| Gas-liquid two-phase flow widely exists in nature and engineering fields,and plays an important role in sea-air exchange,cavitation of high-speed dam flow on buildings,natural gas transportation.Under the combined action of gravity,buoyancy and surface tension,bubbles will deform,rupture and coalesce,which will cause high-speed jet flow and strong turbulence in nearby water bodies,thus affecting air-sea exchange and pollutant movement.The interaction between bubbles is closely related to the motion state and properties of a single bubble.Therefore,understanding the motion characteristics of a single bubble is an important part of better understanding basic problems such as the prediction of sea-air flux exchange and the influence of bubbles on the force of structures,which provide theoretical basis for Marine engineering construction.In this research,a large number of experiments were carried out to investigate the bubble motion characteristics in still water.The movement process of bubbles is captured by High Speed Camera,and data processing is carried out based on MATLAB.The rising speed,the shape and the trajectory changes of bubbles are analyzed.Finally,the prediction model is constructed based on machine learning technology,and the data set is constructed using the experimental results.The bubble motion speed,bubble drag coefficient and bubble shape are studied and trained.The obtained results are compared with previous empirical formulas to evaluate the accuracy of the model.The results can be summarized as follows:After the Rising Air Bubbles in still water is produced,it generally goes through three stages:bubble acceleration,bubble deceleration and bubble stable fluctuation.The stages of bubble acceleration and deceleration all occur in the straight rising section of bubble and last for a short period of time less than 0.1s.When the bubble is in the stable fluctuation section,the final bubble velocity VT decreases with the increase of d,and the fluctuation amplitude of bubble velocity increases with the increase of d;The bubble detachment frequency P mainly affects the bubble velocity by affecting the initial kinetic energy of the bubble.The larger P,the larger VT.VT is more correlated with bubble Eo number.And the drag coefficient of bubbles firstly decreases and then increases with the increase of Reynolds number,and the drag coefficient is relatively discrete with the increase of Reynolds number;The drag coefficient has a positive correlation with the change of Eo number,and the correlation degree is higher than Reynolds number.At the same time,the agreement between the experimental drag coefficient and the calculated value of Dijkhuizen formula is the best.Bubble trajectory is related to bubble size and initial disturbance.The shape of bubble trajectory changes from zigzag to disordered with the increase of d;With the increase of d,the bubble offset increases first and then decreases.Bubble shape is described by bubble aspect ratio and circularity.when the bubble diameter is small(d<4mm),the correlation between bubble aspect ratio and bubble diameter is large.when d>4mm,the bubble CT is small with the increase of d,and the change of bubble shape is complicated.While the bubble inclination angle increases first and then remains unchanged with the increase of d.Finally,BP neural network is used to construct the relationship between three variables(the final bubble velocity,drag coefficient,the shape of bubble)and three dimensionless number(Reynolds number,Eo number,Weber number).The results show that the BP neural network can reflect the relationship between the final bubble velocity,drag coefficient and three dimensionless number very well.At the same time,the relationship between shape of bubble and Reynolds number,Eo number,Weber number is constructed by other algorithms such as:KNN,random forest and logistic regression.It is found that logistic regression makes a more accurate prediction than random forest.Considering the results of random forest,Weber number can see that the Eo number has the greatest influence on the shape of the bubble,and Reynolds number is second,and the We number has minimal influence. |
