A sand particle characterization method for water-bearing high-production gas wells based on a multifrequency collision response

Excessive erosion caused by the continuous collision of sand-carrying annular flow with the gas well wellbore can lead to serious production accidents. This study combined the multifrequency response characteristics of sand particle-wall collision with a deep learning algorithm to improve the recognition accuracy of sand particle information in annular flow. The findings showed that sand-wall collision strength was closely related to the velocity, size, and number of sand particles and that the shielding effect generated by the collision behavior between multiple particles had a protective effect on the elbow. In addition, sand-wall collision strength increased with increases in gas velocity and particle size and decreased with an increase in liquid velocity. The shear effect, the secondary flow effect, and the liquid film buffering effect were shown to be key factors affecting the transportation behavior and spatial distribution of sand particles in annular flow. Furthermore, the fast Fourier transform (FFT) and short-time Fourier transform (STFT) analysis results showed that the multifrequency collision response characteristics of sand carrying annular flow were complex and that the main frequency response of sand-wall collision was concentrated in the high frequency range of 50–80 kHz. Moreover, the recognition accuracy results of convolutional neural network (CNN) models for particle size, gas velocity, and liquid velocity were 93.8%, 91.7%, and 91%, respectively, which were significantly higher than the results for the long short-term memory (LSTM) model. The combination of multifrequency collision response and deep learning effectively characterized sand particle feature information in strong gas-liquid turbulence, providing a reference for the accurate monitoring of sand particle information in high-yield water-bearing gas wells.