A generator acceptance testing procedure is the witnessed, documented sequence that proves a generating set meets the agreed specification twice: once at the manufacturer’s works before despatch, and again at the site after installation. The second run is not a repeat for its own sake. The first gate proves the machine, the second proves the installation, and each one ends with a signature that releases something.
Here is the part most buyers miss. An acceptance test is optional.
ISO 8528 separates a functional test, which must always be done, from an acceptance test, which is discretionary and normally takes place on site. Avtron, which builds the load banks used to run these tests, puts it plainly: acceptance tests are optional, may be done on site, and are often witnessed by the customer or a representative. If your contract doesn’t ask for one, you don’t get one.
Daniel, a plant engineer at a cold-storage facility in Mombasa, found that out the slow way. His 1,250 kVA set arrived with a “factory tested” certificate that ran to a single page: no serial number, no load figures, no signature. Within a month it was dropping frequency every time a large refrigeration compressor started, and when the dispute reached the contract there was nothing to measure against.
The set may well have passed. The paperwork couldn’t prove it.
This guide walks through the load steps, the pass and fail line, and the evidence that closes each gate. Shandong Huali has built and tested diesel generator sets from 8 kVA to 4,000 kVA since 1999, and we sit on the manufacturer’s side of this table every week. If you’re holding a FAT protocol you’ve been asked to sign, send it to our engineers and we’ll tell you what it does and doesn’t prove.
Key Takeaways
- An acceptance test is a contractual gate, not a maintenance activity. The factory test releases the set to despatch; the site test releases it to handover and payment.
- The gate is optional unless your contract requires it. ISO 8528 treats it as separate from the mandatory functional test and expects it to be witnessed.
- Expect two gates on most projects: FAT at the works and SAT on site. Integrated systems testing (IST) is added where sets parallel and feed UPS or life-safety loads.
- A purely resistive load test is an incomplete test, because sets are rated at a lagging power factor, usually 0.8.
- The acceptance test report is the deliverable. A set that passed with no signed record did not, in practice, pass anything.
FAT, SAT and IST: The Three Acceptance Gates
Acceptance is a contract word before it’s an engineering word. It moves money and starts a warranty clock, which is why the most important preparation happens before any fuel is burned: agreeing what will be tested, to what limits, and who signs. Most projects need two gates; larger or paralleled installations add a third.
| Gate | Where and when | Who runs it | Who witnesses | What it releases |
|---|---|---|---|---|
| FAT (Factory Acceptance Test) | At the manufacturer’s works, on the completed set before shipment | Manufacturer’s test engineers | Buyer, buyer’s representative, or a third-party inspector; in person or by live video | Releases the set to despatch |
| SAT (Site Acceptance Test) | At the installation site, after the set, cabling, fuel system and transfer switch are in place | Installing contractor and commissioning engineer | Commissioning authority, owner’s engineer, manufacturer’s representative | Releases the set to handover, and often to final payment |
| IST (Integrated Systems Test) | On site, with the whole power system live | Commissioning team | Owner, consultant, and often the utility where a grid interface exists | Releases the system to operation |
A standalone standby set feeding a building usually stops at SAT. IST becomes necessary when sets run in parallel, when they carry UPS or life-safety loads, or where the site has to survive a black-building start with no mains present. PowerTech Generators, which commissions data-centre plants, stages the programme this way and notes that the whole generator commissioning procedure runs one to three weeks per plant.
Running the Generator Acceptance Testing Procedure at Each Gate
The generator factory acceptance test is your one chance to catch problems when they’re cheapest to fix. Once the set is in a container on a ship, every defect becomes a site visit, a spare part, and a delay. That’s why the protocol matters more than the paperwork.
At the Works: The Factory Acceptance Test
Before the test starts. The protocol should be agreed and approved before the day, not negotiated on the day. Instruments must be calibrated with traceable certificates, typically dated within the last twelve months. The load bank must be the right type and size for the set, and our guide to load bank testing for diesel generators covers resistive, reactive and combined banks and how to size them.
Identity and configuration. Nameplate kVA, duty rating, alternator class and controller firmware version are recorded and matched to the order, and serial numbers go on the camera. Governor stability and AVR configuration are verified next, along with every alarm and protective shutdown.
