Statistical characterization of the flow structure of air-water-solid particles three-phase flow in the airlift pump-bubble generator system

Abstract

The purpose of the present study was to investigate the performance and flow structure of an airlift pump-bubble generator during the lifting of gas-water-solid particles three-phase flow experimentally. Here, a mechanistic model was also developed to predict the performance of an airlift pump bubble-generator on the basis of power balance. In the experimental part, the flow structure was analyzed visually, and taken from the extraction of the differential pressure signal at both the bottom and top test sections. Time series of differential pressure normalization was analyzed by using wavelet transform to determine the wavelet energy distribution. The wavelet energy was used as input the artificial neural network method to clustering the flow regime.The results indicate that under a constant of superficial gas velocity, the discharged both the water and particle increase with the submergence ratio (SR). SR is defined as the ratio of the distance from the injector to the water surface and the distance from the injector to the outlet side. Next, under a constant gas superficial velocity, the increase of SR will increase the solid fraction, but the fractions of both gas and water will decrease. The flow patterns were classified in the clustered bubble, homogeneous bubble, cap bubble, bubbly-stable slug, bubble unstable slug, and slug churn. Furthermore, the bubble flow is indicated by the peak wavelet energy on an eighth-level decomposition of approximation signal (a8), and the energy of the fourth-level decomposition of detail signal (d4) closes to the energy of the fifth-level decomposition (d5). The wavelet energy concentrated on the seventh and/or eighth level decomposition of detail signal (d7 and/or d8) is the indicator of slug flow in the riser pipe. The clustering flow patterns by using the ANN with input from wavelet energy gives a better approach than that of stochastic parameters of time series of the pressure differential. Moreover, the developed mechanistic model shows a good agreement with the experimental data.