A power transformer survives a nearby short-circuit fault. It rides through the event, the protection clears the fault, and the transformer goes back to service. Oil tests come back normal. Insulation resistance is healthy. DGA shows no concerning gas evolution. Everything looks fine.
But inside the tank, the windings have shifted. Not enough to fail immediately. Not enough to show on any standard electrical test. Just enough that next time a similar fault occurs, the weakened windings will fail catastrophically.
This is the problem Frequency Response Analysis (FRA) was developed to solve. FRA detects mechanical changes inside transformers that other tests miss — winding deformation, axial displacement, core movement, loose connections. It’s become a standard part of transformer diagnostics for utilities, large industrial operators, and asset managers worldwide.
This article covers FRA from physical principles to field interpretation. Standards referenced include IEC 60076-18 (the FRA-specific transformer standard), IEEE C57.149 (the IEEE FRA guide), and CIGRE working group recommendations.
Table of Contents
What FRA Detects (and Why Other Tests Can’t)
To understand FRA’s value, you need to understand what it sees that other tests don’t.
The mechanical damage problem
Power transformers experience mechanical stress throughout their service life:
- Through-fault currents — when a fault occurs downstream of the transformer, the resulting short-circuit current passes through the windings. The electromagnetic forces (proportional to current squared) push windings axially and radially. Severe through-faults can permanently displace conductors, loosen winding clamping, or deform the geometric structure.
- Transportation events — moving a transformer (especially the heavy core-and-coil assembly) creates mechanical stress that can shift internal components.
- Seismic events — earthquakes apply lateral acceleration that can move windings or core sheets.
- Internal faults — even minor internal arcing creates pressure waves and electromagnetic forces that can damage internal mechanical structures.
These mechanical changes don’t necessarily cause immediate failure. Per Vahidi & Teymouri:
In most cases, the mechanical displacement or mechanical deformation of the windings does not prevent the transmission of energy in the system, but there is a risk that the mechanical damage to the insulation may lead to an insulation failure ultimately.
Mechanical damage progressively weakens the transformer’s ability to withstand future stresses. A transformer with shifted windings has less mechanical clearance for the next fault, less insulation margin between conductors, and reduced ability to handle thermal cycling. Eventually, this leads to insulation failure — sometimes years after the original mechanical event.
Why other tests miss it
Standard transformer tests don’t see mechanical changes:
- Insulation resistance measures bulk insulation. A 1 mm shift in winding position doesn’t change bulk insulation enough to detect.
- Turns ratio detects shorts between turns but not geometric changes.
- DGA detects active fault byproducts but not past mechanical events.
- Tan delta / power factor measures dielectric losses but is insensitive to small mechanical shifts.
- Impedance measurements detect major electrical changes but require significant winding movement to register.
Per Vahidi & Teymouri:
It is very difficult to detect the deformation of power transformer windings using the traditional methods such as turns ratio measurements, impedance and inductance measurements. However, the deformation of the windings can cause minor changes in the values of the inductance and the capacitance of the windings.
FRA fills this gap by measuring those minor inductance and capacitance changes through their effect on the transformer’s frequency response.
The Physics: Transfer Functions and Transformer Windings
FRA is fundamentally a transfer function measurement.
What a transfer function is
A transfer function describes how a linear system responds to inputs across a range of frequencies. Apply a known signal at the input, measure the resulting output, and the ratio at each frequency defines the transfer function.
For a transformer, the transfer function captures how the winding network responds to electrical signals. Per Vahidi & Teymouri:
The basis of the transfer function method is the bipolar network theory. In this model, the transformer is assumed to be a linear, coherent and passive network.
The basic transfer function definitions:
Voltage transfer function:
H_V(f) = U_output(f) / U_input(f)
Current transfer function:
H_I(f) = I_output(f) / U_input(f)
Where U(f) and I(f) are the frequency-domain (FFT) representations of voltage and current signals.
Why transformers have rich frequency responses
A transformer winding isn’t a simple component — it’s a complex distributed network of:
- Inductances — self-inductances of each winding section, mutual inductances between sections
- Capacitances — between turns, between layers, between windings, to ground
- Resistances — copper resistance, iron losses
This network has multiple natural resonant frequencies. The frequency response shows characteristic peaks and valleys at these resonances, with patterns specific to each transformer’s geometry.
