Thermal Conductivity Prediction of Pure Liquids Using Multi-Layer Perceptron Neural Network

Abstract

Widespread application of a neural network has been obvious in many fields of chemical engineering over the last years. The thermal conductivity of liquids has been predicted by using this kind of network and compared with experimental outcomes. Heat transfer of fluids is important in many industrial sectors, including energy supply, transportation, production, and electronics. To model the heat transfer process, thermal conductivity data are required. Using the Bayesian regularization method, the network parameters are adjusted with the aim of minimizing the sum of the squared errors and the sum of the squared weights. Changing the number of neurons in the hidden layers iteratively, the optimum performance for the network was obtained. Using the test dataset including 124 data points that were not previously used for the network training, the performance of the developed network with one hidden layer containing 15 neurons was evaluated.