The load and transient tests. The set runs through a stepped load schedule, holds at full load, and then takes the acceptance and rejection tests. The sequence follows below.
The evidence pack. You should leave with a signed test matrix, calibration sheets, transient curves and test certificates, which PowerTech describes as the factory output chosen specifically to de-risk the site test that follows.
On Site: The Site Acceptance Test
The generator site acceptance test is where the set meets your fuel, your cooling air, your cable and your transfer switch for the first time. Shenton Group describes the factory stage as proving the equipment is correctly assembled and operates to specification under simulated conditions, and the site stage as verifying it after installation. You need both because simulated conditions aren’t real ones.
Beyond the load test, a site acceptance test adds the items only a real installation can prove: a cold start and a hot start, every safety shutdown simulated for real (low oil pressure, high coolant temperature, overspeed, emergency stop), fuel transfer interlocks, ventilation dampers, protection settings against the approved list, metering accuracy, the BMS or SCADA points, and open and closed transfer with bypass isolation and controlled retransfer timing. If any of that groundwork is still outstanding when the test is booked, our diesel generator installation and ATS guide covers it. The stepped load schedule then runs again on site, usually with a rented load bank.
If you need a document to work from on site, our free diesel generator commissioning checklist works as a generator acceptance test checklist for the site half, covering the pre-commissioning inspection, first start, load bank steps, transfer tests and protection trips across eight tabs and more than fifty checkpoints. This guide covers the wider gate, including the factory half that the checklist doesn’t reach.
The Step-by-Step Generator Acceptance Testing Procedure
Read this as one sequence. The factory version stops before the transfer tests; the site version runs the whole list.
- Agree to the protocol. Test scope, pass limits, load bank type, instrument calibration, and witness rights are approved before the day.
- Inspect before you start. Safety checks, earthing and bonding, guarding, insulation resistance, fluid levels, and torque records on power connections.
- Verify the emergency stop first. Before anything else runs, the stop circuit is proven.
- Start and run at no load. Check frequency, voltage, phase rotation, and look for leaks, unusual noise and vibration.
- Step the load up. Apply load in steps and hold at each point until readings stabilize.
- Hold at full load. Around an hour at the works, longer on site; this is the heat run that proves cooling, fuel delivery and thermal margin.
- Run the transient acceptance test. Apply the specified load step and record the voltage and frequency dip and the recovery time.
- Run the load rejection test. Drop the full load in one step and confirm the set stays within its limits and keeps running.
- Trip every protection. Prove each alarm and shutdown by simulating the fault.
- Exercise the transfer switch. Simulate a mains failure, confirm auto-start, transfer, return and cool-down.
- Cool down, record and sign. Readings, curves, punch list, signatures.
How to run the generator load acceptance test. A commonly specified factory schedule runs 25%, 50%, 75% and 100% of rated output, holding half an hour at each middle step and an hour at full load, then taking 110% overload for half an hour. Overload is capped at an hour in any twelve. Site profiles are shorter and heavier, and a widely used emergency-power profile runs at 50% and 75% for half an hour each, then at 100% for at least two hours. Confirm the governing profile before the load bank is booked.
Split the scope into checks and measurements. ISO 8528-6, the test-methods part of the family, organizes the extent of an acceptance test that way, and it’s a good test of any protocol you’re handed: a document listing only measurements is unverifiable, and one listing only checks is unprovable.
| Checks, verified by presence | Measurements, recorded with a pass line |
|---|---|
| Nameplate data, serial numbers, duty rating, alternator class | Steady-state voltage and frequency at each load point |
| Fluid levels, fuel quality, filter and belt condition | Voltage and frequency dip on load application |
| Earthing and bonding continuity, cable terminations and torque | Voltage and frequency rise on full load rejection |
| Insulation resistance and rotation | Recovery time back into the specified band |
| Guards, covers, alarm annunciation | kW, kVA, current, power factor, fuel consumption rate |
| Protection settings against the approved list | Oil pressure, coolant temperature, exhaust temperature |
| Controller firmware and configuration version | Total harmonic distortion, noise level, vibration |
If the protocol you have been asked to sign is missing an entire column, ask which one and why.
