AUTHOR=Pu Shengyang , Liu Haolin , Ba Yinmengke , Chen Zhiqing , Yu Jiawen , Cui Chunliang 

TITLE=Research on mobile ultrasonic stratified flow velocity measurement based on machine learning algorithms

JOURNAL=Frontiers in Water

VOLUME=Volume 7 - 2025

YEAR=2025

URL=https://www.frontiersin.org/journals/water/articles/10.3389/frwa.2025.1590124

DOI=10.3389/frwa.2025.1590124

ISSN=2624-9375

ABSTRACT=A mobile ultrasonic stratified flow velocity measurement device, which utilizes a pair of ultrasonic transducers, was characterized by its low power consumption and the ability to measure multi-layer flow velocities in channels, offering advantages in water measurement applications in agricultural irrigation areas. The study combined experimental research with computational modeling to investigate the impact of ultrasonic propagation time in downstream and upstream flows and the effect of measurement position on flow velocity prediction. Traditional time-difference methods were found to be less effective for calculating water flow velocities using the proposed mobile ultrasonic transducers. By employing the AdaBoost algorithm with decision tree algorithms as weak classifiers, the relative error of the testing set was <5% in 85.71% of cases, achieving an R2 value of 0.951, an RMSE value of 0.071, an MSE value of 0.795, and an MAE value of 0.654. The experimental results demonstrated that the use of the AdaBoost algorithm from machine learning for the mobile ultrasonic stratified flow velocity measurement device was feasible and effective.