When inspecting large oil-immersed transformers

the inspection robot faces the severe challenge of positioning itself inside the transformer due to the transformer’s large volume and metal enclosed structure. To address this issue

this paper proposes a three-dimensional ultrasound localization method. This method is based on a Grey Wolf Optimization-Variational Mode Decomposition (GWO- VMD) denoising method and a Minimum Variance Distortionless Response (MVDR) beamforming algorithm improved by an AutoRegressive Model (AR) model. This paper addresses the strong on-site noise interference by decomposing the ultrasonic signals emitted by the transformer robot using an improved GWO-VMD method. This method adaptively searches for the optimal center frequency and bandwidth of each mode. By extracting the mode of the positioning signal frequency band and suppressing the mode of the noise frequency band

the autonomously selected excellent Intrinsic Mode Function (IMF) is reconstructed to obtain a high signal-to-noise ratio signal. This paper adopts an AR model-improved MVDR beamforming algorithm to address the low spatial positioning resolution of the inspection robot using conventional beamforming methods. The proposed method uses a nine-element L-shaped sensor array to accurately position the internal inspection robot inside the transformer. Simulation and experimental results demonstrate that compared with traditional EMD denoising methods

the proposed GWO-VMD shows significantly superior denoising performance. The signal-to-noise ratio of the ultrasonic positioning signal increases from 4.2763 to 7.9012 dB

the wave shape similarity increases from 81.27% to 92.48%

and the root mean square error reduces from 0.1009 to 0.0609. In a transformer test box with three-dimensional sizes of 1 200 mm×1 000 mm×1 100 mm (length×width×height)

the GWO-VMD denoising algorithm and the AR-MVDR positioning method are sequentially used to achieve millimeter-level positioning

with the final spatial positioning coordinates of 490

392

394 mm. The relative positioning error of the inspection robot inside the transformer is less than 3%

and the maximum positioning error of less than 20 mm

meeting the engineering positioning requirements.