Abstract:

This paper proposed and validated a prediction model for interfacial wave velocity in annular flow within vertical pipes based on ensemble learning. By optimizing the parameters of empirical wavelet transform (EWT) using whale optimization algorithm (WOA), the signal processing accuracy was enhanced. Key features of the interfacial wave velocity signal were extracted from both time and frequency domains based on theoretical models. Further, models such as support vector regression, decision tree regression, random forest regression, least squares boosting (LSBoost) regression, and Bagging regression were combined to develop a high-precision wave velocity prediction model, with gas-liquid interfacial wave velocity as the prediction target under varying flow rates and pressure conditions. The analysis results showed that the proposed model effectively captured the complex flow characteristics in gas-liquid two-phase flow, with the LSBoost regression achieving a root mean square error (RMSE) of 3.96×10^-7, indicating excellent stability and predictive accuracy. The model provided effective support for the analysis of gas-liquid two-phase flow behavior and engineering practice, especially under complex operating conditions, where the model could still predict wave velocity stably. The model hold broad application prospects and potential, with significant value in practical engineering applications.

Key words: fluid mechanics, annular flow, interface, ensemble learning, prediction

摘要：

提出并验证了一种基于集成学习的垂直管环状流界面波速预测模型。通过鲸鱼优化算法（WOA）优化经验小波变换（EWT）的参数，以提高信号处理的精度。本文结合理论模型分别从时域和频域两个方面提取出环状流界面波速信号中的关键特征，进一步结合支持向量回归、决策树回归、随机森林回归、梯度提升树回归、Bagging回归等模型，以不同流速、压力工况条件下的气液界面波速为预测目标，建立了高精度的波速预测模型。分析结果表明，所提模型能有效捕捉气液两相流中的复杂流动特征，其中梯度提升树回归模型均方根误差（RMSE）为3.96×10^-7，说明模型具有较好的稳定性和预测精度，能够为气液两相流的流动行为分析和工程实践提供有效的支持，特别是在复杂工况下也能够稳定地预测波速，具有广泛的应用前景和潜力。

关键词: 流体力学, 环状流, 界面, 集成学习, 预测