Each programme answers a question the other two cannot. Commissioning answers whether the asset survived transport, storage and installation. Maintenance answers whether it is drifting from its own baseline. Diagnostic testing answers why a specific deviation appeared. Buying the wrong one produces a thick test file and a decision that still depends on opinion.

Which test proves which condition

Each test maps to one condition, not to overall health. A ratio measurement proves winding and tap position integrity, a DC resistance measurement proves winding and connection integrity, and insulation resistance proves bulk dryness and contamination. No single reading proves that a transformer is fit for another decade of service. The practical value of this mapping is that it tells a procurement team what a test file can and cannot defend. When a transformer fails in service three months after a clean commissioning programme, the review usually asks which test should have caught it. The honest answer is normally that no test in the programme was aimed at that condition. Ratio and resistance testing do not see partial discharge; an oil sample does not see a loose bushing connection. The table is also a procurement checklist. A test list that covers four of these six rows leaves two failure routes unmonitored, and the cheapest way to close the gap is usually an oil sample rather than another electrical instrument.

What commissioning testing can and cannot see

Commissioning is an installation audit. Its purpose is to prove that the unit that arrived on site is the unit that was ordered, that it was assembled correctly, and that nothing happened to it between the factory gate and the moment of energisation. It is done once, against a factory baseline, under time pressure, and every result is compared to a nameplate or a factory value rather than to a trend. What a commissioning programme sees well is gross deviation. A winding connected to the wrong tap, a reversed polarity on an instrument transformer, a phase swapped during cable termination, a bushing with a cracked porcelain, a winding with an open circuit: all of these produce an obvious reading. This is why the sequence matters and why IEC 60076-3 separates dielectric routine tests from the tests that are repeated after installation. What commissioning cannot see is drift, because there is no history yet. A transformer that arrives with 1.5 percent moisture in the paper insulation and a transformer that arrives bone dry both pass the same acceptance threshold on the first reading. The difference only becomes visible on the second sample, months later. The same limitation applies to partial discharge that only becomes active above service voltage, and to tap changer mechanisms that work correctly on the day but have a worn follower that will misalign after two hundred operations. Two practical consequences follow. First, the commissioning record has to be detailed enough to act as a baseline: the same conditions, the same corrected values, the same raw traces. Second, the first maintenance test is not optional housekeeping, it is the test that gives the commissioning result meaning.

Where maintenance testing adds new evidence

Maintenance testing is a comparison exercise. Nothing in a maintenance programme is judged against a nameplate limit in isolation; the verdict comes from the change since the last reading, corrected to a common temperature and reported with its measurement conditions. A unit that reads slightly high but perfectly stable is normally healthier than a unit that reads comfortably inside limits and has moved twenty percent in two years. The evidence maintenance testing adds is time. It shows whether the asset is stable, which is what actually determines the interval. A utility running a fleet of forty transformers cannot sample every unit every quarter on the basis of the standard alone, but it can use the slope of the trend to justify a longer interval on the quiet units and a shorter one on the units that are moving. Maintenance programmes also expose conditions that commissioning never had a chance to test. Tap changers accumulate operations and wear the diverter contacts. Oil oxidises and the acidity rises. Gaskets age and moisture ingress increases. Bushings and surge arresters accumulate thermal cycles. Each of these produces a slow, measurable change, and the standard reference for interpreting that change in mineral oil is IEC 60599, which frames the interpretation in terms of concentration and rate of rise rather than as a single threshold.

Diagnostic testing: when to escalate

Diagnostic testing is triggered, not scheduled. It starts when a routine reading moves, when an alarm scores an asset higher than its neighbours, or when a fault has already occurred and the question is whether the transformer can be returned to service. The test is selected to answer that specific question, and the report closes with a verdict rather than with a table of numbers. Escalation is justified when the cost of the test is smaller than the cost of being wrong. A dissolved gas reading that has doubled in six months is worth an SFRA sweep and a repeat oil sample, because the alternative is discovering a deformed winding during a forced outage. A single insulation resistance value that is a few percent below the neighbouring phase, measured in the rain, is not, because the difference is more likely to be a surface leakage artefact than an internal defect. Sequence matters as much as selection. Electrical results are compared against the same unit’s earlier data wherever possible, because two healthy transformers of the same design can differ simply in winding geometry and measurement position. Where no fingerprint exists, a sister unit makes a usable substitute. For frequency response work, the accepted method for comparing a measurement against a reference is set out in IEC 60076-18, which is deliberately written around deviation bands rather than a single pass or fail number.

Evidence each programme produces: comparison

The three programmes differ in more than the tests they contain. They differ in their reference, their repeatability expectation and the decision they support. Reading them side by side is the quickest way to see which one is missing from an existing strategy. The comparison also shows why the three cannot be collapsed into one budget line. Commissioning evidence is produced by the project, maintenance evidence by the operating budget, and diagnostic evidence by whichever budget carries the risk of an unplanned failure.

