DOI: 10.11799/ce202512021
An intelligent fault diagnosis method for shearer rocker arm gears based on SwinT-SKNet dual-branch fusion
YUE Dong
Abstract：To overcome the insufficient feature extraction in existing fault diagnosis models when handling nonlinear,non-stationary signals under high-noise in underground mines,which compromises diagnostic accuracy,a dual-branch fusion model based on SwinT and SKNet is proposed for fault diagnosis of shearer rocker arm gears.The model extracts both global and local features through two parallel branches:a simplified Swin Transformer branch and a lightweight SKNet branch.The SwinT branch incorporates a Bottleneck Attention Module(BAM)to enhance its global feature representation.The SKNet branch utilizes depthwise separable convolutions and replaces fully connected layers with 1-dimension convolutions to reduce computational complexity,making the model suitable for deployment on mobile and edge devices.Tests conducted on the Taiyuan Heavy Industry rocker arm loading test bench show that the proposed method outperforms comparative models in both accuracy and precision.It maintains over 97％recognition accuracy even at a signal-to-noise ratio(SNR)of-6dB,demonstrating strong noise robustness.
Keywords：shearer rocker arm gearfault diagnosisSwin TransformerSelective Kernel Network
Publication Date：2025-12-20
Online Publishing Date：2026-01-29(First online date of this platform, not the publication date of the document)
Pages：9( 162-170 )