Research Papers: Natural and Mixed Convection

Prediction of Local Heat Transfer in a Vertical Cavity Using Artificial Neutral Networks

[+] Author and Article Information
M. Ebrahim Poulad1

Department of Mechanical and Industrial Engineering, Ryerson University, 350 Victoria Street, Toronto, ON, M5B 2K3, Canadampoulad@ryerson.ca

D. Naylor, A. S. Fung

Department of Mechanical and Industrial Engineering, Ryerson University, 350 Victoria Street, Toronto, ON, M5B 2K3, Canada

The difference between the actual value of the target column and the corresponding network output.


Corresponding author.

J. Heat Transfer 132(12), 122501 (Sep 22, 2010) (9 pages) doi:10.1115/1.4002327 History: Received January 07, 2010; Revised June 28, 2010; Published September 22, 2010; Online September 22, 2010

A time-averaging technique was developed to measure the unsteady and turbulent free convection heat transfer in a tall vertical enclosure using a Mach–Zehnder interferometer. The method used a combination of a digital high speed camera and an interferometer to obtain the local time-averaged heat flux in the cavity. The measured values were used to train an artificial neural network (ANN) algorithm to predict the local heat transfer. The time-averaged local Nusselt number is needed to study local phenomena, e.g., condensation in windows. Optical heat transfer measurements were made in a differentially heated vertical cavity with isothermal walls. The cavity widths were W=12.7mm, 32.3 mm, 40 mm, and 56.2 mm. The corresponding Rayleigh numbers were about 3×103, 5×104, 1×105, and 2.7×105, respectively, and the enclosure aspect ratio (H/W) ranged from A=18 to 76. The test fluid was air and the temperature differential was about 15 K for all measurements. ALYUDA NEUROINTELLIGENCE (version 2.2) was used to generate solutions for the time-averaged local Nusselt number in the cavity based on the experimental data. Feed-forward architecture and training by the Levenberg–Marquardt algorithm were adopted. The ANN was designed to suit the present system, which had 4–13 inputs and one output. The network predictions were found to be in a good agreement with the experimental local Nusselt number values.

Copyright © 2010 by American Society of Mechanical Engineers
Your Session has timed out. Please sign back in to continue.



Grahic Jump Location
Figure 9

Performance of different networks

Grahic Jump Location
Figure 1

Experimental geometry and coordinate system

Grahic Jump Location
Figure 2

Steps in the current experiment to calculate Nusselt number

Grahic Jump Location
Figure 3

The high speed video camera mounted on a x,y,z positioning stage that is equipped with micrometers

Grahic Jump Location
Figure 4

The structure of the feed-forward ANN

Grahic Jump Location
Figure 5

Process map of the ANN to predict the local Nusselt number using NEUROINTELLIGENCE

Grahic Jump Location
Figure 6

Effects of structure on R2 and the correlation

Grahic Jump Location
Figure 7

Scatter plot of the training network B4

Grahic Jump Location
Figure 8

Local Nusselt number distribution predicted by network B4 compared with the experimental data




Some tools below are only available to our subscribers or users with an online account.

Related Content

Customize your page view by dragging and repositioning the boxes below.

Related Journal Articles
Related eBook Content
Topic Collections

Sorry! You do not have access to this content. For assistance or to subscribe, please contact us:

  • TELEPHONE: 1-800-843-2763 (Toll-free in the USA)
  • EMAIL: asmedigitalcollection@asme.org
Sign In