Frequency Response Analysis (FRA)

FRA measures the transfer function of a transformer winding by injecting a sinusoidal signal across a wide frequency range (typically 20 Hz to 2 MHz) and measuring the output. The resulting magnitude and phase plots form a unique frequency-dependent signature that reflects the distributed impedance network of resistance, inductance, and capacitance (RLC). Any mechanical displacement of the core or windings alters these parasitic parameters, changing the signature.

FRA is fundamentally a comparative test, not an absolute measurement. Diagnosis relies on three comparison methods:

• Time-based: Comparing a current trace to a baseline fingerprint taken when the transformer was known to be healthy.
• Type-based: Comparing the trace to an identical sister transformer.
• Phase-based: Comparing traces between phases of the same transformer. Deviations indicate mechanical changes.

FRA is uniquely capable of detecting mechanical faults that other electrical tests miss. It can identify:

• Winding deformation: Axial collapse or radial buckling due to short-circuit forces.
• Core displacement: Shifting of the core relative to windings.
• Clamping pressure loss: Relaxation of the pressboard structure.
• Turn-to-turn movement: Localized conductor displacement without full short circuit.

FRA is strictly an off-line diagnostic technique. The transformer must be completely isolated and de-energized before testing. This ensures that the low-voltage sweep signal is not contaminated by power frequency interference and that the test set is safely connected directly to the bushing terminals. The need for an outage makes FRA a scheduled maintenance activity rather than a continuous monitoring solution.

To quantify trace deviation objectively, numerical indicators are applied:

• CC (Correlation Coefficient): Measures the similarity of curve shape; a value near 1 indicates high similarity.
• SDD (Standard Deviation Difference): Quantifies the logarithmic difference between traces.
• ASLE (Absolute Sum of Logarithmic Error): Aggregates deviation across frequency bands. These metrics are defined in standards like IEC 60076-18 and IEEE C57.149.

Different frequency ranges reveal different physical phenomena:

• Low frequency (< 10 kHz): Dominated by core magnetizing inductance; sensitive to core deformation and residual magnetism.
• Mid frequency (10 kHz – 500 kHz): Dominated by winding capacitance and inductance interactions; most sensitive to winding buckling and hoop deformation.
• High frequency (> 500 kHz): Dominated by local stray capacitances and lead connections; sensitive to grounding issues and small local displacements.

The most common FRA variant that injects a sinusoidal signal swept continuously across a wide frequency range (typically 20 Hz to 2 MHz). The resulting transfer function magnitude and phase plots are compared against a baseline fingerprint. Deviations in resonant peaks indicate winding displacement, while shifts at low frequencies suggest core movement. SFRA is standardized in IEC 60076-18 and is the industry-preferred method due to its high repeatability and sensitivity to subtle mechanical changes.

A complementary low-frequency test measuring the leakage reactance of a transformer winding at power frequency (50/60 Hz). While FRA detects geometric displacement, short-circuit impedance quantifies the electrical consequence of that deformation. A change exceeding 1–3% from nameplate values is a strong indicator of winding compression or buckling. This test is often performed alongside FRA after a through-fault event to correlate mechanical damage with electrical parameter shifts.

The initial reference FRA measurement taken when a transformer is in a known healthy state, typically post-manufacturing or after successful commissioning. All subsequent measurements are compared against this fingerprint. Vector subtraction and correlation coefficients (e.g., Relative Factor, Cross-Correlation) quantify deviation magnitude. Without a reliable baseline, FRA interpretation becomes subjective, relying on phase-to-phase comparison or sister unit benchmarking.

An alternative FRA excitation technique using a fast-rising impulse signal rather than a swept sinusoid. The impulse contains broadband frequency content, allowing simultaneous excitation across the spectrum. LVI is faster to execute but offers lower signal-to-noise ratio at high frequencies compared to SFRA. It is useful for field testing where speed is prioritized, though SFRA remains the reference method for detailed diagnostic analysis.

Quantitative metrics derived from FRA traces to automate interpretation:

• Relative Factor (RF): Ratio of compared trace areas
• Cross-Correlation Coefficient (CC): Statistical similarity score (0–1)
• Absolute Sum of Logarithmic Error (ASLE): Frequency-weighted deviation metric Values below defined thresholds trigger investigation. These indicators reduce reliance on expert visual interpretation but should never replace engineering judgment in final assessment.

The practice of triggering FRA testing specifically after a transformer experiences a high-magnitude external short circuit. Through-fault currents produce intense electromagnetic forces that can cause hoop buckling, axial telescoping, or spiral winding deformation. Comparing post-fault FRA to baseline reveals whether the mechanical structure withstood the stress. This event-driven testing strategy is a cornerstone of condition-based maintenance for power transformers.