Paradigm predictive analysis of two-phase Eyring–Powell fluid flow over a vertical stretching sheet with temperature-dependent viscosity by multilayer neural networks

Recently, the accuracy and efficiency of solving complex fluid problems have benefited greatly from these latest advances in AI-integrated fluid mechanics. Deep-learning neural networks optimized with the Levenberg-Marquardt algorithm (DLNNs-LMA) is a supervised AI-based approach that is used to ana...

Full description

Saved in:
Bibliographic Details
Main Authors: Zahoor Shah, Hamza Iqbal, Waqar Azeem Khan, Taseer Muhammad, Muhammad Shoaib
Format: Article
Language:English
Published: Elsevier 2025-08-01
Series:Case Studies in Thermal Engineering
Subjects:
Online Access:http://www.sciencedirect.com/science/article/pii/S2214157X25005490
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849387091596673024
author Zahoor Shah
Hamza Iqbal
Waqar Azeem Khan
Taseer Muhammad
Muhammad Shoaib
author_facet Zahoor Shah
Hamza Iqbal
Waqar Azeem Khan
Taseer Muhammad
Muhammad Shoaib
author_sort Zahoor Shah
collection DOAJ
description Recently, the accuracy and efficiency of solving complex fluid problems have benefited greatly from these latest advances in AI-integrated fluid mechanics. Deep-learning neural networks optimized with the Levenberg-Marquardt algorithm (DLNNs-LMA) is a supervised AI-based approach that is used to analyze the Eyring Powell fluid in a two-phase flow (EPFM-TPF) with dust particles and at temperature-dependent viscosity. PDEs are converted into nonlinear ODEs with similarity transformations, which are solved using the Adam method in Mathematica with varied physical parameters to generate a reference dataset. Finally, we examine graphically velocity and temperature profiles of both fluid and dust particles and for significant parameters. The novel DLNNs-LMA were used for the numerical computations on the reference data. The error analysis indicates a close correlation with the proposed databases and with reference datasets, having values between 10-3 and 10-9, which indicates a high accuracy of the proposed methodology. Its consistently low MSE (mean square error) values (at the scale of E-09 to E-10) make it a good approximation for most scenarios; there is very little deviation and strong predictability. The results obtained validate the robustness of the DLNNs-LMA approach to solve complex fluid dynamics problems. Further, there is variation in epochs (from 54 to 585) to account for the adaptability of the model, leading to the best convergence despite differences in scenarios. It is observed that the velocity profile gradually increases with the rising trend of M and β, and the absolute error was found in the range between E-04 to E-07 and E-02 to E-07, respectively.
format Article
id doaj-art-89cab2d0cf1e469bbd8b228525ad83cd
institution Kabale University
issn 2214-157X
language English
publishDate 2025-08-01
publisher Elsevier
record_format Article
series Case Studies in Thermal Engineering
spelling doaj-art-89cab2d0cf1e469bbd8b228525ad83cd2025-08-20T03:55:22ZengElsevierCase Studies in Thermal Engineering2214-157X2025-08-017210628910.1016/j.csite.2025.106289Paradigm predictive analysis of two-phase Eyring–Powell fluid flow over a vertical stretching sheet with temperature-dependent viscosity by multilayer neural networksZahoor Shah0Hamza Iqbal1Waqar Azeem Khan2Taseer Muhammad3Muhammad Shoaib4Department of Mathematics, COMSATS University Islamabad, Islamabad Campus, PakistanDepartment of Mathematics, COMSATS University Islamabad, Islamabad Campus, PakistanDepartment of Mathematics, Mohi-ud-Din Islamic University, Nerian Sharif, 12010, Azad, Jammu & Kashmir, Pakistan; Corresponding author;Department of Mathematics, COMSATS University Islamabad, Islamabad Campus, PakistanDepartment of Mathematics, College of Science, King Khalid University, Abha, Saudi ArabiaYuan Ze University, AI centre, Taoyuan, 320, TaiwanRecently, the accuracy and efficiency of solving complex fluid problems have benefited greatly from these latest advances in AI-integrated fluid mechanics. Deep-learning neural networks optimized with the