Prediction of thermal conductivities of dry granular media using artificial neural networks
Piotr Furmański , Grabarczyk Marcin
AbstractThermal conductivity of heterogeneous materials is a complex function not only of properties and amounts of constituents but also of many structural factors. Therefore it is difficult to predict. An attempt to predict thermal conductivity of granular media using the Artificial Neural Network (ANN) model is undertaken in the paper. It was assumed that it is a function of a ratio of thermal conductivities of the constituents, medium porosity as well as the coordination number describing the mean number of the nearest neighbours to each grain. Several configurations of the ANNs were tested while developing the optimal model. As a measure of prediction accuracy the coefficient of linear regression and the mean squared error were used. The optimal model of ANN was found to consist of three hidden layers with eight neurons in each layer for both types of media. Some problems associated with application of ANN were pointed out. The predicted values of thermal conductivity obtained with ANN were compared with values calculated from an analytical formula. It was found that the ANN predictions show identical trends and similar values as the analytical formula for all factors affecting thermal conductivities of the granular media.
|Journal series||Journal of Power Technologies, ISSN 2083-4187|
|Keywords in English||granular materials; thermal conductivity; prediction; artificial neural network|
|Score|| = 10.0, 05-09-2019, ArticleFromJournal|
= 12.0, 05-09-2019, ArticleFromJournal
|Publication indicators||: 2015 = 0|
* presented citation count is obtained through Internet information analysis and it is close to the number calculated by the Publish or Perish system.