Installing an industrial pipe water hammer monitoring system means more than adding a pressure reading to a dashboard. A short pressure transient after a pump trip or rapid valve closure must be captured together with the operating sequence before anyone can establish a cause. This guide is for liquid piping in Thai factories, including cooling water, process water and wash lines. It covers measurement points, event recording, procurement, factory acceptance tests (FAT), site acceptance tests (SAT) and the operating routine. Condensate induced water hammer in steam piping has a different mechanism and requires a steam system investigation.
Define what the monitoring system must achieve
Water hammer, also called a hydraulic or pressure transient, occurs when liquid velocity changes abruptly and a pressure wave travels through the piping. Pump start and stop, loss of power, valve movement and check valve closure are possible triggers. KSB’s surge pressure reference explains that rapid valve closure can raise pressure upstream while causing low pressure and even liquid column separation downstream. A dashboard showing only maximum pressure therefore cannot describe the whole event.
There are three separate objectives. First, record when and where a transient occurred and which operation preceded it. Second, compare pressure peaks, troughs and repetition before and after an operating or equipment change. Third, collect evidence for hydraulic design. Monitoring does not itself prevent water hammer. Valve closure characteristics, pump run down, check valves, piping and any surge protection must be designed separately. KSB describes several control measures and stresses that a system specific choice is needed.
A loud bang is a useful report, but its location may differ from the origin of the wave. Reflections at branches, tanks, valves and changes in pipe diameter complicate the picture. Assemble the P&ID, pump run records, valve events, check valve types and previous leaks, instrument failures or loose supports. Each waveform needs an equipment tag and a reliable timestamp. The US EPA utility reference also links rapid pressure changes to pump and valve operation and explains the role of transient analysis.
Choose one piping circuit first
A closed cooling water loop, a raw water supply and a chemical transfer line differ in pipe material, allowable pressure, fluid and hazard. An RFP for “factory wide water hammer” blurs the instrument range and safety work. Start with one circuit that has recurring failures or a high consequence of stoppage. Mark its boundaries, normal flow direction, steady pressure range, pump combinations, valve sequence, source and return on the P&ID.
Before modifying the system, review flange leaks, failed gauges, check valve noise, seal damage, moving supports and vibration. None is proof of water hammer in isolation, but together they help select operating modes to test. Keep equipment ID, time, symptom, action and recurrence in maintenance records so that later pressure events can be compared with actual damage. Include recent piping and PLC changes; old drawings may omit a new branch or an altered closure time.
Why an ordinary pressure trend may miss the event
A SCADA value saved every minute is no proof that a brief peak did not occur between readings. Averaging can hide both a high peak and a low trough. A US EPA report on force mains notes that transients lasting less than a second can contribute to damage if repeated. That report is evidence about the phenomenon in force mains, not a prediction that every factory pipe will fail in the same way.
Capture depends on the whole measurement chain: pressure transducer response, logger acquisition, filtering, timestamp accuracy, trigger logic, pre and post event window, memory and communication. A fast sensor attached to a slow logger cannot preserve a fast waveform. A nominally fast logger is equally unhelpful if a long, narrow impulse tube or trapped air changes the signal before it reaches the diaphragm. Include the mechanical pressure tap in the specification. HWM’s pressure transient logger illustrates an architecture with slow background logging and a faster event window; its maximum rate is a product example, not a universal requirement.

Specify the event duration before the sampling rate
Describe the anticipated event duration, the peak and minimum pressure of interest, the rise and fall of the waveform, possible reflections and the required time relationship to pump and valve events. A catalogue’s sampling rate may differ from the rate actually saved or transmitted. HWM’s setup guidance distinguishes ordinary trend logging from triggered fast recording. This permits a long term trend without storing a high rate stream continuously.
If the event duration is unknown, temporarily instrument a representative point and capture normal start, stop and changeover operations. Use that evidence to set the required end to end bandwidth. Neither “one second is enough” nor “100 Hz is mandatory” is a defensible universal rule. FAT and SAT should demonstrate effective performance through the sensor, input electronics, filter and stored record. A missed peak understates the risk; noise mistaken for a peak overstates it.
