Abstract:

To address the issue of low accuracy in identifying abnormal states of high-voltage direct current contactors and to enhance the generalization ability of the state identification method,a state identification method based on adaptive sensitive mode screening and a hybrid architecture model is proposed.Firstly,the collected vibration and acoustic signals are subjected to adaptive noise complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN),and the sensitive mode components are screened through the variance weighted strategy.Secondly,the reconstructed signals are fused and visualized using the symmetrized dot pattern (SDP) method.Finally,by integrating the efficient local feature extraction ability of EfficientNetV2 and the global modeling advantage of visual Transformer,an EfficientNetV2-ViT model is constructed by introducing the MobileViT module to solve the problem of image feature dispersion of SDP.Compared with other models,the proposed model has better computational efficiency and recognition accuracy.

Key words: high voltage DC contactor, state recognition, acoustic signal, symmetrized dot pattern, lightweight model