Pass/Fail Criteria for Generator Acceptance Testing
The pass line comes from three places, and they don’t always agree: your contract, the performance class named in the specification, and the site’s own load profile. When they conflict, the contract wins. Name the class in writing and require the test report to state it, which is the clause our generator set specification for tender guide shows how to write so bidders can’t dodge it.
That class letter is the acceptance line. A set specified to ISO 8528-5 G2 must hold roughly ±2.5% steady-state voltage and ±1.5% frequency, with a dip no deeper than about 20% and recovery in the region of six seconds. G3 tightens those to about ±1% and ±0.5%. Our generator performance class G1 G2 G3 G4 guide sets out the full limits, and the generator fails load bank test guide tables the steady-state criteria a pass has to meet.
Then check the power factor. This is the most common way an acceptance test is quietly invalidated. Sets are rated at a lagging power factor, usually 0.8, because real loads are inductive. Avtron notes that ISO 8528 requires test reports to state when tests were run at a power factor different from the rated one, and concludes bluntly that tests done with a purely resistive load can usually be considered incomplete.
Priya, an electrical engineer at a textile mill in Tamil Nadu, learned this one at the worst possible moment. At the witnessed factory test, a 60% block step was taken on a resistive-only load bank, and she signed the matrix.
Eight months later, running real inductive load, the AVR showed voltage excursions well outside the G2 window her specification named. Nothing had degraded. The set had never been tested at its rated power factor. The retest cost a load bank rental, a site visit, and a week of arguments about who paid.
The same logic applies to the step size. The load step is tied to the engine and its match to the alternator, historically 100% for naturally aspirated engines and 60% for turbocharged ones, now set by a formula. Define your own load steps instead of the ISO ones and the set falls outside the fixed G1 to G3 classes.
Witnessing, Evidence and the Acceptance Test Report
A generator witness test is only worth the paper it ends up on. If nobody with an interest in the outcome watches it, you have a manufacturer’s report, not an acceptance test.
Put the witness clause in the contract. State who may attend, how much notice is required, and what happens if the buyer can’t travel. Shenton offers factory tests attended in person or by live remote video feed, and notes that a factory acceptance test isn’t legally required but is strongly recommended for critical power systems. The remote option isn’t a downgrade if you specify it properly: an unedited continuous feed from the start of the test to the signature, serial number visible in frame and the instrument display and data logger in shot throughout. An edited highlights reel proves nothing, because the interesting failure is usually the gap between clips.
Use a third-party inspector when distance wins. For a buyer eight thousand kilometres away, an inspection agency on the factory floor is often the most cost-effective witness available. HQTS, which provides this service, is direct about why it exists: some factories might misinterpret test reports or manipulate testing procedures, and an independent inspector has no motive to do either.
Collect the evidence pack. The test is a means to the report, and a report missing any of the items below is worth arguing about before you sign, not after.
| Evidence item | Why it matters |
|---|---|
| Signed test matrix | Ties each required test to a result and a signature |
| Recorded measurements at each load point | Shows the set held its limits under load, not just at no load |
| Transient curves for acceptance and rejection | Proves the transient performance the class or contract demands |
| Power factor used for the test | A resistive-only run cannot support a rated-PF claim |
| Calibration certificates for every instrument | Without traceable calibration the whole test is void |
| Protection settings and firmware version | Confirms the set left the works configured as specified |
| Nameplate photographs and serial numbers | Ties the report to your machine and no other |
| As-built drawings | The record the site team will work from for the next 20 years |
Then work the punch list. Many contracts add a prove-in period after the site test, during which anything that surfaces is put right at the contractor’s cost. Don’t release final payment against an open punch list.