Keep the trough, waveform and operating context
Each record should contain point ID, absolute timestamp, pressure unit, calibration information, pre and post event waveform, maximum, minimum and pump and valve states. State whether pressure is gauge or absolute and whether the unit is bar or MPa. A high pressure alarm alone misses low pressure behavior. An EPA paper on low pressure transients addresses public water quality risks. Its public health conclusions do not transfer directly to a factory process loop; it illustrates why the low pressure side also deserves engineering review.
A list of peaks cannot easily distinguish a reflected wave, a second valve action or instrument noise. Demand raw waveform export and PLC events on a common timeline. If a power outage is a suspected cause, the logger needs a way to retain data just before the outage. A communication gap must be marked “not observed”, never interpreted as “no transient”. Store the time of measurement separately from the later upload time.
Place sensors around likely sources and reflections
Candidate points include pump discharge, either side of a fast valve, near a check valve, the end of a long branch and a high point. Final locations depend on the P&ID, a hydraulic assessment, access and a safe pressure tapping point. Synchronized recordings from two points may help infer propagation direction. In a branched network, a time difference alone may still be insufficient to prove the source.
An existing gauge nozzle is useful only if it measures what the project needs. Long impulse tubing, dead legs, a closed isolation valve or an air pocket can attenuate a fast event. Check fluid temperature, corrosion, solids, cleaning chemicals and pipe material when selecting wetted components and seals. Prepare isolation, depressurization, leak testing and restoration procedures. Hazardous or high temperature fluids call for an installation safety review before the instrumentation choice.

Integrate timestamps and data quality, not just values
Decide whether the gateway sends event summaries or complete waveforms. One sensible design sends an event ID and severity to SCADA while storing full traces in a historian. Whatever the architecture, ensure a waveform can be retrieved after an incident. Distinguish capture time from transmission time. Test delayed upload and replay after network recovery, including event order and duplicate handling.
Use explicit quality flags for valid, sensor open circuit, out of range, calibration, communication loss and unsynchronised clock. Do not display all failures as a pressure of zero. On the operations side, distinguish pump start command from actual running confirmation, and valve close command from measured valve position. Their time difference may explain the transient. Make OT network access read only where feasible and keep monitoring faults from interrupting pump control.
Separate hydraulic protection from monitoring
One waveform does not automatically mean that a slower valve is the right fix. Reverse flow after stopping, check valve response, column separation and reflections can interact. KSB’s transient flow discussion lists power loss, pump changeovers, valve operations and check valve slam among possible causes. A qualified designer may need pipe length and diameter, material, elevation, flow, pump curves, valve characteristics and tank conditions for transient analysis.
Possible measures include a revised closure profile, a pump sequence, a suitable check valve, an accumulator or surge vessel, and improved support or protection. The choice is circuit specific. KSB’s surge risk overview explicitly warns against a universal solution. Verify that a measure does not create a different low pressure problem or transfer the peak elsewhere. The instrument supplier’s obligation to deliver data and the designer’s obligation to specify protection should be clear in the contract.
For an existing plant, divide work into measurement and control modification. Installing a temporary logger can often precede a production critical control change. Acceptance of the first phase means valid waveforms matched to events. Acceptance of the second requires comparison against allowable pressure and process requirements. Production, maintenance, engineering and safety teams should approve operational changes together.
The equation is a starting point, not a design warranty
The basic expression for a rapid velocity change is Δp = ρaΔv, where ρ is density, a is wave speed and Δv is velocity change; KSB presents this relationship. Real systems add valve closure time, pipe elasticity, entrained gas, branching, reflections and cavitation. Substituting nominal velocity into one equation is not a reliable approval of maximum pressure. Power failure with multiple pumps stopping may need a dedicated transient model. If the model and measurement disagree, review both the model boundary conditions and the instrument chain before discarding either result.
