Forensic FRA: Differentiating Manufacturing Defects from Operational Winding Damage

When a transformer fails or shows abnormal FRA deviations, determining whether the root cause is a manufacturing defect or service-induced damage has significant warranty, insurance, and operational implications. A Transformer Frequency Response Analyzer, when interpreted with forensic rigor, provides evidence to distinguish between these origins. This article presents pattern recognition criteria that differentiate inherent manufacturing anomalies from acquired damage.

Manufacturing Defects: Characteristic FRA Signatures

Defects present at manufacture produce FRA patterns that are typically:

• Symmetrical and phase-consistent: A manufacturing error affecting all phases equally (e.g., incorrect number of turns on all HV windings) produces symmetrical deviations.

• Present from first test: The anomaly appears in the factory baseline itself, not as a change between baselines.

• Stable over time: Repeat FRA tests before and after transport or short-term operation show no progression.

• Correlated with design specifications: Deviation may match predictable effects of a known design error (e.g., different wire gauge than specified, incorrect spacer count).

Examples include:

• Consistently low correlation coefficient (0.80–0.90) between phases that persists across all test modes.

• Resonant frequencies that systematically differ from design calculations by >10%.

• Missing resonant peaks that design simulations predict should exist.

Service-Induced Damage: Characteristic FRA Signatures

Damage acquired during operation or transport produces patterns that are:

• Asymmetrical: One or two phases affected, reflecting the non-uniform nature of fault events or shipping shocks.

• Progressive: Correlation coefficient declines over successive tests as damage accumulates.

• Event-correlated: Deviation appears after a known through-fault, lightning strike, or transport event.

• Localized spectral effects: Specific frequency bands affected based on the type of event (mid-band for winding displacement, low-band for core movement, high-band for lead/bushing damage).

Case Example: Manufacturing Defect – Wrong Spacer Material

A utility received 10 identical 25 MVA transformers from a manufacturer. Commissioning FRA on the first unit showed all phases with CC = 0.97–0.98 between phases (acceptable). However, the resonant frequencies were consistently 15% higher than design calculations. The manufacturer claimed this was within tolerance. When the second unit showed identical upward shift, FRA comparison between units revealed they were nearly identical (CC > 0.96). This consistency across units indicated a systematic manufacturing issue, not random damage. Investigation found that the factory had substituted a higher-permittivity spacer material than specified, increasing winding capacitance and raising resonant frequencies. The manufacturer replaced all 10 transformers under warranty. Without FRA's ability to compare across identical units, the issue would have remained undetected.

Case Example: Service Damage – Through-Fault on Single Phase

A 50 MVA transformer experienced a single-phase-to-ground fault on Phase B of the HV side. Post-event FRA showed Phase B mid-band CC = 0.72 compared to baseline, while Phases A and C remained at 0.96 and 0.97. The asymmetry (one affected phase) and the event correlation (fault occurred) clearly indicated service-induced damage. The manufacturer's warranty claim was denied, and the utility's insurance covered the repair. The asymmetrical FRA pattern provided forensic evidence of the fault's role.

Time-Based Differentiation: Single Baseline vs. Multiple Baselines

The most powerful differentiator is having multiple FRA baselines:

• Factory baseline: Establishes as-manufactured condition.

• Post-installation baseline (before energization): Reveals transport damage.

• Post-commissioning baseline (after 30 days of operation): Reveals early service issues.

• Routine periodic baselines (every 3–5 years): Track progressive damage.

If a deviation appears between factory and post-installation baselines, the cause is transport or installation. If it appears between post-installation and post-commissioning, the cause is early service (e.g., through-fault during startup). If it appears gradually over multiple periodic tests, the cause is progressive wear.

Statistical Differentiation Using Fleet Data

For transformers without individual baselines, compare to fleet statistics:

• A transformer whose FRA signature falls outside the fleet mean by >3 standard deviations in a pattern consistent across all phases suggests a manufacturing batch issue.

• A transformer whose signature is within fleet norms but shows asymmetry between phases suggests service-induced damage.

Forensic Documentation for Legal and Insurance Purposes

When using FRA for root cause determination, ensure the following documentation:

• Complete chain of custody for all FRA data files

• Timestamps on every measurement (factory, site, post-event)

• Metadata including temperature, tap position, lead configuration, and technician identity

• Statistical index values (CC, SDR, ASLE) for each comparison

• Overlay plots with frequency axis scaled logarithmically and amplitude axis in dB

• Expert witness report interpreting the FRA evidence in the context of known failure modes

Limitations of FRA in Root Cause Analysis

FRA cannot differentiate all failure origins:

• It cannot determine the exact date of damage—only that damage exists between two test dates.

• It cannot distinguish between identical damage patterns caused by different events (e.g., shipping shock vs. through-fault may produce similar mid-band deviations).

• It requires a baseline; without a baseline, root cause attribution is speculative.

Combine FRA with event logs, DGA trending, and visual inspection for definitive root cause determination.

The Transformer Frequency Response Analyzer is a powerful forensic tool when deployed with systematic baseline collection and rigorous pattern analysis. By distinguishing manufacturing defects from service-induced damage, FRA protects utilities from unwarranted warranty denials and helps insurers correctly assign liability.