A new comprehensive model of thermal conductivity for hydrofluoroolefins refrigerants using feed-forward back-propagation neural networks
N. Ghalem1,2, S. Hanini1, A. Amrane3, M. Hamadache1, M. Laidi1, M.W. Naceur2
1Medea University, Medea, Algeria
2Blida University, Blida, Algeria
3Rennes University, Rennes, France
Keywords: refrigerant, pure system, thermal conductivity, modeling, MLP-ANN, predictive model, artificial neural network
Abstract
In this work, the thermal conductivity of refrigerants systems from three different hydrofluoroolefins including R1234yf, R1234ze (E), and R1233zd(E) were studied using artificial neural network. A total of 4395 data points of liquid and vapor thermal conductivity at several temperatures (241.92 to 344.46) K and pressures (0.068 to 21.73) MPa were used to train and test the model. Five neurons were used in the input layer, fifteen neurons at hidden layer and one was used in the output layer. Bayesian Regulation back propagation algorithm, logarithmic sigmoid transfer function, and linear transfer function were used at the hidden and output layer, respectively. Temperature, pressure, applied heating power; acentric factor and dipole moment were considered as input variables of the networks. The optimal parameters were obtained through the weights searching method. The average absolute relative deviations and correlation coefficient were 1.48 and 0.9998, respectively. This study shows therefore that the artificial neural network model represents an excellent alternative to estimate the thermal conductivity of different refrigerant systems with a good accuracy.
|