Crucially, the frequency response depends on the physical geometry of the winding. If a winding shifts, deforms, or develops a short, the inductances and capacitances change — and the resonant frequencies shift. FRA detects these shifts by comparing the measured frequency response against a reference.
The fingerprint concept
Each transformer has a unique frequency response signature — a “fingerprint” based on its specific geometry. This fingerprint:
- Is stable over time for a healthy, undamaged transformer
- Changes when geometry changes (winding shift, deformation, internal damage)
- Is specific to each transformer (even identical units have slightly different fingerprints)
- Is repeatable under controlled conditions
The diagnostic value comes from comparing today’s fingerprint to a reference. Significant deviations indicate a change has occurred between tests.
The Two Measurement Methods: SFM and LVI
Per IEC 60076-18 and as detailed in Vahidi & Teymouri Section 5.18, two methods produce equivalent results in different ways.
Swept Frequency Method (SFM)
The dominant method in modern FRA testing. A sinusoidal voltage of known amplitude is applied to the winding, swept across a wide frequency range — typically 20 Hz to 2 MHz. At each frequency, the output is measured.
Mathematical representation:
Magnitude: K = 20 × log₁₀(T/R) [dB]
Phase: φ = arctan(T/R) [degrees]
Where T = output signal and R = input (reference) signal.
The result is plotted as magnitude (dB) and phase (degrees) versus frequency on a log scale, producing the characteristic FRA “fingerprint” curve.
SFM advantages (per Vahidi & Teymouri):
- High signal-to-noise ratio with effective white noise filtering
- Wide frequency range scan
- Good frequency accuracy at low frequencies
- Repeatable, well-suited for trending
SFM disadvantages:
- Only one transfer function measured per test
- Longer measurement time (typically 5–15 minutes per measurement)
This is the method virtually all commercial FRA equipment uses today (OMICRON, Megger, Doble, and others).
Low Voltage Impulse Method (LVI)
Older method, less common today but still used in some applications. An impulse voltage is applied to the winding, the response is captured in time domain, then converted to frequency domain via Fourier Transform.
Per Vahidi & Teymouri:
In the LVI method, the impulse domain is usually between 100 and 2000 V and the front time is in the range of 200 ns to 1 μs and the half-value time is in the range of 40–200 μs.
LVI advantages:
- Multiple transfer functions (admittance, impedance, voltage) can be measured simultaneously
- Each measurement takes only a few minutes
LVI disadvantages (per Vahidi & Teymouri):
- Fixed frequency accuracy — fault detection at low frequencies is difficult
- White noise filtering is challenging
- Requires multiple measurement equipment
Which method to use
For modern field testing per IEC 60076-18 and IEEE C57.149, SFM is the standard. The combination of better noise rejection, wider frequency coverage, and repeatability makes it the practical choice. LVI persists in some research and specialized industrial applications but is rarely seen in routine utility FRA testing.
The two methods produce equivalent diagnostic information when applied correctly — but SFM is more practical for trending and field comparison.
The Four Reference Comparisons
A single FRA measurement is meaningless. The diagnostic value comes from comparing the current measurement against a reference. Four types of references exist, in decreasing order of reliability.
1. Time-based comparison (best)
Compare today’s measurement to the same transformer’s previous FRA measurement (commissioning baseline or previous routine test).
Strengths:
- Same transformer, same physical configuration
- Eliminates manufacturing differences
- Sensitive to small changes
- Most reliable diagnostic comparison
Limitations:
- Requires historical data (commissioning baseline is essential)
- Lost reference data is unrecoverable
This is why establishing FRA baselines at commissioning has become standard practice for new transformers above ~10 MVA.
2. Type-based comparison
Compare to another transformer of the same design (same manufacturer, same vintage, same specification).
Strengths:
- Doesn’t require historical data on the specific unit
- Works for legacy transformers without baselines
- Useful when comparing fleet units
Limitations:
- Manufacturing variations between “identical” units
- Less sensitive than time-based comparison
- Requires access to other matching units
3. Phase-to-phase comparison
Compare the responses of phase A, phase B, and phase C of the same transformer. In healthy three-phase transformers, the three phases should produce nearly identical responses.
