Year: 2025
Type: article
Abstract: In order to comprehensively consider the time correlation of the signal and avoid converting the signal into the frequency domain, a higher classification success rate can be obtained. This paper proposes a decision-making model based on LBT-ODF space and multi-channel network for transformer voiceprint fault diagnosis. First, one-dimensional local features are extracted from the signal to acquire significant data of the temporal characteristics and a blended feature vector is derived as a result, then converting the signal from the time domain to a custom feature domain. Subsequently, a multi-channel one-dimensional convolutional neural network (1D-CNN) is used to process single feature and mixed feature channels respectively. At the same time, a multi-channel decision-making network is used to make majority decision-making judgments on the identification categories from each one-dimensional CNN model, and output the final fault category. The experimental outcomes demonstrate that this approach exhibits commendable recognition proficiency and resilience against noise. Consequently, the terminal model's identification precision escalates to a remarkable 96 %. This article provides a new idea for fault identification based on refined feature fusion measurement.
Authors: Jiaqi Peng, Yulin Ma, Haiping Ye, Xianrui Che, Shou Li, Bin Ai
Institution: State Grid Corporation of China (China)
Language: English
Subfield: Control and Systems Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $808
OpenAlex ID: W4407680537