A probe scanning a weakly excited core can respond to the applied magnetic field, geometric leakage and a local current path. The resulting signal is a mixture. A peak does not become fault evidence merely because the excitation is low.
Establish a phase reference from a measured excitation quantity and define the acquisition convention. If the reference changes between scans, a phase difference can be an instrument or timing change rather than a specimen change.
Retain both signal components
Represent the synchronized signal through its real and imaginary components relative to the stated reference, or through amplitude and phase with equivalent information. Do not assume that one component is always background and the other always fault current. Resistance, inductance, geometry and probe response can change their relationship.
Scan an intact representative region and a qualified controlled specimen with an independently known discontinuity. Compare spatial pattern and phase under the same excitation and probe arrangement. The controlled specimen should test a relevant mechanism, not merely create an exceptionally large signal.
| Validation step | Quantity preserved | Interpretation enabled |
|---|---|---|
| Excitation reference | Actual waveform and phase convention | Comparable complex signals |
| Intact scan | Position-dependent background | Geometry and main-field contribution |
| Probe-orientation comparison | Axis and standoff | Sensitivity to inductive pickup |
| Known-defect trial | Independent defect geometry | Demonstrated fault-related difference |
| Repeat scan after repositioning | Repeatability distribution | Confidence in a local indication |
The known-defect work belongs in a controlled professional laboratory procedure. It is not a recommendation to introduce a conductive bridge on an operating or production core.
Use direction as a diagnostic variable
Rotate or reposition the probe in a controlled comparison while preserving its location relative to the feature. Main-field pickup and a local perturbation may respond differently, but the expected transformation should follow the probe’s calibrated sensitivity and a justified field model.
Record standoff and track coordinates. Near a joint, small position changes can alter background pickup substantially. A difference scan is reliable only when the two trajectories are registered closely enough for the intended resolution.
Demonstrate the detection boundary locally
Vary relevant trial features such as depth, orientation or contact state and determine where the candidate response remains distinguishable from background and scan variation. Report ambiguous regions rather than forcing every signal into a binary fault class.
Criteria developed for another machine topology cannot be imported unchanged. Generator stator methods and their acceptance limits depend on their own geometry, excitation and sensor arrangements. A transformer packet requires a relevant capability demonstration.
The resulting scan record should include raw complex data, reference phase, trajectory and the comparison responsible for each indication. This allows an investigator to decide whether a feature is supported as a local conductive-path response, explained by pickup, or unresolved by the selected method.
Subtract the registered complex background
Use nominal 0.23 mm sheet thickness [1] as an assumed scan-position error. Assume the intact signal is 10.0 + j1.0 mV and the suspect signal 10.2 + j1.3 mV at the same low-flux excitation; the registered difference is 0.2 + j0.3 mV, magnitude 0.361 mV. With an assumed 2.0 mV/mm background gradient, one-sheet placement error contributes 0.460 mV and can dominate that difference. Reducing placement error to 0.05 mm reduces this term to 0.100 mV. Preserve both components and common phase reference, then combine repositioning, noise and excitation uncertainties [2]. A selected quadrature channel alone is no universal fault indicator.
Manufacturing controls. Joint shape, packet offsets, surface access and probe standoff define ordinary pickup. Qualify low-flux scanning on representative contact paths and fixed processing before using it to judge a repaired transformer core.
References
- Nippon Steel Corporation, Grain-Oriented Electrical Steel Sheets, catalogue D004je_03_202004, 2020, Specifications; Size and Form Tolerances.
- JCGM, JCGM 100:2008, Evaluation of measurement data — Guide to the expression of uncertainty in measurement, 2008.