Strengths:
- Available on every three-phase transformer
- No external reference data needed
- Effective for detecting single-phase damage
Limitations:
- Insensitive to symmetric damage (affecting all three phases equally)
- Some asymmetry is normal and design-dependent
4. Sister-unit comparison (least reliable)
Compare to a transformer at a different site nominally of the same design.
Strengths:
- Available when nothing else is
- Better than no comparison
Limitations:
- Manufacturing differences between sites
- Different installation history
- Significantly less sensitive than other comparisons
The hierarchy in practice
For a critical transformer, the diagnostic hierarchy is:
- Measure FRA today
- Compare to commissioning baseline → primary diagnostic
- Compare phase-to-phase → secondary check
- Compare to similar units → tertiary check (especially after a phase asymmetry is detected)
If only one comparison is available, time-based is most valuable. If only phase-to-phase is available, asymmetric damage is detectable but symmetric damage may be missed.
Frequency Band Interpretation
Different frequency bands correspond to different physical phenomena in the transformer. Understanding which band is affected helps localize the problem.
Low frequency: < 2 kHz
Dominated by: Magnetizing inductance and the iron core
What changes here:
- Core damage (broken laminations, displaced sheets)
- Residual magnetization changes
- Core ground connection problems
Changes below 2 kHz typically indicate core-related issues.
Mid frequency: 2 kHz – 20 kHz
Dominated by: Bulk inductive properties of the windings
What changes here:
- Bulk winding deformation
- Major axial displacement
- Significant clamping problems
Changes in this band are typically diagnostic of significant mechanical issues affecting the entire winding.
High frequency: 20 kHz – 1 MHz
Dominated by: Distributed inductance and capacitance within windings
What changes here:
- Local winding deformation
- Disc-to-disc capacitance changes
- Inter-winding capacitance changes
- Localized turn-to-turn problems
This band has the highest diagnostic value for detecting partial damage or localized changes.
Very high frequency: > 1 MHz
Dominated by: Lead arrangement and bushing connections
What changes here:
- Lead movement or breakage
- Bushing connection problems
- External tap changer connections
Above 1 MHz, the response is dominated by what’s outside the main winding structure.
Reading FRA results
A trained interpreter reads the FRA trace by:
- Observing where in the frequency spectrum changes occur
- Identifying which physical phenomena that band corresponds to
- Correlating with known transformer history (recent through-faults, transportation, age)
- Combining with other diagnostics (DGA, tan delta, oil tests) for confirmation
This is why FRA interpretation requires expertise — the patterns alone don’t tell the story without context.
What FRA Can Detect — The Defect Catalog
Per Vahidi & Teymouri Table 5.1, FRA detection capability for various defects:
| Defect | FRA Detection |
|---|---|
| Winding axial displacement | Easily |
| Winding holder loosening | Easily |
| Winding deformation | Easily |
| Core sheet deformation | Easily |
| Winding loop short circuit | Easily |
| Poor tank-to-ground connection | Easily |
| Closed loops in tank (circulating currents) | Detectable |
| Weak connections | Easily |
| Extra loops in windings | Easily |
| Multiple grounding points in the core | Easily |
This is what makes FRA distinctively valuable. Most of these defects are difficult or impossible to detect with other tests.
Common scenarios where FRA finds problems
Post-fault diagnostics: A through-fault has occurred. Tank pressure was normal. DGA shows no concerning gases. But did the windings shift? FRA comparison against the pre-fault baseline answers this question definitively. If the response has changed, mechanical damage occurred — even if no other test shows it.
Post-transportation testing: A transformer has been moved (factory delivery, replacement, relocation). FRA testing before shipment and after installation confirms the transportation didn’t cause internal mechanical damage. This is now standard practice for all transformers above ~10 MVA.
Aging assessment: A 30-year-old transformer is being evaluated for life extension. Multiple FRA measurements over its life show the response has remained stable — strong indicator of mechanical integrity. Conversely, a noticeable drift suggests the transformer is approaching end of mechanical life.
