Research On Dynamic Characteristics Of Nonlinear Tank Sloshing And Parameter Prediction Based On Machine Learning Method

Liquid sloshing refers to the phenomenon of the fluctuation of the liquid in the partially loaded tank and the interaction with the bulkhead structure under the action of external excitation.Especially when the external excitation frequency is close to the natural frequency of the tank,the liquid in the tank will have a violent resonance response,which may change the dynamic characteristics of the main structure and bring potential safety risks.This problem has a wide range of applications in engineering,such as tanks for transporting liquefied natural gas and rocket fuel tanks.Therefore,it is necessary to study the characteristics of sloshing liquids.The physical model experiment is used to study the liquid sloshing problem under horizontal harmonic excitation.The hydrodynamic characteristics of the liquid sloshing in the tank under the conditions of different liquid depths,excitation frequencies and excitation amplitudes are studied based on the continuous wavelet transform method.The time-frequency characteristics of the sloshing pressure time history curve on the side wall of the tank are analyzed,and the time-frequency change law of the sloshing pressure at the static liquid level on the side wall of the tank is obtained.The study found that the main frequency component of the sloshing pressure is the excitation frequency in the initial stage and the steady stage of the pressure.In the attenuation stage,the main frequency components of the sloshing pressure are related to the liquid depth and the excitation frequency.A numerical model of tank sloshing is established based on potential flow theory and timedomain boundary element method.Apply instantaneous boundary conditions to track largescale free surface deformations.An artificial viscous method is introduced into the model to make the sloshing process include viscous dissipation effects.A completely nonlinear simulation of liquid sloshing in a tank is studied.The viscosity coefficients of numerical models are often empirical and need to be calibrated with experimental data.In order to adaptively correct the relevant viscosity coefficients,this paper proposes a machine learning method based on back-propagation neural networks,and uses the database constructed by the physical model tests in this paper to train and test the network.Studies have shown that the back propagation neural network prediction model can accurately predict the viscosity coefficient of nonlinear sloshing with high probability,and has great potential to reproduce the nonlinear sloshing process under various working conditions.Based on the linear and nonlinear methods,the pressure jump up phenomenon that occurs when the tank is braked in the physical model experiment is studied.Considering the influence of different braking methods,viscosity coefficients and braking moments on the frequency range and amplitude of pressure sudden changes.The study found that the faster the braking speed of the tank,the smaller the viscosity coefficient,and the more obvious the pressure sudden change.In different excitation frequency ranges,the reasons for the pressure jump up phenomenon are different.Under non-linear conditions,the pressure jump up phenomenon in the tank is more pronounced than under linear conditions.