This work presents a literature survey of the available experimental data regarding the thermal conductivity of organic liquids. Experimental data are regressed with the most reliable semiempirical correlating methods existing in the literature, and a set of 5010 data are finally selected, belonging to 164 compounds in the following families: bromide derivatives, chlorine derivatives, condensed rings, fluorine+chlorine+bromidefluorine+chlorine+bromide derivatives, F-derivatives, hydrocarbon chains, monocyclic compounds, carboxylic acids, cycloalkanes, cycloalkenes, esters, and ketones. A new correlation to represent the thermal conductivity of pure liquids is presented. A factor analysis is performed for the data selected in order to select the physical parameters to adopt. Optimal coefficients and a different version of the recently proposed equation are presented. The correlation is very simple and is able to predict the thermal conductivity with very low deviation for all families studied. The correlation reproduces the selected data with an average absolute deviation of 7.5%. The same physical parameters considered in the equations are also adopted as input parameters in a multilayer perceptron neural network to predict the thermal conductivity. The multilayer perceptron proposed has one hidden layer with 39 neurons, which were determined according to the constructive approach. The model developed is trained, validated, and tested for the set of data collected, showing that the accuracy of the neural network model is very high and confirming the validity of the parameters selected for the proposed correlation. The artificial neural network reproduces the selected data with an average absolute deviation of 3.5%.

Equation for the Thermal Conductivity of Liquids and an Artificial Neural Network

Petrucci, Giulio;
2016-01-01

Abstract

This work presents a literature survey of the available experimental data regarding the thermal conductivity of organic liquids. Experimental data are regressed with the most reliable semiempirical correlating methods existing in the literature, and a set of 5010 data are finally selected, belonging to 164 compounds in the following families: bromide derivatives, chlorine derivatives, condensed rings, fluorine+chlorine+bromidefluorine+chlorine+bromide derivatives, F-derivatives, hydrocarbon chains, monocyclic compounds, carboxylic acids, cycloalkanes, cycloalkenes, esters, and ketones. A new correlation to represent the thermal conductivity of pure liquids is presented. A factor analysis is performed for the data selected in order to select the physical parameters to adopt. Optimal coefficients and a different version of the recently proposed equation are presented. The correlation is very simple and is able to predict the thermal conductivity with very low deviation for all families studied. The correlation reproduces the selected data with an average absolute deviation of 7.5%. The same physical parameters considered in the equations are also adopted as input parameters in a multilayer perceptron neural network to predict the thermal conductivity. The multilayer perceptron proposed has one hidden layer with 39 neurons, which were determined according to the constructive approach. The model developed is trained, validated, and tested for the set of data collected, showing that the accuracy of the neural network model is very high and confirming the validity of the parameters selected for the proposed correlation. The artificial neural network reproduces the selected data with an average absolute deviation of 3.5%.
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11582/307061
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
social impact