Write an RFP that can be tested
An RFP saying only “water hammer monitoring system, complete” invites incomparable bids. Provide the target P&ID, fluid, temperature, normal and design pressure, pipe material, candidate points, failure history, operating sequence and PLC or SCADA interface. If unknown, procure a site survey as a first stage and freeze sensor positions afterward. Include shutdown windows and the factory’s permit to work process.
| RFP item | What the buyer defines | Evidence required |
|---|---|---|
| Detection objective | Start, stop and valve scenarios; peaks and troughs | Event based measurement design |
| Instrument chain | Fluid, pressure, tap, response, storage rate | Combined sensor and logger specification |
| Time | Synchronisation with PLC and acceptable offset | Simultaneous record test |
| Recording | Trigger, pre and post window, gaps, retention | Sample trace and export |
| Alarms | High, low, communications, sensor fault | Alarm test matrix |
| Installation | Isolation, pressure release, leak test, insulation | Method statement and responsibility |
| Acceptance | FAT, SAT, retest and failure procedure | Signed criteria and records |
| Operation | Calibration, power, spares and training | Annual cost and support scope |
Separate site survey, sensors and loggers, piping and electrical work, network and storage, testing and training, and maintenance in the price. Comparing a per point price without waveform retention or acceptance testing favors an incomplete offer. Compare initial and annual costs over five years as a planning example, including batteries, connectivity, storage and on site support. Our related guide on factory cooling water leak detection addresses make up water and leakage; short pressure transients require a different capture and test design. The factory IoT implementation guide explains the broader integration layers. An existing dashboard can be reused only after its effective waveform capture is verified.
FAT and SAT prove different things
At FAT, feed simulated pressure and operation signals through the real input path. Test a normal trace, high and low excursions, a broken sensor, communication loss, clock error, full memory and retransmission. Check the exported raw trace for sample count, timestamp, unit and correct pre and post trigger window. A dashboard demonstration alone does not prove that the measurement chain stores the waveform.
At SAT, check the actual installation: tapping point, leaks, power, communications, time synchronisation and mapping to pump and valve signals. Record normal baseline events and only those starts, stops or changeovers allowed by the plant’s safe operating procedure. Do not deliberately create a dangerous hammer event. If an accident scenario cannot be reproduced, accept normal event capture and simulated abnormal signal processing as separate tests. Do not claim that an untested emergency stop has been proven safe.

Possible acceptance statements include: every tag matches the P&ID; time offset is within the agreed range; specified events retain the entire pre and post window; sensor fault and communication loss produce distinct alarms; and records remain traceable after power restoration. Agree numerical limits from the site requirement before ordering. Do not combine an instrument acceptance criterion with a separate promise that a later hydraulic modification will reduce actual peaks by a given percentage.
Handle a failed FAT or SAT explicitly
A test record must contain more than screenshots of passing screens. For a failure, record the simulated input or operating condition, expected trace, trace actually stored, instrument settings, software version, timestamp and witnesses. Isolate whether the cause is sensor, pressure tap, input filter, logger, network or display, and record every correction. Making an alarm disappear by widening a threshold is not acceptance. After a change, repeat the failed case and related normal and fault cases, including communications loss.
If a safe pump changeover cannot be tested on the SAT date, leave that item marked “real event not yet verified” and agree who will witness the next planned changeover or shutdown. Passing a simulated FAT does not prove that the site pressure tap transmits a fast trace. Capturing a normal SAT event likewise does not guarantee that a dangerous historic peak would be detected. Put outstanding tests, interim operation and completion dates in separate fields in the acceptance record.
When testing sampling frequency, consider where individual sample instants fall around a narrow peak. A change in alignment can alter the saved maximum even for the same physical event. Repeat a test trace with different start offsets to check stable trigger detection and window capture. The purpose is to prove that the selected chain observes the intended phenomenon, not to prescribe one sampling rate for every plant.
Hand over the complete operating record
The handover pack should include point photographs, tapping and isolation details, tag list, calibration certificates, sensor ranges, sampling and trigger settings, clock method, storage location, backup, notifications and recovery procedure. Identify who holds the configuration software and administrator access. For a Thai floor team and Japanese management, align Thai alarm wording with the equipment IDs used in Japanese reports. Record pump, valve and piping changes on the same timeline as the pressure events. After a system change, review thresholds and verify a subsequent normal operating event.