Why Generators Fail Acceptance, and How to Avoid a Retest
Most acceptance failures are procedural, not mechanical.
| Failure mode | What it costs | How to prevent it |
|---|---|---|
| Instruments out of calibration | The whole test is void and must be rerun | Require traceable certificates dated within twelve months, checked on the day |
| Protocol not approved before the test | The buyer cannot enforce the limits, and the test proves nothing binding | Approve and sign the protocol in advance |
| Load bank the wrong type or too small | Transients cannot be verified; a resistive-only run invalidates a rated-PF claim | Match load bank type to the rated power factor and size it to the set’s kW rating |
| Set tested before it was ready | Wasted test slot, pressure to accept a partial result | Confirm fluids, fuel quality and instrumentation before the set starts |
| Site not ready, so the SAT runs at partial load | A partial run gets recorded as a pass | Separate partial-load reference runs from the acceptance test, and label them |
| Records lost or unsigned | The set passed, and nobody can prove it | Treat the signed report as the deliverable, not the test |
One more trap sits on the buyer’s side. A set that passes acceptance isn’t a set that will pass next year. Acceptance testing proves the machine at handover, and NFPA 110 acceptance test requirements for the emergency power supply system apply before a separate ongoing testing regime takes over. Our NFPA 110 generator maintenance guide covers the testing that follows.
Tunde, a procurement lead for a mining contractor in West Africa, pulled this together on a 4,000 kVA containerised set. His travel budget wouldn’t stretch to Shandong, and the contract’s witness clause said “in person,” which in practice meant no witness at all.
The fix took one line in the purchase order: remote witnessed FAT, live and unedited, serial number taped to the frame and the instrument display in shot, plus an inspection agency on the factory floor. The site test ran at the rated power factor with a rented load bank. Two gates, two signatures, no argument six months later.
If you want the proof in person, our engineers can host you at the factory for a witnessed FAT on your own set, or bring you in live by video.
Generator Acceptance Testing: Frequently Asked Questions
What Is a Generator Acceptance Test?
A generator acceptance test is a witnessed, documented procedure that verifies a generating set against the agreed contract specification, at the works before despatch and again at the site. Unlike the mandatory functional test, it is optional unless the contract requires it, and it produces the signed evidence pack that releases the equipment.
What Is the Difference Between FAT and SAT?
A FAT, or factory acceptance test, runs at the manufacturer’s works on the completed set before shipment and proves the machine meets its rated and specified performance. A SAT, or site acceptance test, repeats the key verifications after installation and proves the set works with the actual fuel supply, cooling, cabling, transfer switch and site loads. The FAT releases to dispatch; the SAT to handover.
Is a Generator Acceptance Test Mandatory?
No. ISO 8528 distinguishes a functional test from an acceptance test. The functional test must always be performed and usually takes place at the factory; the acceptance test is optional, normally runs on site, and is often witnessed by the customer. It becomes mandatory only when your purchase contract or specification requires it, which is why the acceptance clause is worth writing carefully.
What Tests Are Performed at a Generator Acceptance Test?
Pre-start inspection, emergency stop verification, a no-load run, and stepped load testing at typically 25%, 50%, 75% and 100% of rated output. Then a sustained full-load hold, acceptance and rejection transient tests, protection and alarm trip tests, and transfer switch operation under simulated mains failure. Every result is recorded against the specified limits.
How Long Does a Generator Acceptance Test Take?
A site acceptance load test commonly runs one to four hours including load steps, holds and cool-down, depending on the profile and the fuel available. The full gate is longer, since preparation, start-up checks and documentation add most of a day. Factory tests usually fit into one working day per set; a parallel plant program runs for one to three weeks.
What Happens If a Generator Fails Acceptance?
The defects go on a punch list, are corrected, and the affected tests are rerun until they pass. Failure at one step usually doesn’t void the whole test; the set is fixed and retested to closure. Get the retest scope agreed in writing before any repair starts, and treat an open punch list as a reason to withhold final payment.
Conclusion
The generator acceptance testing procedure answers one question at each stage: has this set proved, on the record, that it meets what we agreed? The factory test proves the machine and releases it for dispatch. The site test proves the installation and releases it for handover. Both are optional unless you write them into the contract, and both are only as strong as the evidence they leave behind.
Three habits carry most of the value. Name the performance class in the specification so the pass line isn’t negotiable. Require the load test at the rated power factor. And treat the signed, serial-numbered report as the deliverable.
The set that passes twice with paperwork behind it is the set that runs when the mains drop. Send our engineers your duty rating, load profile and site conditions, and we’ll issue the FAT protocol, the acceptance criteria and the test schedule before you order, on sets from 8 kVA to 4,000 kVA at 50 or 60 Hz.
