FAT and SAT for vibration monitoring systems: acceptance plan

Minimum package

• FAT: identification of components, calibration, reference samples, trigger, full record, analysis, alarm, export and behavior after fault.
• FIT: connection of the recorder, platform, email, SMS and required integrations before dispatch to the site.
• SAT: point, mounting, axes, cable, power, transmission, time, test alarm and user access under field conditions.
• SIT: handover of events to external systems or procedures, if the project requires it.
• Every failed test ends with a deviation, deadline and retest. "To be fixed after acceptance" is not a pass.

Vibration monitoring systems are often accepted from the end. Three axes appear on the screen, someone triggers a test alarm, the email arrives. A signature appears under the protocol. No one checked whether the first peak of the event fit into the record, whether the frequency came from the signal, whether a missing fragment was exposed, and whether the calibration factor was assigned to the correct axis.

That kind of acceptance confirms that the interface works. It does not confirm measurement capability.

IEC 62381:2024 defines requirements and checklists for FAT, optional FIT, SAT and SIT for automation systems in the process industry. Structural monitoring is not exactly the same application, but the logic of the standard is very useful: before testing, the parties define scope, responsibilities, documentation and criteria, then adapt the plan to the specific system. An improvised vendor demo is not enough.

FAT, FIT, SAT and SIT are not four names for the same test

Factory Acceptance Test (FAT) takes place before shipment or deployment. The goal is to demonstrate that the components and functions meet the specification in controlled conditions. On the bench, you can inject a known signal, simulate packet loss and change a threshold without the pressure of the construction site.

Factory Integration Test (FIT) checks the connection of subsystems before delivery. For monitoring, this may include the recorder, communication gateway, platform, email service, SMS, time synchronization and the required external interface. IEC 62381:2024 added FIT as an optional, clearly named phase.

Site Acceptance Test (SAT) takes place after installation. It checks what cannot fairly be confirmed in the lab: the point and stiffness of mounting, orientation, cable route, coverage, local interference, power supply, weather, user accounts and actual notification time.

Site Integration Test (SIT) covers on-site integration with the client process. Example: the alarm reaches the dispatcher, the event is linked to the work log, and the shift operator performs the step from the response plan. If the project has no external integrations, the SIT scope may be small. It should not be added for decoration.

FAT does not replace SAT. Replacing a sensor, cable, bracket, configuration version or interface after FAT means the tested system has changed. You must state which tests need to be repeated.

Freeze the acceptance basis before the test exists

A test cannot settle an unspecified requirement. First you need an approved list of requirements and document versions. The structure of the data layer is worth taking from the data specification for the OPZ, and the minimum for the vibration chain includes:

• measurement purpose and operational decision;
• points, axes, physical quantity and unit;
• band, amplitude range, sampling and filtering;
• trigger, pre-trigger, post-trigger and maximum record;
• required metric, calculation method and tolerance;
• definition of completeness and response to a missing fragment;
• thresholds, hysteresis and alarm states;
• recipients, channels and maximum notification time;
• retention of waveforms, results and logs;
• export, roles, access and exit procedure;
• list of as-built documents.

If the requirement says "analysis according to the standard", the test has no unambiguous result. You must specify the document, edition, scope, point, quantity, tolerance and evidence. If the requirement says "real-time alarm", set the stopwatch: from which moment to which, and what limit in seconds or minutes.

A test sheet that leaves no room for gestures

Each test should have an identifier and one measurable objective. A good sheet includes:

Avoid the result "pass with comments" if the comment concerns a critical function. A minor documentation deviation with a deadline can be accepted, but not a missing pre-trigger, wrong unit or alarm without a recipient.

FAT plan: from input to evidence

1. Identification and configuration

Compare serial numbers, types, ranges, versions, calibration certificates, wiring and accessories with the delivery list. Save the configuration export, not just a screenshot. Check whether the factors and axes are assigned to the correct points.

2. Signal chain

Apply signals with known frequency and amplitude at several places within the band. Add a level close to noise, a typical level and a level near the upper limit. Check polarity, unit, range and saturation. If the sensor and generator do not allow laboratory excitation of the full set, state which part is simulated and which part is physical.

