Rotor fault diagnosis of rotating machinery based on SDP images and VGG network

WU Hai-bin, BU Ming-long, LIU Yuan-yuan, HAO Hui-min
(School of Mechanical and Vehicle Engineering, Taiyuan University of Technology, Taiyuan 030024, China)

Abstract: Aiming at improving the identification accuracy of rotating machine rotor fault diagnosis, a deep learning visual geometry group (VGG) network method combined with symmetry dot pattern (SDP) image feature was proposed. The multichannel vibration signals of rotor in fault state were transformed by SDP, and the SDP images of different rotor fault shown distinguish features. The SDP images were input the deep learning VGG network adaptively, and the fault features were extracted by VGG. The results show that the SDP images combined with VGG obtained more accurate diagnosis and fault recognition of rotor fault than extreme learning machine (ELM). The results indicate that the method for fault diagnosis of rotating machinery rotors based on SDP images and VGG networks solves the problems of high complexity, nonlinearity and instability in vibration signals of rotor faults, and has higher recognition accuracy than traditional machine learning methods ELM.
Key words: deep learning; visual geometry group (VGG) network; symmetry dot pattern (SDP) images; multi-channel information fusion; rotor fault diagnosis；extreme learning machine (ELM)