Acceptance testing: A used transformer is being purchased for re-installation. FRA testing as part of due diligence reveals whether the unit’s mechanical structure is intact, complementing routine electrical and dielectric tests.
Differential diagnosis: The transformer is showing elevated DGA gases but no clear cause. FRA testing identifies whether mechanical damage is contributing — if FRA matches baseline, the gas evolution is from a different cause; if FRA shows changes, mechanical damage is implicated.
What FRA Cannot Detect
Per Vahidi & Teymouri Table 5.1, two specific defects fall outside FRA capability:
| Defect | FRA Detection |
|---|---|
| No connection between core and ground | Indiscoverable |
| Foreign particles in tank | Indiscoverable |
Floating core ground
If the core ground connection is completely lost, FRA may not detect this directly. The floating core can produce its own issues (induced voltages, spurious discharges) but the FRA fingerprint may remain unchanged. Other tests — particularly insulation resistance and partial discharge measurements — are better for this.
Foreign particles
Conductive particles loose inside the tank don’t change the inductive or capacitive structure of the windings. They may cause partial discharges, contaminate oil, or even cause arcing in extreme cases — but they don’t shift the FRA fingerprint. Particle detection requires DGA or PD testing, not FRA.
Other limitations
- Active electrical faults — FRA is performed on de-energized transformers, so active arcing or discharge isn’t directly measured
- Cell-level chemical changes — paper degradation, oil oxidation, moisture accumulation don’t show in FRA
- Loose connections at terminals — external connections affect FRA but only at very high frequencies; detail-level diagnosis is harder
- Tap changer status — FRA is sensitive to tap position; the tap must be at the same position as the reference for valid comparison
These limitations don’t reduce FRA’s value — they just clarify that FRA is one tool in a broader diagnostic toolkit, not a complete diagnostic by itself.
Test Setup and Connections
Standard FRA per IEC 60076-18 measures specific transfer functions across the transformer’s terminals.
Test arrangements
End-to-end open circuit: Apply signal to one end of a winding, measure response at the other end with all other terminals open. Sensitive to the winding’s series inductance and turn-to-turn capacitance.
End-to-end short circuit: Apply signal to one end of a winding, measure at the other end with the secondary winding shorted. Sensitive to leakage inductance and capacitive coupling between primary and secondary.
Capacitive inter-winding: Apply signal to one winding, measure response on the other winding. Sensitive to inter-winding capacitance.
Inductive inter-winding: Similar but with one winding shorted. Sensitive to magnetic coupling between windings.
The four arrangements together provide diagnostic visibility into different aspects of the transformer’s structure.
Connection requirements
Per IEC 60076-18 best practices:
- Disconnect from system — both HV and LV sides, including ground connections to neutral
- Tap position — record exactly; subsequent measurements must be at the same tap
- Bushing connections — clean, tight, properly torqued for repeatability
- Test cables — short, high-quality coaxial cables with consistent grounding
- Ground reference — single, low-impedance ground point at the test instrument
- Measurement conditions — record temperature, oil level, humidity (these can subtly affect results)
Repeatability requires consistency across measurements. Small differences in connection arrangement, tap position, or grounding can cause apparent changes in the FRA trace that aren’t related to actual transformer damage.
Modern FRA equipment
A modern field FRA system includes:
- Sinusoidal signal generator (sweep capability across the full frequency range)
- High-impedance voltage measurement at input and output
- Digital signal processing for FFT analysis
- Software for trace comparison, pattern recognition, and reporting
- USB or network interface for data management
Major vendors include OMICRON (the FRAnalyzer device referenced in the Vahidi textbook), Megger, Doble, and Megger MIT-series equipment. The principles are the same; software and analysis tools differ.
A typical field-portable FRA system runs $25,000–$75,000. This is much less than offline PD testing equipment, making FRA economically accessible to most utilities and large industrial operators.
Standards: IEC 60076-18 and IEEE C57.149
Two main standards govern FRA testing:
IEC 60076-18:2012
Power transformers — Part 18: Measurement of frequency response
The international standard for FRA testing. Specifies:
- Measurement procedures and connections
- Test equipment requirements
- Reporting requirements
- Recommended frequency ranges (typically 20 Hz to 2 MHz)
- Quality assurance for measurements
This is the dominant international standard for transformer FRA. Most commercial FRA equipment is designed to comply with IEC 60076-18 procedures.
