Research on an improved VMD-RCMDE multi-feature fusion diagnosis method for external pump faults

Downloads

Authors

  • Liang Ge Southwest Petroleum University, China
  • Nvqing Zhou Southwest Petroleum University, China
  • Nengji Jiang Baikouquan Oil Production Plant, Xinjiang Oilfield Branch, PetroChina, China
  • Lie Li Baikouquan Oil Production Plant, Xinjiang Oilfield Branch, PetroChina, China
  • Xiaojun Wang Baikouquan Oil Production Plant, Xinjiang Oilfield Branch, PetroChina, China
  • Yi Teng Southwest Petroleum University, China

Abstract

This paper proposes a multi-feature fusion fault diagnosis method based on the global search whale optimization algorithm(GSWOA) to address the issues of single feature extraction, low detection efficiency, and insufficient accuracy in the actual operation of external pumps in oil and gas gathering and transmission. This method optimizes the variational mode decomposition(VMD) and the refined composite multi-scale dispersion entropy(RCMDE). First, GSWOA optimizes the VMD parameters to decompose the pressure signal and extract its intrinsic mode function(IMF) components during external pump operation. Second, to improve the accuracy and stability of feature extraction, RCMDE values are calculated for the pressure and temperature signals. Then, the extracted features are fused with the vibration signals to construct a multi-source feature fusion model. Finally, a classifier for fault identification and diagnosis of the constructed feature model is used. This classifier is a least squares support vector machine(LS-SVM) optimized by the snow geese algorithm(SGA). Experimental results demonstrate that the improved VMD-RCMDE multi-feature fusion diagnostic method achieves a 91% fault identification accuracy, which is at least 3.5% higher than other methods and significantly enhances the accuracy and robustness of external pump fault diagnosis. These results demonstrate the feasibility of efficiently and reliably diagnosing external pump faults and have significant theoretical and engineering application value.

Keywords:

External pump, Fault diagnosis, variational modal decomposition(VMD), refined composite multiscale dispersion entropy(RCMDE), least squares support vector machine(LS-SVM), Snow Goose algorithm(SGA)