Transformer Test Benches: Factory Acceptance Versus Site Testing
Send your test requirement — get a tailored solution and a quote within one working day.
A transformer is tested twice in its life before it carries load: once in the factory, where the measuring conditions can be controlled, and once on site, where they cannot. Projects that confuse the two end up paying for loading tests that can never be accurate on a construction site, or skipping the post-transport checks that are the only evidence the unit survived the journey. Getting transformer test bench factory acceptance scope and site scope right is a procurement decision with a direct cost consequence.
The two stages answer different questions. The factory stage proves the unit meets its guaranteed characteristics under reference conditions. The site stage proves the unit is still the same unit after shipment, storage and installation, and that nothing was disturbed in the process.
What a test bench takes over
A transformer test bench is an integrated measurement station that runs the core electrical tests on an assembled unit without reconnecting between them. It brings together turns ratio, winding resistance, insulation resistance, capacitance and dielectric loss measurement, and in larger configurations it also manages the dielectric tests and the loss measurements performed with external equipment.
The value of the integration is not speed in isolation; it is consistency. Each test is performed with the unit in the same configuration, at the same tap positions, with the same temperature reference, and the results are recorded in one file against one unit identification. That consistency is what allows a factory result to be used as a reference years later, because the conditions under which it was taken are documented along with the values.
A bench also removes a class of error that is otherwise common in production testing: the wrong-tap and wrong-winding mistake. On a three-winding transformer with a multi-position tap changer, the number of possible connection combinations is large, and a systematic connection set managed by the bench reduces the opportunity to test something other than what the record claims.
The measurement definitions that a bench is designed to satisfy come from the transformer standards. Routine and type tests, and the conditions under which each is valid, are set out in IEC 60076-1, with the dielectric test framework in IEC 60076-3 and the thermal basis for the reference temperatures in IEC 60076-2.
Factory acceptance scope and its limits
Factory acceptance testing covers the routine tests performed on every unit and, where the contract requires it, the type and special tests performed once per design. The routine group establishes ratio, winding resistance, dielectric strength, no-load loss and current, and short-circuit impedance and load loss. The type group adds temperature rise, impulse withstand and, for some designs, sound level and short-circuit withstand evidence.
The limits of the factory stage are structural rather than technical. The unit is tested in a controlled environment with a supply designed for the purpose, and the results are referenced to defined conditions. That is what makes them comparable against guaranteed values, and it is also why the conditions cannot be reproduced on a construction site. A load loss measurement taken from a site generator will not match a factory figure, and the difference will not indicate a fault.
A second limitation is that the factory result reflects a state that no longer exists once the unit is drained, shipped and refilled. Oil condition, moisture content and the mechanical state of the windings after transport are outside the factory test scope by definition. The factory test is therefore a reference point, not a statement about the delivered condition.
| Test | Factory | Site | Reason for the split |
|---|---|---|---|
| Turns ratio across the tap range | Yes, against guaranteed values | Yes, compared with factory | Transport can disturb tap connections; ratio is fast to repeat |
| Winding resistance | Yes | Yes, corrected to the same reference | Detects connection damage and gives a post-transport baseline |
| Insulation resistance and polarisation index | Yes | Yes | Moisture ingress during transport and oil filling is a real risk |
| Capacitance and dielectric loss | Yes | Yes | Detects winding movement and moisture; needs a comparison value |
| No-load loss and current | Yes, under reference conditions | Not meaningful | Requires a low-distortion supply and accurate power measurement |
| Load loss and short-circuit impedance | Yes | Not meaningful | Same reason, plus the need for a stable high-current source |
| Temperature rise | Type test where required | No | Requires a controlled load and a long stable thermal period |
| Frequency response fingerprint | Where specified | Yes, and it is the reference for the future | The post-transport sweep becomes the practical baseline |
| Oil dissolved gas and moisture | Before filling | Yes, after filling and standing | Different stages of the oil lifecycle |
What must be repeated on site
The site programme is a comparison programme. Its job is to establish that the unit delivered matches the unit tested, which is why the tests repeated on site are the ones that are sensitive to transport and installation: ratio, winding resistance, insulation resistance and polarisation index, capacitance and dielectric loss, and a frequency response sweep.
Two tests deserve particular attention. The frequency response sweep taken after installation is the fingerprint that all future diagnostics will be compared against, so the conditions under which it is taken matter as much as the trace itself. If the lead arrangement and grounding are not recorded, the reference is worth less than the paper it is printed on.
Dissolved gas analysis on the oil after filling and standing is the second. Transport and oil processing can leave the oil in a state that differs from the factory sample, and the post-filling sample is the baseline for every later comparison. It is also the sample that catches a unit that has been damaged during shipment, because a mechanical disturbance that produced no immediate electrical change may still generate gas over the following weeks.
Routing a unit through both stages without duplication
The duplication risk sits in the middle of the two programmes. Both stages can measure ratio, resistance and loss, and a specification that lists the same tests in both places produces a site programme that repeats work the factory already did well, at a cost and with a worse result.
The routing rule that avoids this is to assign each test to the stage where it is valid, then record the handover explicitly. Ratio, resistance, insulation and loss measurements belong to the factory as guaranteed-value tests and to the site as comparison tests. The comparison is only meaningful if the factory value is handed over in a form that can be compared, which means on the same reference temperature and with the same test conditions stated.
Where a test genuinely cannot be repeated on site, the handover record should say so and identify the factory value as the only evidence. Where a test is repeated, the site record should identify the factory value it is being compared with, so that a later reviewer does not have to reconstruct the comparison from two separate documents. Independent acceptance practice such as the material published by NETA follows the same split between installation acceptance and equipment verification.
