Machine learning the kinematics of spherical particles in fluid flows

Abstract

Numerous efforts have been devoted to the derivation of equations describing the kinematics of finite-size spherical particles in arbitrary fluid flows. These approaches rely on asymptotic arguments to obtain a description of the particle motion in terms of a slow manifold. Here we present a novel approach that results in kinematic models with unprecedented accuracy compared with traditional methods. We apply a recently developed machine learning framework that relies on (i) an imperfect model, obtained through analytical arguments, and (ii) a long short-term memory recurrent neural network. The latter learns the mismatch between the analytical model and the exact velocity of the finite-size particle as a function of the fluid velocity that the particle has encountered along its trajectory. We show that training the model for one flow is sufficient to generate accurate predictions for any other arbitrary flow field. In particular, using as an exact model for trajectories of spherical particles, the Maxey–Riley equation, we first train the proposed machine learning framework using trajectories from a cellular flow. We are then able to accurately reproduce the trajectories of particles having the same inertial parameters for completely different fluid flows: the von Kármán vortex street as well as a two-dimensional turbulent fluid flow. For the second example we also demonstrate that the machine learned kinematic model successfully captures the spectrum of the particle velocity, as well as the extreme event statistics. The proposed scheme paves the way for machine learning kinematic models for bubbles and aerosols using high-fidelity DNS simulations and experiments.

Machine Learning-Driven Multiobjective Optimization: An Opportunity of Microfluidic Platforms Applied in Cancer Research

“…ML algorithms are applied to microfluidic platforms with different formats. They can be classified into three major types: droplet microfluidics (see Figure 2 b), trajectory prediction [ 40 , 41 ], and blood cell counting. Implementing ML in the format of droplet microfluidics is the most popular of these.…”

Deep learning-based prediction of piled-up status and payload distribution of bulk material

“…Chen S. et al [46] established a CNN model with parameters of particle residence time, diameter and density as inputs, and classification of particle height in the initial packing as output. Wan Z. Y. et al [47] combined a fluid mechanics model with machine learning to propose a method describing the motion state of spherical particles in fluid. Goldstein E.B.…”

High-resolution fluid–particle interactions: a machine learning approach

“…With the recent developments in machine learning (ML) methods and their successful application to classical engineering problems, various advances have been made to accelerate numerical methods (Kachrimanis, Karamyan, & Malamataris 2003;Ariana, Vaferi, & Karimi 2015;Benvenuti, Kloss, & Pirker 2016;Chaurasia & Nikkam 2017;Liang et al 2018a,b;Figueiredo et al 2019;Brevis, Muga, & van der Zee 2020;Prieto 2020). This capacity has also been extended to problems related to fluid dynamics and granular flow (Radl & Sundaresan 2014;Kutz 2017;Wan & Sapsis 2018;Fukami, Fukagata & Taira 2019;Li et al 2020a;Park & Choi 2020;Aghaei Jouybari et al 2021) where its applications towards the former has been extensively reviewed (Brenner, Eldredge, & Freund 2019;Brunton, Noack, & Koumoutsakos 2020;Fukami, Fukagata, & Taira 2020a). For example, a ML approach was used for the estimation of gravitational solid flows (Garbaa et al 2014).…”