Short-Circuited Turn Fault Diagnosis in Transformers by Using Vibration Signals, Statistical Time Features, and...

Area of Science:

• Electrical Engineering
• Mechanical Engineering
• Signal Processing

Background:

• Transformers are critical electrical components susceptible to various stresses, leading to failures.
• Short-circuited turns (SCTs) are a common transformer winding fault, often diagnosed using vibration signals.
• Diagnosing SCTs is challenging due to varying severity levels and high ambient noise.

Purpose of the Study:

• To develop and validate a robust methodology for diagnosing transformer short-circuited turns (SCTs).
• To utilize statistical time features (STFs) and support vector machines (SVM) for automated fault detection.
• To implement the diagnostic system on an FPGA for a practical, system-on-a-chip solution.

Main Methods:

• Computed 19 statistical time features (STFs) from transformer vibration signals.
• Selected discriminant features using Fisher score analysis and reduced dimensions with linear discriminant analysis.
• Employed a support vector machine (SVM) classifier for automated diagnosis.
• Implemented the methodology on a Field-Programmable Gate Array (FPGA).

Main Results:

• The proposed methodology effectively diagnosed transformer conditions under various SCT severities.
• Achieved a high diagnostic accuracy of 96.82%.
• Validated the effectiveness of the FPGA implementation for real-time application.

Conclusions:

• The STF and SVM-based methodology provides an accurate and effective approach for diagnosing SCTs in transformers.
• The FPGA implementation offers a viable system-on-a-chip solution for real-world transformer monitoring.
• This approach addresses the challenges of varying fault severity and high noise levels in vibration-based diagnosis.