Turn each event into a controlled improvement
When an alarm arrives, secure personnel and equipment under the site’s safety procedure. A pressure trace alone cannot prove that a pipe is undamaged. Compare its time with pump, valve, power and process records. Examine the high and low sides, repetition and order across points. Maintenance and a piping specialist can then test possible causes. Evaluate a proposed change, record its approval and compare before and after traces under similar flow and pump combinations.
Monthly indicators should include events by operating mode, maximum and minimum distributions, missing data, acknowledgement time, completed investigations and repeated events at one asset. Normalise event counts to pump starts or running hours where appropriate. If a change reduces noise but increases low pressure at another point, the system is not necessarily improved.
For economic evaluation, track leaks, gauge failures and stoppages separately from other maintenance work and changes in production. A short period without failures cannot prove that monitoring prevented them. The first deliverable is the ability to connect a pressure wave to an operating action and to test a remedy. Avoided downtime can be assessed later from a sufficiently documented history.
Example: a cooling water pump changeover
Suppose two cooling water pumps alternate and a bang is heard near a check valve just after each changeover. Do not label the check valve defective yet. Collect flow and pressure before switching, stop command and actual stop of the old pump, start command and actual start of the new pump, and the two check valve locations. Verify the as built P&ID and recent setting changes. If timing exists only in operators’ recollection, enable PLC event logging before the temporary pressure test.
Place temporary recorders at the common discharge header and near the suspect check valve. Record approved normal starts, changeovers and stops. Compare an event with noise and one without, including troughs and actual valve timing. An isolated sharp spike on one instrument could be a tap or instrument problem; related traces with a time offset call for propagation analysis. A change to the pump sequence or check valve must be reviewed by engineering, operations, maintenance and safety. Retest under comparable flow and pump combinations. Quieter piping alone does not demonstrate that the low pressure side is acceptable.
This method remains useful even if the initial hypothesis fails. Air entrainment, a manual operation or restart after power loss may prove more relevant. Mark measured and unmeasured operating conditions separately in the report, so an untested emergency scenario is not described as safe.
Define responsibility across suppliers
Instrument supplier, piping contractor, PLC integrator and maintenance software provider may all be different companies. Assign ownership for the pressure tap and leak test, cabling, clock synchronisation, data fields, alarm routing, site test and final drawings. Name witnesses and the person who signs acceptance. A gap here is a safety and delivery risk. Also specify waveform export, data return when a cloud contract ends, support life and backup restore testing for local storage. The purpose is a verifiable link between operating action and physical event, not merely a high specification sensor.
Show the observation boundary alongside the investment case
Before claiming an annual water hammer loss in a budget proposal, distinguish confirmed cause from suspected cause. A two year history of pressure gauge replacements is an actual expenditure, but not all those failures are necessarily due to a transient. Unplanned shutdowns may arise from valve defects, power, control settings or hydraulic events. The direct value of monitoring is a shorter investigation and the ability to test a remedy. Do not count an unproven promise of accidents prevented as a guaranteed benefit.
Attach the measurable and unmeasurable operating conditions to the quotation. A recorder may capture routine pump changeovers but lose the seconds before an external power outage if it lacks backup power and buffering. If power loss is the main concern, include power continuity, memory and clock retention in the scope. If the first stage is a temporary survey at a few points, a plant wide permanent communications contract may be unnecessary. Expand only when the cost and coverage of each stage are understood.
Specify the data delivered under the contract. It should include raw pre and post event readings, time, pressure units, calibration history, missing data flags, PLC tag mapping and pump and valve states. A screenshot alone cannot be reanalysed by a different engineer. Set the same export and ownership conditions when comparing suppliers. Safety should not be ranked solely by the monetary loss: a hot, corrosive or flammable liquid, or a line near people, can make a small leak serious. Review design pressure, relief, pipe supports and isolation with the plant’s responsible engineers. Do not assume that unsafe operation may continue until monitoring is installed.
