Reduced Order Framework for 2D Heat Transfer Simulation Under Variations of Boundary Conditions Based on Deep Learning Algorithm

Articles in Press

Document Type : Original Article

Authors

1 university of qom, qom, iran

2 Department of Mechanical Engineering, University of Qom

3 University of Qom

Abstract

Due to the high time and computational cost of the direct or numerical solution methods of the partial differential equations governing phenomena, methods based on deep learning algorithms are commonly used in modeling phenomena. In this research, to simulate temperature changes in a two-dimensional field, three different methods based on deep learning algorithms have been used, including a model based on convolutional neural networks, model based on convolutional kernel containing the desired phenomenon pattern and a reduced-dimension Model using autoencoder method. The results of two-dimensional equilibrium temperature distribution simulations using the three methods were compared with the results obtained from the finite difference method. The results showed that the Reduced-Dimension Model (with an accuracy of 2.528×10-6 °C) has higher accuracy compared to the other two models. While the kernel-based model, is superior to the other two models in terms of steady-state temperature data consumption (only 400 data of size 8 × 8).

Keywords

Soft Computing Journal

Articles in Press, Accepted Manuscript
Available Online from 08 September 2022

Files

History

  • Receive Date: 26 March 2022
  • Revise Date: 04 August 2022
  • Accept Date: 08 September 2022

Share

How to cite

Statistics