Novel Recurrent neural networks for efficient heat transfer analysis in radiative moving porous triangular fin with heat generation

This paper investigates the use of Artificial Intelligence (AI), notably Recurrent Neural Networks (RNNs), to analyze heat transfer in moving radiative porous triangular systems with heat generation (HTMPTHG). AI-based RNN models are employed to simulate and forecast the complex heat transfer behavi...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Case studies in thermal engineering 2024-12, Vol.64, p.105516, Article 105516
Hauptverfasser: Saqib, Sana Ullah, Farooq, Umar, Fatima, Nahid, Shih, Yin-Tzer, Mir, Ahmed, Kolsi, Lioua
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:This paper investigates the use of Artificial Intelligence (AI), notably Recurrent Neural Networks (RNNs), to analyze heat transfer in moving radiative porous triangular systems with heat generation (HTMPTHG). AI-based RNN models are employed to simulate and forecast the complex heat transfer behavior in these environments, offering a more precise and efficient analysis as compared to traditional numerical methods. The findings of the study highlights the intricate interactions among thermal radiation, porous media, and internal heat generation which plays an integral role in a number of industrial and engineering applications. Recurrent neural network (RNN) is validated to examine the temperature distribution efficiency in a new configuration of triangular, porous, moving fins. Various dimensionless parameters are analyzed for their impact on the effectiveness of portable, transparent, triangular fins. These parameters include permeability, radiation-conduction, Peclet number, thermo-geometric factors, convection-conduction, and surface temperature. The Lobatto III-A numerical technique for HTMPTHG is simulated computationally to provide the synthetic datasets. Then, the RNN supervised computational technique is applied to the generated datasets and the RNN outputs show negligible errors and closely align with numerical observations for all model variant. The effectiveness of Recurrent Neural Networks (RNNs) is rigorously proved through extensive experiments, demonstrating iterative convergence curves for mean squared error, control metrics of optimization and error distribution via histograms.The mean absolute percent error (MAPE), mean absolute error (MAE), and Nash-Sutcliffe efficiency (NSE) are all nearly zero, while the coefficient of determination (R2) is close to 1.Furthermore, there is strong evidence of the prediction accuracy and dependability of the RNN in the regression results for the HTMPTHG model.
ISSN:2214-157X
2214-157X
DOI:10.1016/j.csite.2024.105516