IEEE C57.149:2012 (Reaffirmed 2020)
IEEE Guide for the Application and Interpretation of Frequency Response Analysis for Oil-Immersed Transformers
The IEEE counterpart, focused more on application and interpretation than on measurement procedures. Includes:
- Detailed guidance on test connections
- Pattern interpretation methods
- Numerical comparison criteria (correlation coefficients, standard deviations)
- Case studies of various defect types
The standards are largely compatible. Some utilities reference both, taking the IEC for procedural compliance and IEEE for interpretation guidance.
CIGRE recommendations
CIGRE Working Group A2.26 produced “Mechanical-condition assessment of transformer windings using FRA” (Technical Brochure 342, 2008), which provides additional industry consensus on FRA application. While not a formal standard, it’s widely cited in technical practice.
Numerical comparison approaches
Both IEC 60076-18 and IEEE C57.149 describe numerical methods for FRA comparison:
- Correlation coefficient (CC) — measures similarity between traces; near 1.0 indicates good match
- Standard deviation of differences — quantifies overall dissimilarity
- Frequency band metrics — separate metrics for low, mid, and high frequency bands
Typical interpretation thresholds (from IEEE C57.149 and industry practice):
| Correlation coefficient | Interpretation |
|---|---|
| > 0.9999 | Excellent match — no significant change |
| 0.999 – 0.9999 | Good match — minor changes (often measurement noise) |
| 0.99 – 0.999 | Notable changes — investigate |
| < 0.99 | Significant changes — high probability of mechanical damage |
These thresholds vary by frequency band. The high-frequency band (20 kHz – 1 MHz) typically shows more variation than the low-frequency band, so thresholds are often relaxed there.
Numerical comparison reduces interpretation subjectivity but doesn’t replace expert judgment. A transformer with CC = 0.9995 might be perfectly healthy or might have a localized defect that’s mathematically subtle but mechanically real. Interpretation requires both the numbers and the patterns.
Common Pitfalls in FRA Testing
Several pitfalls can compromise FRA test results.
Tap position mismatch
The on-load tap changer position significantly affects FRA results. A test at tap 5 cannot be directly compared to a baseline at tap 7. Always document the tap position at the time of measurement, and ensure subsequent tests use the same tap position for time-based comparison.
Connection inconsistency
Different bushing terminals, different cable lengths, different grounding points all produce different traces. Establish a consistent connection protocol at commissioning and follow it exactly for all subsequent measurements.
Temperature effects
Oil temperature affects winding capacitances slightly. Significant temperature differences (>20°C) between baseline and current measurement can produce small apparent shifts that aren’t related to mechanical damage. Document and consider temperature in interpretation.
Bushing or external lead changes
If a transformer has had bushings replaced, leads modified, or external connections changed, the FRA trace will shift. The change isn’t from mechanical damage to the windings — it’s from the modified external structure. Establish a new baseline after such modifications.
Tap changer cycling between tests
Switching the on-load tap changer between FRA tests can introduce noise. Some operators recommend cycling the OLTC through several operations before the FRA test to ensure consistent contact engagement.
Equipment differences
Different FRA instruments produce slightly different traces. If switching equipment vendors, expect some differences in absolute values. The pattern shape and frequency-dependent features should remain consistent if the transformer hasn’t changed.
Over-interpretation of small differences
Not every difference is significant. Random measurement noise, small temperature variations, and tap changer contact differences can produce small apparent changes. Only changes outside the noise floor warrant investigation.
Under-interpretation of subtle patterns
Conversely, expert interpreters notice patterns that less experienced engineers miss. Subtle shifts in resonant frequency, small changes in damping, or low-amplitude features can indicate developing problems.
This balance — distinguishing real changes from noise without missing subtle but significant patterns — is what makes FRA interpretation a developed skill rather than a simple measurement.
FRA in the Broader Diagnostic Program
FRA is most valuable when combined with other transformer diagnostics.
