Study On Noncontact Measurement Of Surface Tension Of Supercoling Nanofluid Based On Artificial Neural Network

Supercooling is a process that nanofluids and other phase change materials must go through before freezing.Measuring the surface tension of supercooled nanofluids is of great significance for its practical application and numerical simulation in cold storage systems.Since traditional contact surface tension measurement methods are prone to heterogeneous nucleation of supercooled liquids,it is necessary to use non-contact surface tension measurement methods,such as the acoustic levitation droplet oscillation method,to measure the surface tension of supercooled nanofluids.The principle of the acoustic levitation droplet oscillation method is to calculate the surface tension based on the resonance frequency of the acoustic levitation droplet,which involves many disciplines such as acoustics,thermodynamics,image recognition and spectrum analysis.There are few studies on the acoustic levitation droplet oscillation method at home and abroad,and there is no complete set of experimental equipment and data processing standards,and there are some influencing factors that are difficult to ignore during the experiment,so the acoustic levitation droplet oscillation method still faces some key problems to be solved urgently when used to measure the surface tension of supercooling nanofluids.This thesis has carried out detailed and in-depth research on the subject of surface tension measurement of supercooling graphene oxide nanofluids based on acoustic levitation.The main research contents and conclusions are as follows:(1)Firstly,a large number of optimization analyses were carried out on the four major devices of the acoustic suspension droplet oscillation experiment system:acoustic suspension device,cooling device,oscillation excitation device and image acquisition device,which greatly reduced the cost required to build the experimental system while increasing the experimental efficiency of the experimental system is improved.Then,according to the optimized experimental system,a large number of second-order droplet oscillation images of deionized water and graphene oxide nanofluid in supercooled and non-supercooled states were collected,and obtained according to CANNY edge detection method and fast Fourier transform Resonance frequency.The analysis found that the experimental value of surface tension calculated directly according to the Rayleigh formula has obvious errors.(2)Through analysis,it is found that the influencing factors such as frequency shift phenomenon,temperature fluctuation,viscosity and droplet deformation are the reasons for the error of the experimental value.The 8 most important parameters affecting the surface tension of acoustic suspension droplets are defined:mass concentration?,temperature T,droplet mass M,resonance frequency f_m,temperature fluctuation?T,viscosity?,frequency shift?f,and average aspect ratio?.and fitted with a new Rayleigh formula correction model,which can accurately correct the experimental value of the surface tension of deionized water.(3)Considering the limitations of the modified formula,this paper established a double hidden layer feedforward artificial neural network model to predict the surface tension of the supercooled graphene oxide nanofluid.Compared with the modified model,the artificial neural network model shows higher reliability and adaptability.The prediction results of the surface tension of the supercooled graphene oxide nanofluid have almost no obvious dispersion and deviation.Finally,according to the predicted results,two high-precision empirical formulas for the surface tension of the supercooled graphene oxide nanofluid are fitted.