A transformer repair can accumulate hundreds of measurements and still lack one decisive piece of design knowledge. The missing item may be a winding preload, an insulation interface or the loss basis used for a declared rating. More measurements do not automatically recover it. [1]
A design-gap register makes the missing knowledge explicit. Its purpose is to connect uncertainty to a decision: what can proceed, what needs a revised design and what cannot be claimed from the available record. It should accompany the repair drawings rather than disappear when those drawings look complete.
Identify the authority behind each value
Mark whether a value comes from an identifiable original document, an as-found measurement, a test interpretation or a new repair design. A reverse-measured dimension can be highly reliable without being the original nominal dimension. Wear, compression, fault forces and dismantling can all change the state being measured.
A copied legacy drawing also needs context. Its revision may predate a previous repair. Match its terminal, component and dimensional identities with the unit actually opened. Where they disagree, retain both records and investigate the difference.
| Gap | Available evidence | Decision affected | Resolution route |
|---|---|---|---|
| Original clamping condition unknown | Measured heights and surviving hardware | Mechanical support design | New justified restraint assessment |
| Insulation material identity missing | Location and accessible condition | Reuse or replacement | Qualified material identification or replacement |
| Loss-test basis unclear | Historical total watts without conditions | Performance comparison | Recover conditions or limit the comparison |
| Regulating connection uncertain | Partial lead trace and tap ratios | Voltage schedule | Corroborated circuit reconstruction |
| Cooling design data absent | Passage dimensions and external equipment | Declared thermal duty | New thermal assessment and verification |
Write the consequence, not just the missing item
Unknown spacer grade is incomplete as a register entry. Explain whether that uncertainty affects compression behavior, dielectric performance, dimensional stability or all three. The consequence determines the resolution needed. A replacement material that solves one concern may create a new fit or compatibility issue.
Give each gap an owner and a disposition state. Resolved should mean that evidence supports the relevant decision, not merely that a drawing has been issued. A newly engineered solution resolves the repair design while leaving the original construction unknown; preserve that distinction.
Oria and co-authors’ pressboard compression dataset retains material conditioning, ageing and load-cycle history. A free dimension cannot substitute for the compression response needed to recover a restraint design. [2]
Turn an unknown preload into a decision range
Assumptions: final cold preload is 50 kN, the repair requires at least 40 kN after seating, the stack may settle 0.4 mm, and plausible effective clamp stiffnesses are 20 and 30 kN/mm. Force losses are 8 and 12 kN, leaving 42 and 38 kN respectively. The lower-stiffness case passes by 2 kN, while the higher-stiffness case fails by 2 kN. Because the unresolved stiffness range reverses the release decision, hold release and determine the actual load-displacement response or redesign the restraint. Another accurate unloaded height cannot close this design gap. [2]
Keep assumptions visible through verification
An assumption should carry a range or a stated alternative where possible. If several plausible geometries satisfy the terminal ratio, do not convert one into a recovered fact because it is convenient to manufacture. Assess what changes across the alternatives and which observation could separate them.
References
- U.S. Bureau of Reclamation, FIST Volume 3-31, Transformer Diagnostics, June 2003.
- C. Oria et al., Experimental dataset on the compressive mechanical properties of high-density pressboard used in power transformers spacers, Data in Brief 50, 109471, 2023. DOI: 10.1016/j.dib.2023.109471.