Prediction Of Thermophysical Properties Research Articles

Prediction of Saturation Densities and Critical Properties of n-Decane, n-Pentadecane, and n-Eicosane Using Molecular Dynamics with Different Force-Fields.

Prediction of thermophysical properties of heavy hydrocarbons is important because of the recent increased interest in the extraction of heavy and shale oil to meet the global energy demand. Carrying out experimental work, such as determining the critical properties of heavy hydrocarbons, is challenging due to the possibility of thermal degradation during experimentation. This study focuses on the use of molecular simulations, specifically canonical molecular dynamics, to predict the critical properties of three hydrocarbons: n-decane (n-C10), n-pentadecane (n-C15), and n-eicosane (n-C20). The method uses volume-expansion molecular dynamics (VEMD), where a single box is enlarged in one axis and the vapor-liquid equilibrium is achieved. Three different force-fields (AMBER, COMPASS, and TraPPE) were employed to compare the accuracy of the simulated results and the required computational time. The results from the simulations were compared with available experimental data, equations-of-state, and several correlations. The results indicate that TraPPE is the most accurate and efficient force-field to predict the critical properties followed by AMBER and COMPASS.

Thermophysical property prediction of biodiesel mixtures at extreme conditions using molecular dynamics simulation

Liquid biofuels such as biodiesel are playing an increasing role in renewable energy utilisation, but accurate predictions of fuel properties at extreme conditions remain challenging. In this study, molecular dynamics (MD) simulation with classical force field is performed to obtain the thermophysical properties of biodiesel over extended ranges of temperatures and pressures. The predicted properties include the critical temperature, critical density, critical pressure, surface tension, viscosity, thermal conductivity and diffusion coefficient. The specific MD setup for fuel thermophysical property prediction is examined. It is observed that the long-range dispersion interaction is essential to accurately predicting the surface tension, and averaging over a sufficient number of independent replication simulations is required to eliminate the statistical error when using the Green-Kubo method for viscosity calculation. The reliability of MD approach is validated against experimental data at normal temperature and pressure. The equilibrium MD simulation has an advantage over nonequilibrium MD simulation when building databases of fuel properties. Moreover, the properties of biodiesel are compared with conventional diesel to elucidate the effects of fuel composition and molecular structure of the fuel surrogates on fuel thermophysical properties. For biodiesel utilisation, higher values of critical temperature, surface tension and viscosity, lower diffusivity together with the increased aggregation tendency in bath gas indicate the needs of further optimisation of injection system to achieve the desired mixing in combustion applications.

Prediction of thermophysical properties of chlorine eutectic salts via artificial neural network combined with polar bear optimization

The eutectic salt as latent heat storage (LHS) media has been widely employed in various fields, but its thermophysical properties are difficult to be predicted accurately. Here, a prediction model of thermophysical properties of eutectic salts based on the backpropagation (BP) artificial neural network method combined with bio-inspired algorithms (polar bear optimization (PBO) or genetic algorithm (GA)) is proposed. The average prediction error of melting temperature of 26 chlorine eutectic salts is the lowest by employing the BP-PBO algorithm, which is reduced by 42 % and 38 % compared to the BP algorithm and BP-GA, respectively. The advantage of BP-PBO is more obvious in predicting phase change enthalpy of chlorine eutectic salts, whose average error decreases by 71 % and 68 % compared with BP and BP-GA, respectively. For more intuitive comparisons, experiment measurements, BP-PBO, phase diagram, and molecular dynamics are used for characterizing thermophysical properties of fabricated KCl-LiCl eutectics. Again, BP-PBO demonstrates extremely accurate prediction performance and unique advantage of avoiding the difficulty of choosing appropriate phase equilibrium data or potential functions. The work provides accurate and fast prediction methods of thermophysical properties of chlorine eutectic salts and guides their designs for different applications.

Simulation-assisted AI for the evaluation of thermal barrier coatings using pulsed infrared thermography

The development of predictive models for the accurate estimation of thermo-physical properties of the Thermal Barrier Coated (TBC) aero-engine components is critical in assessing component life and maintenance. TBCs are multi-layer systems applied on metallic structures operating at higher temperatures, such as aero-engine parts and gas turbine blades. These thermally insulating materials prolong the component life by limiting the thermal exposure of structural components. In this study, simulation-assisted Artificial Intelligence (AI) is developed to predict thermal conductivity (k), heat capacity (ρCp), and thickness measurement of TBC from thermal responses of samples with varying topcoat layer thicknesses. The dataset used in the AI model is a low-fidelity thermal profile from a multi-layer heat transfer model of the TBC system for training the neural network and high-fidelity thermogram from pulsed thermography experiments that are used for validation of the trained neural network. The proposed method demonstrated potential in the prediction of thermo-physical properties for real samples with a newly coated topcoat layer of thickness measurement varying from 24 to 120 μm, with a mean absolute percentage error (MAPE) for k and ρCp predictions of 1.71% and 1.37%, respectively, and for thickness prediction, MAPE ranges from 0.81% to 6.14%. This work explores the possibilities of merging a large set of low-fidelity simulation data and a small set of high-fidelity experimental data to train the deep neural network to achieve promising results in real-world thermography experiments.

