Quantitative Structure-Property Relationship Studies On The Thermal Conductivity Of Liquid Organic Compounds

Thermal conductivity is the basic physical property of a material that characterizes the heat transfer ability of a substance.Generally,the greater the thermal conductivity of a substance,the better the heat transfer ability.Based on the accurate thermal conductivity,the heat transferred or exchanged by the heat transfer medium can be calculated accurately.Therefore,thermal conductivity is the basic data required in engineering fields,such as chemical engineering,petroleum,and energy.Accurate thermal conductivity data can make the heat balance in the production process more accurate,improve the comprehensive utilization of energy,and reduce the costs of production.Organic compounds are widely used in daily life with huge quantity.Organic compounds act as important raw materials and intermediates in chemical production.Meanwhile,some organic compounds have the characteristics of low freezing point,high boiling point,and high thermal conductivity which are widely used in heating,heat preservation,heat transfer,heat dissipation,refrigeration and other projects.Under this circumstance,thermal conductivity data provide significant reference to find a suitable heat transfer medium.Therefore,the accurate measurement and calculation of the thermal conductivity of organic compounds is of theoretical and practical importace.At present,many experimental methods and empirical formulas have been used to obtain thermal conductivity data of liquid organic compounds.However,their limitations of the experimental methods and empirical formulas estimation leaded to considerable errors of the obtained thermal conductivity data.Therefore,how to accurately predict the thermal conductivity of liquid organic compounds is of practical significance.In this study,quantitative structure-property relationship(QSPR)method was used to explore the molecular structure characteristics that affect the thermal conductivity of liquid hydrocarbons,halogenated alkanes,aliphatic oxygen compounds,and nitrogen-containing organic compounds,based on the optimized molecular structure of organic compounds.The thermal conductivity estimation models of the diverse organic compounds were developed,of which the predicting results were both great.The main research works are as follows:(1)560 thermal conductivity data of 56 liquid hydrocarbons(alkanes,olefins and alkynes)compounds at different temperatures were ollected.Using the Kennard Stone algorithm,the dataset was divided into a 392-member training set and a 168-member test.Calculation of molecular descriptors was performed based on the optimized molecular structure and the genetic function approximation method(GFA)was used to screen parameters.Based on the training set,with multiple linear regression method,a 5-parameter QSPR model for predicting the thermal conductivity of liquid hydrocarbon compounds was developed.The squared correlation coefficients(R~2)of the training set and test set were0.9574 and 0.9745,respectively.And the average absolute relative deviations(AARD%)were3.37%and 4.10%,respectively.The predictive ability,model validation and applicability domain test demonstrated the reliablility and robustness of the developed model.From analysising the physical meaning of model parameters,it can be found that the temperature of environment,molecular weight of hydrocarbon compounds,the multiplicity of different bond lengths and the structural complexity of each vertex of the molecule have great effects on the thermal conductivity of liquid hydrocarbon compounds.(2)410 thermal conductivity data of 37 liquid alkyl halides compounds(alkanes partially or completely substituted by fluorine,chlorine,bromine,and iodine atoms)at different temperatures were ollected.Using the Kennard Stone algorithm,the dataset was divided into a307-member training set and a 103-member test.The molecular descriptors calculation was performed based on the optimized molecular structure and the GFA method was used to perform parameters screening.Based on the training set,with MLR method,a 5-parameter QSPR model for predicting the thermal conductivity of liquid alkyl halides was developed.The R~2 of the training set and test set were 0.9808 and 0.9745,respectively.And the AARD%values were 3.37%and 4.10%,respectively.From analysising the physical meaning of model parameters,it can be found that fluorine interatomic interaction,temperature,molecular vibration and rotation,molecular shape,the size and distribution of charge and electron intensity scatteration have great effects on the thermal conductivity of liquid alkyl halides.(3)992 thermal conductivity data of 112 kinds of liquid aliphatic oxygen-containing organic compounds(alcohol,ether,aldehyde,ketone,acid and ester)at different temperatures were ollected.Using the Kennard Stone algorithm,the dataset was divided into 694-member training set and a 298-member test set.In order to further prove the predictive ability of the developed model,72 experimental values were measured experimentally as the prediction set.The molecular descriptors calculation was performed based on the optimized molecular structure and the GFA method was used to screen parameters.A multivariate linear regression QSPR model with 6 parameters was established to predict the thermal conductivity of various aliphatic and oxygen-containing organic compounds.The R~2 of the training set,test set and prediction set were 0.9138,0.8922 and 0.8816,respectively.And the AARD%values were4.14%,4.41%and 4.16%,respectively.From the physical meaning analysis of model parameters,it can be found that molecule group interaction,molecular branches,number of oxygen atoms,temperature and main chain of molecules have great effects on the thermal conductivity of aliphatic oxygen-containing organic compounds.(4)142 thermal conductivity data of 22 nitrogen-containing liquid organic compounds at different temperatures were collected.The dataset was divided into a 99-member training set and a 43-member test set with the Kennard Stone algorithm.Molecular descriptors calculation was performed based on the optimized molecular structure,and parameter selection was performed using the best subset regression method.A multivariate linear regression QSPR model with 5 parameters was established to predict the thermal conductivity of liquid nitrogen-containing organic compounds.The R~2 of the training set and test set were 0.9650and 0.9420,respectively.And the AARD%values were 5.99%and 6.33%,respectively.From analysising the physical meaning of model parameters,it can be found that the existence of nitrogen atoms,the electron scattering intensity,the topological distance between nitrogen and oxygen atoms,and the temperature have great effects on the thermal conductivity of nitrogen-containing liquid organic compounds.The developed models have good predictive ability and robustness.The present study is of great significance not only by providing a robust model for predicting thermal conductivity of diverse liquid organic compounds,but also by shedding light on other thermodynamic data estimation.