The identification of gas–liquid two-phase flow patterns in a centrifugal pump with an imbalanced sample and variable rotational speeds
Su-Ting Li,Rui-Lin Li,Deng-Hui He,Rui Huang,Shuai-Hui Sun
DOI: https://doi.org/10.1063/5.0197688
IF: 4.6
2024-02-01
Physics of Fluids
Abstract:Due to the effects of the working medium and external conditions, centrifugal pumps often produce complex gas–liquid two-phase flows during operation. The entrainment of gas can lead to a significant decrease in the pressure head of the centrifugal pump and a serious reduction in conveyance efficiency. Therefore, in the absence of visualization or noninvasive techniques, the question of how to predict and model the flow status inside the pump is a key issue that needs to be urgently addressed. Using experimental data for gas–liquid two-phase flows in centrifugal pumps, this paper proposes a flow pattern identification model based on one-dimensional pressure increment signals. This deep learning model consists of convolutional neural networks and bidirectional gated recurrent units. Bayesian optimization is used for automatic hyperparameter tuning, and under the premise of balanced data distribution, the model achieves an identification rate of 0.98 on the test set. Subsequently, addressing the issue of the uneven sample distribution in centrifugal pumps due to their poor gas-carrying capacity, a data augmentation model based on the conditional Wasserstein generative adversarial network with gradient penalty approach is proposed based on the predictive model. Finally, to establish a flow pattern identification method with good extrapolation capability, a transfer learning strategy that accounts for the domain adaptation problem is introduced to the dataset with variable speeds based on the previous two studies. This method achieves knowledge transfer from the source domain to the target domain with a recognition rate of over 0.988 without hyperparameter optimization.
mechanics,physics, fluids & plasmas