Power transformer fault diagnosis method based on multi source signal fusion and fast spectral correlation DOI Creative Commons
Shan Guan, Mingyu Shi, Fuwang Wang

et al.

Scientific Reports, Journal Year: 2025, Volume and Issue: 15(1)

Published: Feb. 27, 2025

Addressing the issues that signal measured by a single sensor can not provide complete description of power transformer fault states and problems selection features relies on manual experience, method based multi source fusion Fast Spectral Correlation is produced for diagnosis. At first, vibration signals from different locations surface case are collected array synchronously, Function Weighting proposed to fuse multi-source multiple sensors in order obtain fused signal; then, subjected belonging cyclic smooth theory construct sample set images; finally, image samples fed into MobileNetV3 model training transfer learning fine-tuned neural network model, which completes Experimental results showed overall recognition accuracy reached 98.75%, was 10.52% higher than diagnosis signal, 10.86% other classical images, providing new tool signals.

Language: Английский

Power transformer fault diagnosis method based on multi source signal fusion and fast spectral correlation DOI Creative Commons
Shan Guan, Mingyu Shi, Fuwang Wang

et al.

Scientific Reports, Journal Year: 2025, Volume and Issue: 15(1)

Published: Feb. 27, 2025

Addressing the issues that signal measured by a single sensor can not provide complete description of power transformer fault states and problems selection features relies on manual experience, method based multi source fusion Fast Spectral Correlation is produced for diagnosis. At first, vibration signals from different locations surface case are collected array synchronously, Function Weighting proposed to fuse multi-source multiple sensors in order obtain fused signal; then, subjected belonging cyclic smooth theory construct sample set images; finally, image samples fed into MobileNetV3 model training transfer learning fine-tuned neural network model, which completes Experimental results showed overall recognition accuracy reached 98.75%, was 10.52% higher than diagnosis signal, 10.86% other classical images, providing new tool signals.

Language: Английский

Citations

0