ORIGINAL RESEARCH article
Front. Mech. Eng.
Sec. Fluid Mechanics
This article is part of the Research TopicTechnical Briefs in Mechanical Engineering: Volume 2View all articles
Fluid Flow and Mass Transport along a Microchannel with Physics-Informed Neural Networks
Provisionally accepted
Mehmet Melih Tatlisoz
Cetin Canpolat*- Cukurova Universitesi Muhendislik Fakultesi, Adana, Türkiye
Select one of your emails
You have multiple emails registered with Frontiers:
Notify me on publication
Please enter your email address:
If you already have an account, please login
You don't have a Frontiers account ? You can register here
In the present study, fluid and mass transport along a microchannel are aimed to be solved using physics-informed neural networks (PINN), which is an emerging solution method for partial differential equations (PDEs). For this purpose, Continuity, Navier-Stokes, and Nernst-Planck Equations, which mainly characterize momentum and mass conservation, are solved simultaneously for pressure-driven and electroosmotic flow conditions. In both solutions, PINN is used to solve steady-state and time-dependent equations for various Reynolds numbers within 5x10-6≤ Re ≤ 5x10-2. The non-dimensional form of the PDEs is employed, and the pressure variable is made dimensionless with its maximum magnitude, in contrast to previous studies, to minimize the error level better. For comparison purposes, the same cases under identical conditions are modeled with a solver for finite element analysis (FEA), COMSOL Multiphysics. It is observed that both Navier-Stokes and Nernst-Planck Equations can be solved with PINN in a single model, regardless of the flow generation method and the time dependency of the model. However, the pressure scale must be determined as inlet pressure magnitude, instead of the usual scaling of viscosity multiplied by the velocity scale divided by the length scale, to descend within reasonable error margins. Reasonably low calculation errors are reached, such as 3.04x10-5, 4.76x10-3, 3.60x10-4, 6.02x10-3 for steady pressure-driven flow, time-dependent pressure-driven flow, steady electroosmotic flow, and time-dependent electroosmotic flow, respectively. Therefore, the results of the FEA solver exhibit excellent overlap with the data obtained from PINN. For PINN, species concentrations of 0.52-1.75-2.71-3.36 mol/m3 are accumulated at the outlet under pressure-driven flow for time points of 50-100-150-200 seconds, respectively. Under electrokinetic flow, species concentrations are varied as 0.72-1.93-2.83-3.46 mol/m3 for the same time scales, respectively. In consideration of FEA, species concentrations are calculated as 0.52-1.75-2.71-3.39 mol/m3 and 0.55-1.85-2.84-3.52 mol/m3 under similar conditions for pressure-driven and electroosmotic flows, respectively. As a result, PINN is a notable alternative for simultaneous modeling of the flow and mass transport under pressure-driven and electrokinetic conditions in microfluidic applications.
Keywords: Microfluidics, Physics-informed neural network, navier-stokes equation, Nernst-Planck equation, Electrokinetics
Received: 21 Jun 2025; Accepted: 05 Dec 2025.
Copyright: © 2025 Tatlisoz and Canpolat. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
* Correspondence: Cetin Canpolat
Disclaimer: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.