Vibration detection method for core loosening degree of high-voltage direct current converter valve saturable reactor based on variational mode decomposition-symmetrized dot pattern and image matching

Abstract

Saturable reactor is the key equipment in converter valves; it is important to evaluate the core condition. In this paper, a recognition model of saturable reactor core loosening based on variational mode decomposition (VMD)-symmetrized dot pattern (SDP) feature fusion images and scale-invariant feature transform (SIFT) was proposed. Firstly, the saturable reactor vibration test under high frequency pulse excitation was carried out, and the vibration signals in different core loosening degrees were collected. Secondly, the VMD algorithm was used to decompose the broadband vibration signal into multiple narrowband functions, which were used to reflect the characteristics of each frequency band. Thirdly, each function and the original signal were transformed by SDP, the generated spiral arms were fused into a new image, and the typical templates in different loosening degrees were selected. Finally, the improved SIFT algorithm was used to obtain the matching results between test sets and templates. The results show that the recognition accuracy of the proposed model for core loosening is 97.5%, which is better than traditional algorithms. It can find the core loosening defect early and avoid further failures such as water pipe break and discharge, which can provide an important basis for saturable reactor monitoring.