Specification points for a test bench
A bench specification is usually written around the tests it must perform, which leaves out the properties that determine whether its records are usable. The measurement uncertainty for each quantity, the accuracy class of the reference dividers and shunts, the temperature measurement and correction method, the resolution of the ratio measurement, and the way raw data is stored all matter more over the life of the bench than the test list does.
The second group of specification points concerns the interface to the unit under test. The number of independent voltage and current channels, the rated voltage and current of each, the tap selection capability and the connection hardware all determine which transformer classes the bench can handle without external equipment. A bench that covers distribution units but needs a separate arrangement for a 500 kV class winding is not necessarily a problem, provided the limitation is known before purchase rather than discovered on the first large order.
The third group is the record itself. The bench should export a machine-readable file per unit, with the test conditions embedded, so that the values can be compared rather than merely referred to. A printed certificate is a deliverable to the customer; the data file is the asset that supports every future comparison.
Comparison of factory and site evidence
The two evidence sets are not interchangeable and should not be presented as if they were. Factory evidence establishes performance against a guarantee. Site evidence establishes condition against a reference. A factory test failing a guaranteed value is a commercial event; a site test disagreeing with a factory value is an engineering question.
The most useful way to present the pair is side by side, with the conditions stated for each. A table showing the factory value, the site value, the temperature basis for each and the deviation allows a reviewer to judge whether the difference is explained by conditions. Without the conditions, the same table invites a conclusion that the data does not support.
Where the deviation is larger than the conditions can explain, the next step is almost always a repeat measurement under controlled conditions rather than an escalation. Repeating the site test at a lower temperature, or with a cleaner connection set, resolves more disagreements than any amount of analysis.
Documentation handover
The handover that matters is from the factory test bay to the site commissioning team, and from the commissioning team to the operating asset record. Each step loses information unless the transfer is defined, and the loss is usually in the conditions rather than in the values.
A workable handover package contains the factory test report with test conditions, the agreed guaranteed values, the reference temperature basis for every corrected quantity, the raw frequency response files where they exist, the oil test certificates for the oil as delivered, and a statement of which tests are to be repeated on site. The commissioning team returns a comparison against each of those values, plus the post-installation frequency response fingerprint that becomes the baseline for the asset.
The commercial reason to invest in this handover is that it is the only stage at which a transport or manufacturing problem can be attributed with confidence. Once the unit has been in service, attributing a defect to the manufacturer, the transporter or the installer becomes an argument rather than an investigation.
Cost and schedule implications
The cost of the factory stage sits in the unit price, and the cost of the site stage sits in the project’s commissioning budget. Splitting them properly reduces both, because each programme is sized to the tests that only it can perform.
The schedule consequence is more interesting than the cost consequence. Factory testing is on the critical path between assembly and shipment, so any additional test extends the delivery date. Site testing is on the critical path between installation and energisation, and its duration is set by the number of tests and the number of connection changes. A bench that reduces connection changes shortens the factory stage; a site programme that omits invalid loss measurements shortens the site stage. Both are schedule decisions as much as engineering ones.
Where a project is running against a fixed energisation date, the useful question is not how many tests can be fitted into the site window but which tests would produce a decision if they were run. The transformer research programme published by EPRI and the asset management material coordinated through CIGRE both address that question from the utility side, and the short-circuit withstand context for units entering service on networks with high fault levels is defined in IEC 60076-5.
If your site test programme still includes load loss or no-load loss measurements, those results will not be comparable with the factory figures.
Send your factory test schedule and your site commissioning scope to our engineering team and we will mark the rows that should move to the other stage. Integrated test benches are listed on the transformer commissioning test equipment page, within the wider power transformer testing range.
FAQ
Which tests can only be done in the factory?
Tests that need a controlled supply, accurate power measurement and a stable thermal environment belong in the factory. No-load loss, load loss, short-circuit impedance and temperature rise all fall into that group, because their accuracy depends on the quality of the measuring system and on the supply being free of distortion. Repeating them on site produces a number that cannot be compared with the guaranteed value.
What must always be repeated after transport?
Ratio, winding resistance, insulation resistance and polarisation index, dielectric loss, and a frequency response sweep where a fingerprint exists. Oil samples for dissolved gas and moisture content belong in the list because transport, oil filling and standing time all change the oil condition. The purpose of the repeat is comparative, not absolute, and the comparison is against the factory or pre-shipment values.
How does a test bench reduce test time rather than add to it?
By removing the reconnection step between tests. Ratio, resistance, capacitance and dielectric measurements that would otherwise be made with separate instruments and separate connection sets are made in one pass with the unit in one configuration. On a production line where every unit is tested, the saving compounds, and it also removes the class of error that comes from connecting to the wrong winding or the wrong tap.
Does a site test need the same accuracy as a factory test?
No, and demanding it is counterproductive. A site measurement is judged on whether it matches the factory or pre-shipment value within a margin that accounts for temperature, connection and supply conditions. Chasing factory-level precision on site produces a measurement that is more precise than it is accurate, and frequently a disagreement that has to be explained rather than a finding.
Who signs off a factory acceptance test?
Normally the purchaser’s representative or an appointed third party, alongside the manufacturer’s test engineer. The signature confirms the witnessed test, not the design. Because the routine tests are performed once and are expensive to repeat, the witnessing arrangement and the agreed test schedule should be settled before the unit enters the test bay rather than on the day.