3. Sampling and time

Compare the number of samples with the record length. Check time monotonicity, UTC zone, behavior after restart, loss of synchronization and retransmission. For several devices, perform a common excitation and measure the time difference within the required tolerance. The full dependency of record parameters is explained in the guide to sampling, trigger and pre-trigger.

4. Trigger and window

Apply a series of pulses below, at and above the threshold. Verify the trigger moment, pre-trigger, post-trigger, re-arm and two events close together. Separately check a pulse occurring just before the end of the local buffer.

5. Completeness and transmission resilience

Interrupt the link during transmission. Remove one controlled fragment, repeat another and change the order. The system should recognize the duplicate, show progress and not present a partial record as complete.

In Inclify, the terminal incomplete state appears only when the latest time of a given stream has passed the end of the examined window and it is known that older parts will no longer arrive. Silence alone does not close the record, because the device may resume sending hours later. The test must therefore include both delayed closure and confirmed absence.

6. Analysis

Compare the waveform, FFT and band results with the reference calculation. Record the window function, scaling, frequency range, RMS or MAX metric and tolerance. A single-frequency test does not qualify all 21 bands.

7. Alarms and notifications

Run the sequence OK, warning, alarm and return, including hysteresis. Measure the time. Check email and SMS, recipient preferences, acknowledgement, mute for a defined time and trace in the history. Also verify the actual response procedure described in the TARP alarm response plan. A test message sent without an actual state transition does not test the alarm engine.

8. Faults, restart and recovery

Disconnect power, restart the system, restore the link and check the queue. Verify that retransmission does not duplicate the event. Check the procedure for fixing a failed analysis and whether the old, correct result does not disappear before the new recalculation is completed.

9. Export and audit

Export results, waveform or the available event package, alarm configuration and change list. Open the file with an independent tool. Log in with an administrator account and a read-only account to check permissions. Define how to protect a record under dispute from routine retention.

Numerical FAT scenario for one record

Assume a 10 Hz sine with acceleration amplitude 0.20 m/s², sampling at 256 Hz and a 30 s record, including 10 s of pre-trigger. The expected number of samples per axis is:

256 samples/s × 30 s = 7 680 samples

The 10 s pre-trigger should contain 2 560 samples. FFT should indicate a dominant component at 10 Hz within the tolerance resulting from record length and analysis configuration. The third-octave band covering 10 Hz should respond according to the reference calculation.

That is not a complete set of criteria. You still need to define tolerance for amplitude, phase, time, noise and spectral leakage. A generator signal does not prove correct installation on the building. It does, however, expose swapped units, a wrong factor, an incomplete record and wrong processing before dispatch.

In a separate test, remove one fragment. The result must not look like a full 7 680 samples after interpolation. The interface should show the missing part, and the procedure should define whether there will be retransmission, repair from the saved log, or terminal marking of incompleteness.

SAT plan: the laboratory ends at the site boundary

SAT starts with a check of the as-built documentation. The point on the drawing must match the physical point. Check the substrate, mounting, orientation, cable, glands, enclosure, power supply and labeling. You will find the detailed procedure in the guide on vibration sensor installation.

Then perform a safe, repeatable field excitation. Compare the response with the pre-delivery test, taking the difference in conditions into account. Do not try to "correct" the result by changing the factor without explaining the cause.

Test failure of the real link and power supply. Check alarm reception on the phones of people on duty, not on the integrator's number. Confirm that the read-only user sees the correct project, not another client's data.

SAT checklist for signature

• [ ] Points and axes match the as-built drawings.
• [ ] Mounting, cable and environmental protection passed inspection.
• [ ] Calibration and configuration match the supplied devices.
• [ ] The field test has a recorded excitation and criterion.
• [ ] The time of all devices is within tolerance.
• [ ] Every record has the correct window and completeness status.
• [ ] Loss of link does not delete the locally saved event.
• [ ] After the link returns, data arrive without duplicates.
• [ ] Analysis shows the correct unit and band range.
• [ ] The alarm passes through the real threshold and hysteresis.
• [ ] Email and SMS reach the approved recipients.
• [ ] Acknowledgement and mute record the person and time.
• [ ] Accounts and roles restrict access according to the contract.
• [ ] The export opens outside the platform and includes metadata.
• [ ] The service, retention and escalation instruction is handed over.
• [ ] All deviations have an owner and a retest deadline.

