A novel method for recognizing acoustic signals of partial discharge optical fiber sensors using the time-frequency phase composition property is proposed in this paper. The method involves obtaining the Wigner–Ville time-frequency distribution for acoustic signals from partial discharge optical fiber sensors through the Cohen bilinear time-frequency transformation, which provides a high time-frequency resolution. The Wigner–Ville distribution could reflect the insulation defect-related properties in detail, owing to the fact that the intensity distribution in the time domain and energy distribution in the frequency domain is seriously influenced by medium dispersion and acoustic propagation. The time-frequency phase composition property is implemented by combining the Wigner–Ville distributions at different phases in the power cycle, which comprehensively represent the characteristics of the acoustic signals from partial discharge optical fiber sensors. A Vision Transformer with an attention block is introduced to identify the acoustic signals of partial discharge sensors. The attention block ensures that the neural network assigns more weight to the energy concentration areas in the extracted acoustic features. To validate the proposed approach, experiments are conducted to identify the acoustic signals of partial discharge optical fiber sensors. The proposed method achieves an impressive accuracy of 99.56% on three group testing sets. This indicates that the proposed approach is a promising method for identifying acoustic signals of partial discharge sensors to detect various insulation defects using acoustic emission feature analysis.
Jin et al. (Mon,) studied this question.