AN ASYMPTOTIC APPROACH TO ESTIMATE THE THERMAL CONDUCTIVITY OF A SQUARE PLATE UNDER NATURAL CONVECTION

Thermal Engineering

The present work is to develop a reduced order model, using Artificial Neural Networks (ANN) based inverse technique on a transient heat transfer problem. The proposed methodology has been demonstrated on a two- dimensional heat slab to estimate the unknown parameters such as thermal conductivity (k). A 2D model of the material (Teflon) is developed and encapsulated with the fluid domain, is simulated using CFD. To minimize the influence of mesh size on the results, grid independence study is carried out. The input to the CFD model is heat flux and thermal conductivity varying between 400 ≤ q ≤ 600 W/m2 and 0.1 ≤ k ≤ 0.6 W/mK respectively, the output is temperature of the Teflon plate. The input-output data set from CFD simulation are used to train the ANN network, which act as a proxy model. The neural network used in this work is based on the feed forward back propagation algorithm. The optimum number of neurons required for the neural network is determined by neuron independence study. The experimental data are input to the inverse model (ANN) to estimate the thermal conductivity of the Teflon plate for an unknown heat flux. The method can be used for different applications and provides a good accuracy on the result.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.