Business criteria that usually disappear from the protocol

A system can pass the technical test and still not be fit for work. Add operational tests. The night shift operator should recognize the project, point, level, time and required action without calling the dashboard author. The manager must be able to distinguish missing data from no vibration. The administrator should issue a package for the expert without writing a new script. These capabilities should also be assessed in the platform selection checklist before the FAT result becomes a procurement argument.

Measure the workload too. A sequence of one hundred weak triggers should pass without loss, and the team must have a way to prioritize them. If the pilot generates six hours of manual review per day, that is a process deviation even when all functions work.

Record the guaranteed service response time, calibration schedule, battery responsibility, retention and format of the final data issue in the protocol. These items decide whether the capability proven in FAT will survive six months on site.

What Inclify acceptance looks like

Inclify receives frames from devices over HTTP/JSON and stores measurements with UTC time. For recognized transfers, the administrator can inspect metadata, decompressed content and download the raw body during the log retention period, by default seven days. In FAT, this makes it possible to compare the sent input with the response and the result stored in the system.

For dynamic events, the platform stores the full waveform, analysis profile and counters of expected and received fragments. It shows progress in the UI. An incomplete record remains marked, instead of quietly pretending to be complete. A failed analysis can be repeated, and project recalculation publishes results only after all batches are finished.

Available profiles include FFT and 21 third-octave bands from 1 to 100 Hz. The profile with 4th-order Butterworth filters requires an effective sampling rate above 224 Hz and at least 10 s pre-trigger. Calibration is set separately for each axis.

Dynamic alarm evaluates RMS and MAX against the reference envelope. It has warning and alarm levels, plus hysteresis. State transitions, acknowledgement and mute for a defined time leave a trace, and notifications can go by email and SMS. A specific dynamic signal can be exported to XLSX, and the band-threshold configuration to CSV.

Inclify does not physically calibrate the sensor and does not qualify the mounting. It does not generate a PDF acceptance report with one click. It also does not approve the engineering threshold or the conformity of the full installation with the standard. Those proofs are provided by the measurement team and the person responsible for assessment.

Limits of FAT and SAT

A passed test proves performance under the tested conditions and version. It is not a promise for any temperature, band or number of events. If you change the sensor, filter, sampling, algorithm or threshold during operation, perform an impact analysis and repeat the relevant tests.

SAT does not replace an observation period. The battery, adhesive, connector and communications may work on day one and fail after a weather cycle. Set a stability test after one week and after one month, and define final acceptance criteria.

Most important: a positive test alarm does not prove the safety of the structure. It confirms the signal path and the procedure. The engineering decision is based on the approved criterion, the quality of the chain and the site context.

Sources and further reading

• IEC 62381:2024, FAT, FIT, SAT and SIT – current requirements, responsibilities and checklists for acceptance tests of automation systems.
• US Bureau of Reclamation, FIST 4-7 Commissioning Guide – the role of baseline measurements, FAT, test procedures and documented reference data.
• US Bureau of Reclamation, Design Standards No. 13, Chapter 11 – pre-installation tests, calibration, as-built documentation and instrument maintenance.
• ISO 4866:2010, Vibration of fixed structures – principles for measuring and processing structural vibrations.
• ISO 16063-21:2003, vibration calibration by comparison – calibration procedures for vibration transducers by comparison with a reference transducer.
• IEC 61260-1:2014, octave-band and fractional-octave-band filters – requirements for band filters used in third-octave analysis.

Before signing the protocol

Choose one critical requirement and ask for proof of the full chain: known excitation, full record, analysis, exceedance, message to the on-duty person and response entry. If any link cannot be shown, the system has not yet been accepted. Talk to the Inclify team if you want to run such a FAT and SAT scenario on a pilot chain before work starts.