Summary of the Experiment:
The study presented at the 14th International Symposium on Ultrasonic Doppler Methods for Fluid Mechanics and Fluid Engineering aimed to classify flow regimes in bubble columns using the integration of ultrasound technology and the k-nearest neighbors (KNN) algorithm. The experiment focused on obtaining the standard deviation of bubble velocity distributions and echo signal characteristics to develop a KNN model capable of identifying bubbly, transition, and churn-turbulent flow regimes. The experimental setup included a circular tank with a bubble generator and gas hold-up monitoring to provide comprehensive data for model training and validation.
UVP-DUO Usage:
The Met-Flow Ultrasonic Velocity Profiler (UVP) was employed to capture detailed velocity profiles and echo signal characteristics within the bubble column. The ultrasonic transducer used in the experiment was a 4 MHz model, with key parameters set as follows:
- Center frequency: 4 MHz
- Number of cycles: 4
- Emission voltage: 140 Vp-p
- Receiving gain: 30 dB
- Pulse repetition frequency: 8 kHz
- Number of repetitions: 64
The UVP system included a pulser/receiver, digitizer, and computer, controlled via LabVIEW software. The precision of the UVP-DUO allowed for accurate measurement of bubble velocities and echo signal characteristics, which were critical for the development of the KNN classification model.
Result of the Experiment and How UVP-DUO Contributed:
The results demonstrated that the integration of UVP measurements with the KNN algorithm effectively classified flow regimes with high accuracy. Specifically, the model achieved a classification accuracy of 96.4% when tested with 28 data sets covering the three flow regimes. The UVP-DUO's high spatial and temporal resolution enabled the precise determination of bubble velocities and echo signal characteristics, which were essential attributes for the KNN model.
The UVP-DUO's contributions were pivotal in providing the detailed data necessary for accurate flow regime classification. The standard deviation of bubble velocity and the slope of the autocorrelation function of the echo signal were key inputs for the KNN model, demonstrating the UVP-DUO's capability in non-intrusive, high-resolution flow measurement. This study highlights the effectiveness of combining advanced ultrasonic measurement techniques with machine learning algorithms to enhance the understanding and control of multiphase flow systems in industrial applications.
How UVP-DUO Supports Classification of Flow Regimes in Bubble Columns