Experimental Investigation And Prediction Of Turbulent Drag Reduction Rate And Viscoelastic Stress Caused By Polymer Solution

By adding water-soluble polymer or some kinds of surfactant additives into the turbulent flow,the resistance of turbulent flow can be effectively reduced,which is called viscoelastic additive turbulent drag reduction technology.However,it can't be popularized in practical engineering applications because it is impossible to judge the effect of drag reduction accurately.In view of the drag reduction phenomenon,although the drag reduction mechanism has not yet been fully determined,the key role of additive viscoelasticity has been become a consensus.Viscoelastic stress plays an important role in the study of turbulent drag reduction of Non-Newtonian fluids,understanding its distribution in the channel is conducive to understanding the drag reduction mechanism of viscoelastic additives.Based on the self-designed experimental platform of turbulent drag reduction,polyethylene oxide(PEO)solution was injected into the turbulent channel flow of water through a narrow slot on the channel wall,and conducted experimental measurements of the velocity fields and concentration fields of the drag-reducing flow based on the Particle Image Velocimetry(PIV)and Planar Laser Induced Fluorescence(PLIF)techniques,and the micro pressure difference measuring device was used to measure the water flow pressure difference in the observation section and the pressure difference of PEO solution injected.The instantaneous velocity field and concentration field in the downstream of the narrow channel under different Reynolds numbers were obtained.According to the turbulence statistical theory,the statistical calculation of velocity field and concentration field was carried out by using the self-made program,and the statistical characteristics of turbulence in the channel were extracted and analyzed.The drag reduction rate under various conditions was calculated through the pressure difference.Then,based on a large number of experimental data,two BP neural network models were established: one is a three-layer BP neural network with Reynolds number,PEO solution concentration and injection flow rate as input and drag reduction rate as output;the other is a three-layer BP neural network with Re,PEO solution concentration,injection flow rate and vertical distance from the wall as input and viscoelastic stress as output.The influence of neural network structure and different parameter design on neural network prediction model was analyzed.Finally,through the comparison between the predicted value of BP neural network model and the actual experimental value,the coincidence is obtained,in order to analyze the feasibility of the trained BP neural network model.The results show that:(1)On the same flow conditions,with the increase of injection flow rate and concentration of PEO solution,the drag reduction rate gradually increases and tends to be stable,and the maximum drag reduction rate in the experiment is 26.87%;the Reynolds shear stress increases first and then decreases along the vertical wall direction,and decreases with the increase of injection concentration,and the peak position of Reynolds shear stress in the channel moves to the center of the channel move.(2)Under the influence of PEO solution injection,the viscoelastic structure near the wall increases,the viscoelastic stress increases,and the Reynolds shear stress in drag reduction flow decreases.This redistributed stress breaks the symmetry of the average flow velocity on the channel center line,and the change of viscoelastic solution near the wall plays a key role.(3)On the different injection concentrations,the concentration distribution of PEO solution near the wall is quite different,and it is basically the same far away from the wall.With the increase of injection concentration,the concentration near the wall increases,and the viscoelastic stress also increases,which indicates that the effect of viscoelastic stress on turbulent flow is mainly concentrated in the near wall region.The peak value of viscoelastic stress and concentration distribution in the near wall region decrease with the increase of Reynolds number,and there is a positive correlation between them.(4)The drag reduction rate predicted by BP neural network model is similar to that of drag reduction flow formed by injection of PEO solution.The coefficient of determination R~2 = 0.9892,and the prediction accuracy is high,and it has better generalization ability.The results of viscoelastic stress prediction are similar to the actual values.The coefficient of determination R~2 =0.9876.It is expected that BP neural network will be applied to the research of drag reduction mechanism and engineering application.