Levenberg-Marquardt algorithm (DLNNs-LMA) is a supervised AI-based approach that is used to analyze the Eyring Powell fluid in a two-phase flow (EPFM-TPF) with dust particles and at temperature-dependent viscosity. PDEs are converted into nonlinear ODEs with similarity transformations, which are solved using the Adam method in Mathematica with varied physical parameters to generate a reference dataset. Finally, we examine graphically velocity and temperature profiles of both fluid and dust particles and for significant parameters. The novel DLNNs-LMA were used for the numerical computations on the reference data. The error analysis indicates a close correlation with the proposed databases and with reference datasets, having values between 10-3 and 10-9, which indicates a high accuracy of the proposed methodology. Its consistently low MSE (mean square error) values (at the scale of E-09 to E-10) make it a good approximation for most scenarios; there is very little deviation and strong predictability. The results obtained validate the robustness of the DLNNs-LMA approach to solve complex fluid dynamics problems. Further, there is variation in epochs (from 54 to 585) to account for the adaptability of the model, leading to the best convergence despite differences in scenarios. It is observed that the velocity profile gradually increases with the rising trend of M and β, and the absolute error was found in the range between E-04 to E-07 and E-02 to E-07, respectively.http://www.sciencedirect.com/science/article/pii/S2214157X25005490— artificial intelligenceDeep learningNeural networkEyring-powell fluidTwo-phase flowLevenberg-marquardt algorithm
spellingShingle Zahoor Shah
Hamza Iqbal
Waqar Azeem Khan
Taseer Muhammad
Muhammad Shoaib
Paradigm predictive analysis of two-phase Eyring–Powell fluid flow over a vertical stretching sheet with temperature-dependent viscosity by multilayer neural networks
Case Studies in Thermal Engineering
— artificial intelligence
Deep learning
Neural network
Eyring-powell fluid
Two-phase flow
Levenberg-marquardt algorithm
title Paradigm predictive analysis of two-phase Eyring–Powell fluid flow over a vertical stretching sheet with temperature-dependent viscosity by multilayer neural networks
title_full Paradigm predictive analysis of two-phase Eyring–Powell fluid flow over a vertical stretching sheet with temperature-dependent viscosity by multilayer neural networks
title_fullStr Paradigm predictive analysis of two-phase Eyring–Powell fluid flow over a vertical stretching sheet with temperature-dependent viscosity by multilayer neural networks
title_full_unstemmed Paradigm predictive analysis of two-phase Eyring–Powell fluid flow over a vertical stretching sheet with temperature-dependent viscosity by multilayer neural networks
title_short Paradigm predictive analysis of two-phase Eyring–Powell fluid flow over a vertical stretching sheet with temperature-dependent viscosity by multilayer neural networks
title_sort paradigm predictive analysis of two phase eyring powell fluid flow over a vertical stretching sheet with temperature dependent viscosity by multilayer neural networks
topic — artificial intelligence
Deep learning
Neural network
Eyring-powell fluid
Two-phase flow
Levenberg-marquardt algorithm
url http://www.sciencedirect.com/science/article/pii/S2214157X25005490
work_keys_str_mv AT zahoorshah paradigmpredictiveanalysisoftwophaseeyringpowellfluidflowoveraverticalstretchingsheetwithtemperaturedependentviscositybymultilayerneuralnetworks
AT hamzaiqbal paradigmpredictiveanalysisoftwophaseeyringpowellfluidflowoveraverticalstretchingsheetwithtemperaturedependentviscositybymultilayerneuralnetworks
AT waqarazeemkhan paradigmpredictiveanalysisoftwophaseeyringpowellfluidflowoveraverticalstretchingsheetwithtemperaturedependentviscositybymultilayerneuralnetworks
AT taseermuhammad paradigmpredictiveanalysisoftwophaseeyringpowellfluidflowoveraverticalstretchingsheetwithtemperaturedependentviscositybymultilayerneuralnetworks
AT muhammadshoaib paradigmpredictiveanalysisoftwophaseeyringpowellfluidflowoveraverticalstretchingsheetwithtemperaturedependentviscositybymultilayerneuralnetworks