弹状流液弹区含气率分布的试验研究

采用光纤探针法和高速摄影法对垂直上升气液两相弹状流的液弹区含气率分布进行了试验研究，测试管径为15 mm。一种基于机器学习的图像处理技术用来识别气液两相界面，通过搭建气泡边界提取的神经网路系统，使用构建的气泡边界数据库对模型进行多次迭代训练，该方法可以有效地识别多种复杂类型的气泡边界。试验得到了弹状流液弹区的径向含气率分布曲线，结果表明，壁峰分布是液弹区含气率分布的主要形式，其中泰勒气泡的尾迹效应对分布形式有重要影响，尾迹的旋涡中心和含气率分布的峰值相对应。针对弹状流液弹区径向含气率分布的两个主要特征——中心局部含气率和壁峰位置，分别提出了相应的预测公式，且与本文的试验数据吻合良好。

Experimental study on void fraction distribution in liquid slug of vertical upward slug flow

The fiber probe method and the high-speed photography method are used to test the void fraction distribution in the liquid slug of a vertical upward co-current air-water slug flow. Experiments were performed in a polymethyl methacrylate tube of 15 mm inner diameter. An improved image processing technique based on the machine learning was employed, which can effectively recognize various complex types of bubble boundary. By building a neural network system for bubble boundary extraction, the iterative training of the model was performed by using the constructed bubble boundary database. The radial void fraction distribution profiles of the liquid slugs were obtained and the results showed a typicalwall-peak distribution. The wake effect of the Taylor bubble had aprominent influence on the distribution form, and the wall peaks were observed to correspond to the vortex center of the Taylor bubble wakes. Regarding the local void fraction in the tube center and the radial position of the wall peak, two corresponding predictive correlations were proposed and the experimental data in the present study was found to agree well with the correlations.