Computational Fluid Dynamics and Machine Learning Algorithms Analysis of Striking Particle Velocity Inside an Acinar Region of Human Lung

Aug 5, 2026

Upload by Sage

Complementing computational fluid dynamics (CFD) simulations with machine learning algorithms accelerates the classification, prediction, and optimization of how particle properties influence deposition in the lung. The study numerically determines the optimum particle diameter by targeting an ideal striking velocity magnitude and impact time inside a human acinus. CFD computes these two independent properties for three diameters, then ML classifiers infer the diameter corresponding to each case. Healthy and diseased acini are compared through different surface tension conditions.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages36

Reading Time1 h 31 min

AccessFree to download

Abstract

Keywords

Introduction

Pulmonary surfactant and acinar anatomy

Nanoparticles, lung diseases, and drug delivery

Machine learning algorithms and CFD integration

PLEASE CITE THIS ARTICLE AS DOI : 10. 1063/5.0106594 Accepted to Phys. Fluids 10.1063/5.0106594 Computational Fluid Dynamics and Machine Learning Algorithms Analysis of Striking Particle Velocity Magnitude, Particle Diameter, and Impact Time Inside an Acinar Region of Human Lung Isabella Francis1, Suvash C. Saha1, * 1School of Mechanical and Mechatronic Engineering, Faculty of Engineering and Information Technology, University of Technology Sydney, NSW, Australia *Corresponding author: Suvash C. Saha; Email: [Suvash.Saha@uts.edu.au](Suvash.Saha@uts.edu.au) Abstract. Complementing computational fluid dynamics (CFD) simulations with machine learning Algorithms (MLA) is becoming increasingly popular as the combination reduces the computational time of the CFD simulations required for classifying, predicting, or optimizing the impact of geometrical and physical variables of a specific study. The main target of drug delivery studies is indicating the optimum particle diameter for targeting particular locations in the lung to achieve a desired therapeutic effect. In addition, the main goal of molecular dynamics studies is to investigate particle-lung interaction through given particle properties. Therefore, this study combines the two by numerically determining the optimum particle diameter required to obtain an ideal striking velocity magnitude (velocity at the time of striking the alveoli, i.e. deposition by sedimentation/diffusion) and impact time (time from release until deposition) inside an acinar part of the lung. At first, the striking velocity magnitudes and time for impact (two independent properties) of three different particle diameters (0.5 􀟤􀝉, 1.5 􀟤􀝉, 5 􀟤􀝉) are computed using CFD simulations. Then, machine learning classifiers determine the particle diameter corresponding to 1 PLEASE CITE THIS ARTICLE AS DOI : 10. 1063/5.0106594 Accepted to Phys. Fluids 10.1063/5.0106594 these two independent properties. In this study, two cases are compared: A healthy acinus where a surfactant layer covers the inner surface of the alveoli providing low air-liquid surface tension (LST) values (10 􀝉􀜰/􀝉), and a diseased acinus where only a water layer covers the surface causing high surface tension (HST) values (70 􀝉􀜰/􀝉) . In this study, the airflow velocity throughout the breathing cycle corresponds to a person with a respiratory rate of 13 breaths per minute and a volume flow rate of 6 􀝈/􀝉􀝅􀝊 . Accurate machine learning (ML) results showed that all three particle diameters attain larger velocities and smaller impact times in a diseased acinus compared to a healthy one. In both cases, the 0.5 􀟤􀝉 particles acquire the smallest velocities and longest impact times, while the 1.5 􀟤􀝉 particles possess the largest velocities and shortest impact times. Keywords: Computational fluid dynamics (CFD); Machine learning classification; Particle striking velocity magnitude; Time for impact; Surface tension; Surfactant Introduction Pulmonary surfactant is a vital biological barrier residing on the lung’s inner surface. It reduces the risk of infection by protecting the lungs from toxic nanoparticles, preserving lung homeostasis, and preventing alveolar collapse at the end of exhalation by decreasing the surface tension of the air-liquid interface inside the lungs (Chroneos et al. 2010) . A pulmonary acinus is a gasexchanging lung unit located distal to a single terminal bronchus (Haefeli‐Bleuer and Weibel 1988), composed of a group of respiratory bronchioles, alveolar ducts, alveolar sacs and alveoli. The alveolar surface of the lungs is composed of ultrathin lung tissue that allows maximum gas exchange but also poses a threat upon inhaling toxic particles that can easily penetrate through the tissue into the bloodstream. 2 PLEASE CITE THIS ARTICLE AS DOI : 10. 1063/5.0106594 Accepted to Phys. Fluids 10.1063/5.0106594 Toxic nanoparticles mainly originate from the combustion of fossil fuels, wildfires, biomass burning,

FAQs

How are CFD and machine learning combined in this study?

CFD simulations compute striking velocity magnitude and impact time for different particle diameters, and machine learning classifiers map these two independent properties to the corresponding diameter.

What biological conditions are compared to represent healthy versus diseased acini?

A healthy acinus is modeled with a surfactant layer providing low air-liquid surface tension, while a diseased acinus uses only a water layer causing high surface tension.

How do the particle diameter cases differ in velocity and impact time results?

Across both cases, 0.5 μm particles show the smallest striking velocities and the longest impact times, while 1.5 μm particles show the largest velocities and the shortest impact times.

Aug 5, 2026

Upload by Sage