Abstract:

Aiming at the accurate identification of horizontal oil-gas-water three-phase flow pattern, a combined pulse-wave and continuous-wave ultrasonic sensing measurement method and a multi-domain feature extraction scheme were proposed. In this study, horizontal oil-gas-water three-phase flow dynamic experiments were firstly conducted, and a pulse-wave ultrasonic sensor with a single piezoelectric chip and a continuous-wave ultrasonic sensor with two piezoelectric chips were adopted to synchronously acquire the echo intensity data and Doppler frequency shift data during the flow process. By analyzing the signal responses to different flow patterns, the flow characteristics of different oil-gas-water three-phase flow patterns were revealed. Then, the probability distribution of the radial position of the maximum value in the echo intensity profile, the time series of the average flow velocity and the decomposition of the Doppler frequency shift signal were calculated, and several features were extracted from them to quantify the distribution characteristics of phase interfaces in spatial domain and the fluctuation characteristics of the flow velocity in time and frequency domains. Using the extracted multi-domain feature vector, a classifier based on the random forest was finally constructed, and 8 types of horizontal oil-gas-water three-phase flow patterns were accurately identified with a total identification accuracy of 96.6%. The proposed method provides a simple, efficient, low-cost, and non-invasive solution for the flow pattern identification of complex industrial multiphase flows, which has important scientific and engineering significance.

Key words: oil-gas-water three-phase flow, velocity, interface, measurement, flow pattern identification, instrumentation

摘要：

针对水平管道油气水三相流流型的准确辨识问题，提出一种脉冲波/连续波超声组合式测试手段与多域特征提取方案。研究中，首先利用单晶的脉冲波超声传感器与双晶的连续波超声传感器对水平管道油气水三相流开展动态实验测试，同步获取流动过程中的回波强度数据与多普勒频移数据。通过分析测试数据在不同流型下的响应特性，揭示了油气水三相流不同流型的流动特点，并从回波强度剖面中峰值所在管道径向位置的概率分布、平均流速波动时间序列与多普勒频移的频率分解中提取了相关特征，量化了相界面的空间域分布特性与流速的时域、频域波动特性。最终，利用所提取的多域特征向量搭建了基于随机森林的分类器，实现了水平管道中8种油气水三相流流型的准确识别，平均识别率达96.6%。所提方法为复杂工业多相流的流型辨识问题提供了一种简单、高效、低成本、非侵入性的解决方案，具有重要的科学与工程应用价值。

关键词: 油气水三相流, 流速, 界面, 测量, 流型识别, 仪器仪表