The transformer diagnostic pyramid
A modern transformer diagnostic program typically includes:
| Test | Frequency | Detects |
|---|---|---|
| Visual inspection | Continuous | Obvious external damage, oil leaks |
| DGA | Quarterly to annually | Active fault byproducts, paper degradation |
| Oil quality tests | Annually | Oil aging, moisture, acidity |
| Insulation resistance | Major outages | Bulk insulation health |
| Tan delta / power factor | Major outages | Dielectric losses, contamination |
| Turns ratio | Major outages | Winding shorts |
| FRA | Major outages or after events | Mechanical changes, winding deformation |
| Partial discharge | Critical units, occasional | Localized defects, void discharges |
Each test sees a different aspect of transformer health. FRA’s specific niche — mechanical integrity — complements the others.
When to add FRA
FRA testing makes economic sense when:
- The transformer is large enough that failure consequences justify the testing cost (typically >10 MVA)
- The transformer has experienced a through-fault (verify mechanical integrity post-event)
- The transformer has been transported (verify no transportation damage)
- Routine diagnostics show ambiguous results (FRA can confirm or rule out mechanical contributions)
- Asset management requires comprehensive health assessment
- Insurance or regulatory requirements specify FRA as part of acceptance or maintenance
For small distribution transformers below ~5 MVA, FRA is rarely justified — the cost of testing (typically $3,000–$10,000 per unit including outage) approaches the unit’s replacement cost.
Integration with other diagnostics
A comprehensive transformer health assessment uses multiple diagnostic methods together:
Example: Through-fault diagnostic
A 50 MVA transformer experiences a major through-fault. Diagnostic sequence:
- DGA within 24 hours — check for fault gases (acetylene = arcing, hydrogen = PD)
- Oil quality — verify oil hasn’t been compromised
- Visual inspection — check for oil leaks, tank deformation
- FRA — compare to pre-event baseline; identify any winding movement
- Insulation resistance / tan delta — verify insulation hasn’t been damaged
- Decision — return to service, monitor more frequently, or schedule outage for further investigation
If FRA shows no changes, the transformer can return to service with confidence. If FRA shows shifts, further investigation (possibly tank inspection) is warranted before re-energization.
This integrated approach is the standard for modern utility transformer diagnostics.
FAQ
How sensitive is FRA to small mechanical changes?
Modern FRA equipment can detect winding axial displacements of a few millimeters and turn-to-turn capacitance changes of a few percent. The sensitivity depends on the frequency band — high-frequency response is most sensitive to localized changes; low-frequency response shows bulk changes. Per Vahidi & Teymouri, FRA “calculates the values of the frequency dependent variables of transformer windings such as inductance and capacitance of the coils. These parameters change when the windings are short-circuited, open circuited or deformed.”
Does FRA require taking the transformer out of service?
Yes. FRA is an offline test — the transformer must be de-energized, isolated, and connected to the test equipment. Typical outage requirement is 4–8 hours including setup, multiple measurements, and teardown.
Can I do FRA on three-phase transformers without baseline data?
Yes, using phase-to-phase comparison. Without a baseline, you can still detect asymmetric damage by comparing the three phases. Symmetric damage (affecting all three equally) is harder to detect without external reference. Per Vahidi & Teymouri, “by dividing the voltage or current of an arbitrary terminal over another desired voltage or current, a transfer function or conversion function is obtained.”
How does FRA compare to short-circuit impedance measurement?
Both detect winding deformation but at different sensitivities. Short-circuit impedance changes only after significant mechanical damage. FRA detects much smaller changes earlier. The two tests complement each other — FRA for sensitivity, impedance for absolute reference.
Can FRA detect problems caused by lightning or switching transients?
Possibly. Severe lightning or switching surges can deform windings or damage insulation. FRA detects the resulting mechanical changes. Smaller transients that don’t cause physical damage won’t show in FRA.
How long does an FRA test take in the field?
Typically 2–4 hours of actual measurement once setup is complete. Including transformer isolation, lockout/tagout, setup, multiple measurements (different windings, different connections), and reconnection, expect 4–8 hours total outage.
Is FRA suitable for distribution transformers?
For pad-mount distribution transformers (≤5 MVA), FRA is rarely cost-justified. For substation distribution transformers (~10 MVA and above) and power transformers, FRA is increasingly standard practice. The break-even point depends on transformer cost vs. testing cost.