Choosing the programme for a new versus aged asset

A new asset needs a full commissioning programme and almost no diagnosis. The engineering question is whether the installation is correct and whether the baseline is trustworthy, so the money goes into coverage and into recording. A reduced commissioning programme on a new unit saves a small amount of site time and removes the only opportunity to capture a clean fingerprint of the winding geometry before any fault can occur. An aged asset needs the opposite balance. Its commissioning history is long finished, and the useful evidence now comes from frequency, trend and targeted diagnosis. Money spent repeating the full commissioning programme on a fifteen-year-old transformer buys confirmation of what is already known; money spent on oil sampling frequency, on trend review and on one well-chosen diagnostic test buys information about what is going to fail. The transition point is usually the first post-energisation years. During that period the asset behaves like a new unit if it survived transport, and the comparison against the commissioning baseline is the strongest signal available. After the baseline stabilises, the programme should shift weight from measurement coverage towards measurement frequency and interpretation. Wrindu manufactures the transformer test instruments that sit in both halves of this split, including tan delta sets, turns ratio and winding resistance meters, on-load tap changer analysers and sweep frequency response analysers, and supplies them to utilities, EPC contractors and test laboratories working on 33 kV to 500 kV equipment. The category overview is on the power transformer testing hub, with commissioning and maintenance equipment grouped separately.

What the outage window actually buys you

An outage window is a scarce resource, and the programme should be designed backwards from what can realistically be completed inside it. Taking a transformer out of service costs lost transfer capability and, on a network with constrained redundancy, a direct commercial consequence. The tests that need the window should be grouped, and the tests that can be done at another time should be moved out of it. Within a typical outage window, an experienced team can complete winding resistance on all phases at every tap, a turns ratio run across the tap range, insulation resistance and polarisation index, a tan delta measurement across the main insulation paths, and a set of SFRA traces. Adding oil sampling and a bushing check costs little extra time. Adding a full withstand test does not fit comfortably and usually needs its own plan. The order inside the window is not arbitrary either. DC tests magnetise the core, so an SFRA sweep taken immediately afterwards can read incorrectly, and a ratio measurement taken before the DC current has fully decayed can be unstable. Grouping the DC tests together and allowing the core to settle before the frequency response sweep is standard practice, and it is the reason a written sequence is more useful than a test list.

Recording results so they are comparable next time

Most test records fail not because the measurement was wrong but because the conditions were not written down. Temperature correction is the clearest example. A winding resistance value is meaningful only alongside the copper temperature used for the correction, and a ratio value is meaningful only alongside the tap position. When those fields are missing, the next engineer can still read the number but cannot compare it.

A record that survives a change of contractor should carry the test date and time, ambient and oil or winding temperature, the applied current or voltage, the tap position, the winding connections, the instrument model and serial number, the calibration reference, the raw trace where a trace exists, and the corrected value alongside the measured one. The conclusions section should state which standard or acceptance limit was applied, and whether the result met it. Storing the raw data matters more than it appears to. A printed pass or fail verdict cannot be re-analysed when a second measurement disagrees with the first, and re-running a test to recover lost detail costs the outage twice. For trace-based methods, the accepted quality benchmark for the recording system itself comes from IEC 60060-1 for high-voltage test techniques, and from the measurement-system requirements in its companion part.

How do you phase the first test programme on an ageing fleet?

Phase it by consequence, then by information value, then by convenience. Start with the units whose failure would remove the most capability, and give them a full first-pass programme so a baseline exists. Add units in order of criticality after that, not in order of age or location. The worst approach is a single sweep across the whole fleet in one financial year, because it produces a large volume of first-pass data with no follow-up schedule and no capacity left to interpret it. A staged approach costs less in any single year and produces something more useful: a growing baseline that can be compared as it is built. A workable first phase covers the highest-consequence units with a complete electrical set plus an oil sample, records everything as a baseline, and then schedules a second measurement on the same units within twelve months. The second pass is what converts the first into a trend, and the trend is what justifies moving a unit to a shorter or longer interval. Subsequent phases add the next tranche of units to the same baseline-plus-repeat structure. Interpretation capacity, whether in-house or external, should be sized at the same time as the measurement programme, otherwise the data accumulates faster than it can be read. Independent acceptance testing practices, such as those published by NETA, are a useful reference for structuring the acceptance side of the same programme.

Cost per asset of each programme

Cost should be compared on the same basis for all three programmes: the cost per asset per year, including the outage, the labour, the analysis and the instrument time. Purchase price of the instrument is a small part of the picture once the asset has been in service for a decade, which is why total cost of ownership calculations matter more in this category than in most equipment purchases. The economically useful comparison is not the cost of a diagnostic test against the cost of a routine test. It is the cost of the diagnostic test against the cost of the failure it may prevent, weighed by the probability that the trigger was real. Research material aimed at exactly this trade-off is published by the asset management study committees of CIGRE, and it is more useful for justifying an interval change than any single instrument datasheet.

Writing the programme into the maintenance strategy

A test programme becomes policy only when it is written into the maintenance strategy with an owner, an interval, a trigger and a stopping rule. Without those four elements, the programme degrades into an annual activity that produces data nobody reads, and the first time it is cut for budget reasons nobody can explain what was lost. The interval should be stated as a starting value with a rule for changing it, not as a permanent number. A common pattern is to sample and test at a fixed interval for the first few years, then extend the interval for units whose trend is flat and shorten it for units that show movement. Writing the rule down at the start keeps the interval from being renegotiated every budget cycle on the basis of opinion. Ownership should be explicit about the handover between the project phase and the operating phase, including the transfer of the commissioning baseline into the maintenance record system. Where that handover is unclear, the first maintenance result usually arrives without a comparison, which is precisely the situation the strategy was written to avoid. The reference frame for the transformer itself, including its rated characteristics and the tests that define them, remains IEC 60076-1. Finally, the strategy should name the decision each dataset feeds. Trend data sets the interval, trigger data sets the diagnostic test, and diagnostic results set the return-to-service verdict. When each dataset has a named consumer, the programme stops being a cost centre and becomes the mechanism by which the fleet is actually managed.