What a Factory Audit Checks in a Transformer Supplier

Start with who designs the transformer, not who assembles it. Ask how many engineers work on electromagnetic and thermal design, what they are qualified in, and how long they have worked on the voltage class you are buying. Ask which software is used for short-circuit force and temperature rise calculation, and whether those calculations are reviewed and signed by a second engineer. A plant that can show a named design team and a review step is organised differently from one where the sales engineer also produces the drawings.

Then drawing control. The audit question is whether the drawing issued to the shop floor can be traced to a customer-approved revision, and whether change notes exist for every revision. Ask to see one complete package for a unit similar to yours — insulation clearances, winding arrangement, cooling design, tank and bushing details — and check that these are drawn rather than referenced. As a scale reference, one manufacturer's published research and design group is 72 engineers; the number itself matters less than whether it matches the order book.

Most latent defects are process defects, and process control is visible on the shop floor in about an hour. Look for what is recorded, not what is installed: a modern winding machine without a tension record tells an auditor less than an older machine with a chart for every coil.

Each stage should leave a record linked to the serial number of the unit. The audit question is therefore not "do you do this" but "show me the records for the last three units of this class and tell me what you did when a value was out of range". A plant that can answer the second question has a working quality system; a plant that can only answer the first has a certificate.

What a plant can test itself determines what it can catch before the unit ships. Ask for the impulse generator rating and the product classes it covers; the partial discharge measuring system and the background level actually measured in the test bay; the capacity of the test power supply for temperature rise tests at the largest rating the plant builds; and the accuracy class of the loss-measurement instruments.

One manufacturer's arrangement, as an example of industry practice: two impulse generators, 1200 kV for products up to 220 kV and 4800 kV above that; partial discharge background recorded per plant, ≤1 pC at the dry-type plant and ≤10 pC at the oil-immersed plant, against the IEC 60076-3 limit of 100 pC for Um above 72.5 kV; and a testing centre accredited by CNAS, registration L5393. The point for an auditor is whether the same number appears in the calibration certificate and the test report.

Ask also what the plant cannot do. Short-circuit withstand testing is the usual answer: it is destructive, it is performed on a sample of a design rather than on your unit, and it requires a high-power laboratory that most transformer plants — including ours — do not have. The evidence comes from an external laboratory, and the report should name the design tested. A supplier that describes short-circuit testing as an in-house routine is describing something that does not exist.

Certificates prove that a system exists; records prove that it works. Non-conformity handling is the most revealing file in the plant: how a defect is raised, who decides disposition, what is reworked and what is scrapped, and whether the same defect reappears in the next quarter's statistics. Traceability is the second test: from the serial number of a delivered unit, can the plant produce the winding record, the drying chart, the oil test results and the heat numbers of the core and conductor?

Material control is where cost pressure shows up first. Ask where the grain-oriented electrical steel comes from and whether mill certificates are kept for each delivery; the same for insulating paper and board, bushings, tap changers and gaskets. Ask how incoming material is inspected, what the rejection rate has been over the last year, and what happens to a batch that fails. A plant that cannot name its steel source cannot substantiate a core loss guarantee.

Capacity claims are easy to make and easy to test. Ask for annual output in MVA, the number of units shipped in the last twelve months by voltage class, the largest unit built and tested, and the current order book measured against capacity. A plant quoting a short delivery against a full order book is either prioritising you deliberately or estimating. Wood Mackenzie put power transformer lead times at 128 weeks and generator step-up transformers at 144 weeks in Q2 2025, so a delivery date that ignores the order book deserves a follow-up question.

For an export project, add the items domestic audits skip: drawings, test reports and installation manuals in English; packing for sea freight and long inland transport, including nitrogen or dry-air filling, shock and tilt recording, and seaworthy crating; commissioning and after-sales engineers who can work in English; and a spare parts and warranty process that works across borders. Check who prepares the documentation the owner's engineer and customs will ask for — a missing English type test report can delay energisation even when the transformer itself is ready.

A useful discipline is to score the answers rather than accept them. Three findings — an out-of-range drying record with no documented action, a partial discharge background that has never been measured, and a short-circuit report that does not name a design — will tell you more about a supplier than a wall of certificates.