Make retention and reanalysis part of acceptance
Historic waveforms may be needed months later when a pump is replaced or a valve setting changes. Define retention from the maintenance review cycle, equipment renewal plan, data volume and contract cost. Normal trends, high rate event traces, alarm history and communication gaps need not all have the same retention period. Decide what is retained, downsampled and deleted. If memory fills by overwriting old events, test the warning before that happens.
Clock accuracy matters when comparing several measuring points. A vendor screen may show aligned traces after software correction; verify whether it uses actual sensor capture time or an adjusted time. If a logger continues on its local clock during an outage, SAT should confirm that reconnecting and resynchronising does not silently rewrite historical timestamps. Keep clock correction history and distinguish data from before and after the correction.
Write a standard reanalysis procedure. Start from an event ID, open the asset, waveform, pump and valve log and maintenance ticket in an agreed order. Assign responsibility for labels such as “instrument fault”, “operation related” and “unresolved”. Compare traces only after aligning units, gauge versus absolute pressure, flow, open valves and start counts. Two charts from different conditions do not prove the benefit of a change merely because one peak is lower. Retain raw readings as well as processed plots so another engineer can review the conclusion.
When piping is extended, update the P&ID and tag list and reconsider whether old points represent a new branch. A changed connection may alter reflection locations, making the old normal waveform a poor alarm baseline. Periodically review sensor calibration, blocked taps, cables, power, connectivity, clock and storage. Report valid capture by asset. A green screen must tell the operator whether the plant was normal or no usable measurement was available. For night shifts and holidays, train the on-call handover and log the first response; feed questions found in drills back into the procedure and alarm wording.
FAQ: industrial pipe water hammer monitoring
Can existing gauges and SCADA detect water hammer?
Only if the complete chain has sufficient response and retains the event waveform. Check sensor dynamics, filtering, actual logging rate and pre and post trigger storage. A slow trend cannot prove the absence of a short peak.
Where should pressure sensors be installed?
Pump discharge, rapid valves, check valves and long branch ends are candidates. Choose final positions from the P&ID and transient assessment; also review the pressure tapping arrangement. Synchronized points help interpretation.
Is a 100 Hz logger sufficient?
There is no universal rate. Required bandwidth depends on the event and the complete instrument chain. A quoted 100 Hz rate is one product’s available option, not a guarantee for every pipe. Temporary measurement can establish the requirement.
Will monitoring prevent the hammer?
Monitoring identifies the event and supplies evidence. Prevention requires separate hydraulic and control design. After a change, the monitoring trace can verify the result.
What drives cost?
Point count, safe pipe tapping, process conditions, response, connectivity, time synchronisation, waveform storage, FAT/SAT and support all matter. Compare offers under one test scope with initial and annual costs separated.
Conclusion
Industrial water hammer monitoring is an investment in capturing short pressure traces and linking them to operations. Define a target circuit and credible events first. Specify effective response from pressure tap to retained data, sensor positions, clock accuracy and treatment of missing data. Prove simulated capture at FAT and real installation and approved operations at SAT. Use the evidence with a qualified hydraulic designer to choose and verify a protective measure.
If pump or valve changes are repeatedly followed by pipe noise or gauge damage, contact TOMAS TECH to discuss a small temporary measurement study and a testable RFP for your existing plant.
Sources
- KSB, Surge pressure: pressure wave, low pressure and basic equation.
- KSB, Transient flow analysis: pumps, valves, check valves and modelling.
- KSB, Risk of pressure surges: system specific mitigation.
- US EPA, Water Distribution System Analysis Reference Guide: definition and analysis.
- US EPA, State of Technology Report for Force Main Rehabilitation: short transients and repeated loading.
- HWM Global, Pressure Transient Logger: example of event recording architecture.
- HWM Global, Pressure Transient set up: example trigger settings.
- US EPA, Pressure Transients and Contaminant Intrusion: drinking water case informing the low pressure discussion; its public health conclusions are not applied to process water.