night shifts may run different machines, while material and season change the dust load. One normal band across every mode can mark an idle branch as faulty or conceal a real decline during partial operation. If PLC run signals are available, define baselines for meaningful operating combinations. Otherwise, connect the data with shift logs. Group combinations with equivalent flow conditions so the thresholds remain manageable.
A daily average can hide a short, serious low-flow period. Retain the duration and frequency of declines, affected process, time to recovery and rate of missing signals. A day without data must be marked “not evaluated,” not “normal.” This classification helps the monthly review decide whether to invest in duct cleaning, instrument maintenance or a different point layout.
Recommission after modifications and plan a cloud exit
Adding a machine, reshaping a hood, moving a branch, replacing a fan or adding variable-speed control can change the balance. An unchanged sensor reading after such work does not prove that a new hood receives its required flow. The change-control sheet should list affected branches, design flow, damper settings, reference measurements, hood checks and whether alarms must change. Test the altered system in a SAT-like process, and preserve the previous baseline with its date and reason for revision.
If a supplier proposes cloud monitoring, confirm that the plant can export history after network failure or contract termination. Specify a usable format such as CSV, time zone, tag names, units and transfer of access rights. Headquarters, local maintenance and an external contractor may all view the same system in a Thailand plant; assign responsibility for display language and alert recipients. The practical priority is that staff recognize an abnormal condition, act on it and can later reconstruct what happened.
Roll out in three verifiable stages
First, survey the current system. Match the drawing to actual hoods and ducts, map processes, materials and operating combinations, and inspect hood escape, deposits, cleaning history, fan and filter records. Perform reference measurements where needed. If the central problem is poor capture at the hood, improve the hood before adding instrumentation; monitoring can then help preserve that improved condition.
Second, pilot at one representative branch with enough straight run and safe inspection access. Compare the candidate sensor with a portable reference instrument in several operating modes. Check dust adhesion, cleaning effects, seasonal or night-shift variation, and repeatability. State which faults the pilot cannot see before adding points; fitting every branch is not itself a measure of success.
Third, connect the validated points to PLC or IoT data and the operating procedure. Maintenance and production leaders should exercise the screen, alerts and response together, retaining FAT/SAT results in the baseline table. At monthly review, examine the share of alerts confirmed in the field, detection-to-repair time, cleaning evidence and unresolved events. These measures reveal value and expose false alerts or missing points.
Frequently asked questions
Is one main duct velocity point enough?
It can show a total system change but may miss a declining branch or a hood whose geometry has changed. List required fault scenarios first, then add critical branch points, periodic traverses and hood checks as needed.
For example, total flow after two branches join can stay nearly unchanged if one damper closes while the other opens. One hood may then lose needed suction. If using a main-duct point, document which machine combinations define normal operation and what field checks are required under other combinations. State the blind spots before deciding whether a critical branch needs its own sensor.
What m/s should trigger the alarm?
Use the system design, material, duct size, operating modes, reference measurements and instrument position. Published indicative velocities are not universal limits. Confirm the threshold in real production operation.
Does filter differential pressure make airflow monitoring redundant?
No. Pressure drop across a filter and airflow through an individual branch describe different parts of the system. Select the combination based on the fault being monitored; see the filter differential pressure guide.
Can a sensor be put on an existing duct and used immediately?
Installation alone is insufficient. Check whether the position is representative, how a reference reading will be taken, mode-specific baselines, cleaning access and alarm response. If the available straight run is short, limit the interpretation or redesign the point. Before work, confirm safe access around the port, a shutdown and isolation window, and whether the probe interferes with cleaning tools or material flow. A displayed number is not an accepted monitoring system until SAT captures reference measurements under production conditions and responsible staff can demonstrate the abnormal-response procedure.
Discuss your Thailand plant with TOMAS TECH
Share your duct diagram, sizes, active machine and hood combinations, dust type, cleaning and blockage history, existing PLC and collector details. TOMAS TECH can help define the abnormal conditions worth detecting, select practical points, structure the data and alarm screens, and write FAT/SAT evidence requirements. The aim is a monitoring design that drives inspection and improvement, not a velocity number marketed as a safety guarantee. Contact us to discuss the project.
Primary references
- HSE HSG258, Controlling airborne contaminants at work.
- HSE, Commission your local exhaust ventilation system.
- OSHA Technical Manual, Ventilation Investigation.
- Donaldson, Understanding Conveying Velocities.
- Donaldson, Connected Solutions for Dust Collectors.
- TSI, TA465 Operation and Service Manual.
*This article provides general technical guidance. Qualified personnel must evaluate the specific dust, worker exposure, fire and explosion hazards, and applicable local requirements for each installation.*