Abstract:Accurate prediction of the dynamics and deformation of freely moving drops is crucial for numerous droplet applications. When the Weber number is finite but below a critical value, the drop deviates from its spherical shape and deforms as it is accelerated by the gas stream. Since aerodynamic drag on the drop depends on its shape oscillation, accurately modeling the drop shape evolution is essential for predicting the drop's velocity and position. In this study, 2D axisymmetric interface-resolved simulations were performed to provide a comprehensive dataset for developing a data-driven model. Parametric simulations were conducted by systematically varying the drop diameter and free-stream velocity, achieving wide ranges of Weber and Reynolds numbers. The instantaneous drop shapes obtained in simulations are characterized by spherical harmonics. Temporal data of the drag and modal coefficients are collected from the simulation data to train a Nonlinear Auto-Regressive models with eXogenous inputs (NARX) neural network model. The overall model consists of two multi-layer perceptron networks, which predict the modal coefficients and the drop drag, respectively. The drop shape can be reconstructed with the predicted modal coefficients. The model predictions are validated against the simulation data in the testing set, showing excellent agreement for the evolutions of both the drop shape and drag.

What problem does this paper attempt to address?

This paper mainly focuses on the droplet dynamics and drag prediction under limited but subcritical Weber number (We) conditions. When the Weber number is below its critical value, the droplet deviates from a spherical shape and deforms under the acceleration of the gas flow. Accurately simulating the shape evolution is crucial for predicting the droplet velocity and position, as the drag is related to the oscillation of the droplet shape. In this study, the authors conducted two-dimensional axisymmetric interface analytical simulations and generated a comprehensive dataset for developing data-driven models. By systematically varying the droplet diameter and free stream velocity, a wide range of Weber numbers and Reynolds numbers (Re) were achieved. The paper utilized a Neural Network model called Nonlinear Auto-Regressive models with eXogenous inputs (NARX) to train the modal coefficients and drag of the droplet shape. The model consists of two multilayer perceptron networks, which predict the modal coefficients and droplet drag respectively. The droplet shape can be reconstructed by the predicted modal coefficients, and the model shows good consistency with simulated data in the test dataset. In the study, the droplet was exposed to a uniform gas flow, and the shape oscillation, drag variation, and complex interaction with the surrounding fluid make the prediction challenging. The authors developed a model that can predict the evolution of droplet shape and drag through numerical simulations and machine learning methods, especially the NARX neural network. The focus of the research is on the behavior of droplets that do not experience break-up under subcritical Weber number conditions. In conclusion, this paper aims to address how to predict the dynamic deformation and drag of droplets in complex flow fields through data-driven methods without considering the details of each droplet interface. This is of significant importance for understanding the droplet dynamics in applications such as spraying, fuel injection, and cooling.

Data-driven modeling of the aerodynamic deformation and drag for a freely moving drop in the sub-critical Weber number regime

Data-driven modeling of the aerodynamic deformation and drag for a freely moving drop in the sub-critical Weber number regime

Simulation and modeling of the vaporization of a freely moving and deforming drop at low to moderate Weber numbers

Data-driven Nonlinear Constitutive Relations for Rarefied Flow Computations

Analysis and Development of Novel Data-Driven Drag Models Based on Direct Numerical Simulations of Fluidized Beds

Novel Models for Predicting the Shape and Motion of an Ascending Bubble in Newtonian Liquids Using Machine Learning

Drag prediction of rough-wall turbulent flow using data-driven regression

Data-driven modelling for drop size distributions

Drag Model from Interface-Resolved Simulations of Particle Sedimentation in a Periodic Domain and Vertical Turbulent Channel Flows

Models of Interphase Drag Force from Direct Numerical Simulations of Upward Turbulent Particle-Laden Channel Flows

The effect of droplet deformation and internal circulation on drag coefficient

Detailed numerical investigation of the drop aerobreakup in the bag breakup regime

Detailed Numerical Simulation of Unsteady Drag Coefficient of Deformable Droplet

Physics-guided Design and Learning of Neural Networks for Predicting Drag Force on Particle Suspensions in Moving Fluids

Data-driven modeling for unsteady aerodynamics and aeroelasticity

Homogeneous drag models in gas–solid fluidization: Big data analytics and conventional correlation

DrivAerNet: A Parametric Car Dataset for Data-Driven Aerodynamic Design and Graph-Based Drag Prediction

DCC-DNN: A deep neural network model to predict the drag coefficients of spherical and non-spherical particles aided by empirical correlations

Data-driven turbulence modeling in separated flows considering physical mechanism analysis

A consistent volume-of-fluid approach for direct numerical simulation of the aerodynamic breakup of a vaporizing drop

Data-driven time-dependent state estimation for interfacial fluid mechanics in evaporating droplets