A TWO-STEP METHOD FOR DATA PROCESSING IN THE PROCESS OF DETECTING MICRO-DISCHARGE PULSES USING DEEP MACHINE LEARNING

Author(s): Nizamov D.Yu., Lazukin A.V., Zaitsev S.A.

Rubric: Information technology

DOI: 10.21777/2500-2112-2025-2-93-101

Release: 2025-2 (51)

Pages: 93-101

Keywords: machine learning, recurrent neural networks, time series classification, partial discharges, deep learning, event detection, mathematical model of current pulse

Annotation: The article describes an approach to solving the problem of detecting current pulses of a surface barrier discharge at atmospheric pressure under conditions of large volumes of data obtained at a high sampling rate. A brief overview of the collected dataset is presented, an analysis of a typical detection algorithm is carried out, and its shortcomings are identified. A new two-step approach to data processing is proposed: classification of the time series for data pre- processing and detection of the start and end of pulses using a deep learning regression model. The article describes relevant neural network architectures and provides a comparison of the effectiveness of the considered approaches in solving time series classification tasks. The use of machine learning makes it possible to reduce detection errors. This method shows better results than existing approaches due to improved accuracy and noise resistance. The theoretical and practical results of the study are aimed at the development of technical diagnostics and machine learning methods.

Bibliography: Nizamov D.Yu., Lazukin A.V., Zaitsev S.A. A TWO-STEP METHOD FOR DATA PROCESSING IN THE PROCESS OF DETECTING MICRO-DISCHARGE PULSES USING DEEP MACHINE LEARNING // Education Resources and Technologies. – 2025. – № 2 (51). – С. 93-101. doi: 10.21777/2500-2112-2025-2-93-101