Can FRA be done online?
No, in the conventional sense. Online FRA requires special instrumentation that injects test signals while the transformer is energized, which presents safety and operational challenges. Some research has explored online FRA, but it’s not yet standardized or widely deployed.
Do I need an expert to interpret FRA results?
For routine acceptance testing where the transformer is clearly healthy or clearly damaged, automated comparison with standard correlation coefficients can give reliable answers. For ambiguous results, complex patterns, or critical transformers, expert interpretation is valuable. Most large utilities have FRA specialists on staff or contract with diagnostic service providers.
What if I don’t have a baseline FRA for an existing transformer?
Several options:
- Start now — establish a baseline today; future measurements compare against this
- Use type-based comparison — test similar transformers and compare
- Use phase-to-phase comparison — works on the existing transformer alone
- Reference manufacturer’s type test data if available
Without a baseline, you have less diagnostic capability, but FRA still provides value through phase-to-phase analysis.
How often should I do FRA?
For most large transformers:
- At commissioning — establish baseline
- After significant events (through-fault, transportation, major maintenance) — verify integrity
- Every 5–10 years — routine assessment
- As part of major outage diagnostic programs — combine with insulation tests and other diagnostics
Some utilities perform FRA every 3–5 years as routine; others wait for triggering events. Both approaches are reasonable depending on equipment criticality.
Is FRA recognized for warranty or insurance purposes?
Increasingly yes. Manufacturer warranty claims after through-faults often require FRA evidence to determine if winding movement was within acceptable tolerances. Insurance assessors for major losses request FRA data as part of post-event analysis. Documented FRA history adds value to transformer asset management.
Key Takeaways
- FRA detects mechanical changes in transformer windings that other electrical tests miss. This includes winding deformation, axial displacement, core movement, and clamping problems.
- The transfer function method measures how the transformer responds to a swept frequency input. Today’s measurement compared to a reference reveals geometric changes in the windings.
- Two measurement methods exist: Swept Frequency Method (SFM, dominant today) and Low Voltage Impulse (LVI, less common). Both produce equivalent diagnostic information when applied correctly.
- Four reference comparisons are possible: time-based (most reliable), type-based, phase-to-phase, and sister-unit. Time-based comparison against a commissioning baseline is the gold standard.
- Frequency bands have meaning: below 2 kHz indicates core issues; 2–20 kHz indicates bulk winding problems; 20 kHz – 1 MHz indicates localized winding changes; above 1 MHz indicates lead and bushing issues.
- FRA detects easily: winding axial displacement, winding deformation, core sheet deformation, winding loop short circuits, weak connections, multiple grounding points, poor tank-to-ground connections.
- FRA cannot detect: floating core ground, foreign particles in the tank, active electrical faults, chemical changes in oil or paper.
- Standards: IEC 60076-18 (international), IEEE C57.149 (US), CIGRE TB 342 (industry consensus).
- Pitfalls to avoid: tap position mismatches, connection inconsistencies, temperature effects, bushing or lead modifications, over-interpretation of small differences.
- FRA fits in a broader program alongside DGA, oil tests, insulation resistance, tan delta, and partial discharge. Each test sees a different aspect of transformer health.
- Best applied to: transformers ≥10 MVA, after through-faults, after transportation, during major outage diagnostics, for aging assessment.
Standards and References
| Standard / Reference | Content |
|---|---|
| IEC 60076-18:2012 | Power transformers — Part 18: Measurement of frequency response |
| IEEE C57.149-2012 (R2020) | IEEE Guide for the Application and Interpretation of Frequency Response Analysis for Oil-Immersed Transformers |
| CIGRE Technical Brochure 342 | Mechanical-condition assessment of transformer windings using FRA (Working Group A2.26, 2008) |
| IEC 60076-3:2018 | Power transformers — Insulation levels, dielectric tests, and external clearances in air |
| IEEE C57.152-2013 | IEEE Guide for Diagnostic Field Testing of Fluid-Filled Power Transformers, Regulators, and Reactors |
| Vahidi & Teymouri (2019) | Quality Confirmation Tests for Power Transformer Insulation Systems (Springer) — Chapter 5, Sections 5.14–5.20 |