Prediction of Thermophysical Properties of Spherical Top Molecules

Prediction of Thermophysical Properties of Spherical Top Molecules

Comparative evaluation of AI ‐based intelligent GEP and ANFIS models in prediction of thermophysical properties of Fe 3 O 4 ‐coated MWCNT hybrid nanofluids for potential application in energy systems

Hybrid nanofluids are gaining popularity owing to the synergistic effects of nanoparticles, which provide them with better heat transfer capabilities than base fluids and normal nanofluids. The thermophysical characteristics of hybrid nanofluids are critical in shaping heat transmission properties. As a result, before using thermophysical qualities in industrial applications, an in-depth investigation of thermophysical properties is required. In this paper, a metamodel framework is constructed to forecast the effect of nanofluid temperature and concentration on numerous thermophysical parameters of Fe3O4-coated MWCNT hybrid nanofluids. Evolutionary gene expression programming (GEP) and an adaptive neural fuzzy inference system (ANFIS) were employed to develop the prediction models. The model was trained using 70% of the datasets, with the remaining 15% used for testing and validation. A variety of statistical measurements and Taylor's diagrams were used to assess the proposed models. The Pearson's correlation coefficient (R), coefficient of determination (R2) was used for the regression index, the error in the model was evaluated with root mean squared error (RMSE). The model's comprehensive assessment additionally includes modern model efficiency indices such as Kling-Gupta efficiency (KGE) and Nash-Sutcliffe efficiency (NSCE). The proposed models demonstrated impressive prediction capabilities. However, the GEP model (R > 0.9825, R2 > 0.9654, RMSE = 0.7929, KGE > 0.9188, and NSCE > 0.9566) outperformed the ANFIS model (R > 0.9601, R2 > 0.9218, RMSE = 1.495, KGE > 0.8015, and NSCE > 0.8745) for the majority of the findings. The generated metamodel was robust enough to replace the repetitive expensive lab procedures required to measure thermophysical properties. Highlights Predictions of thermophysical properties of Fe3O4-coated MWCNT hybrid nanofluid AI-based ANFIS and GEP models performed well on statistical indices ANFIS and GEP-based prognostic models validated and compared with Taylor diagrams

Estimation of thermophysical property of hybrid nanofluids for solar Thermal applications: Implementation of novel Optimizable Gaussian Process regression (O-GPR) approach for Viscosity prediction.

Solar energy technologies represent a viable alternative to fossil fuels for meeting increasing global energy demands. However, to increase the production of solar technologies in the global energy mix, the cost of production should be as competitive as other sources. This study focuses on the implementation of machine learning for estimating the thermophysical properties of nanofluids for nanofluid-based solar energy technologies as this would make the synthesis of nanofluids cost-effective. The prediction of thermal conductivity has gained a lot of research attention, whereas, the viscosity of nanofluids has less concentration of studies. The accurate prediction of the viscosity of hybrid nanofluids is important in estimating the heat transfer performance of nanofluids as regards their pump power requirements and convective heat transfer coefficient in several applications. The rigor of experimentations of hybrid nanofluids has necessitated the need for developing efficient and robust machine learning models for accurately estimating the viscosity of hybrid nanofluids for solar applications. Several studies were aimed at developing a predictive model for the viscosity of nanofluids; however, these models are limited to specific types of nanofluids. This study is aimed at developing a robust machine learning algorithm for predicting the viscosity of several hybrid nanofluids from reliable experimental data (700 datasets) culled from literature. This study implements a novel optimizable Gaussian process regression (O-GPR), which have not been previously used in this area, and compares the result with other commonly used machine learning algorithms like, Boosted tree regression (BTR), Artificial neural network (ANN), support vector regression (SVR), to accurately predict the viscosity of a wide range of Newtonian-based hybrid nanofluid. The input parameters used in training the machine learning models were temperature (T), volume fraction (VF), the acentric factor of the base fluid (ACF), nanoparticle size (NS), and nanoparticle density (ND). The prediction performance of the machine learning algorithms was tested using statistical metrics and was compared with theoretical models. The O-GPR model showed superior predictive performance with an R2 of 0.999998 and an MSE of 0.0002552. The study conclusively states that the high accuracy prediction of thermophysical properties of nanofluid using robust machine learning models makes the design of nanofluid-based solar energy technologies more cost-effective.

Screening of important parameters in optimal design of compressed air energy storage system using an ensemble learning method

Accurate prediction of thermophysical properties of compressed air is specifically crucial in optimal design and analyzing performance of a Compressed Air Energy Storage (CAES) system. In this regard, an advanced machine learning method, namely stochastic gradient boosting (SGBTree) technique was examined based on reported data points in the literature for estimation of specific heat ratio of air, water content of atmospheric air, water content of compressed saturated air, viscosity of compressed air as well as thermal conductivity of compressed air. The used ensemble learning method precisely functions for a wide range of temperature, pressure and relative humidity. The SGBTree models estimate the aforementioned target values well, matching to those available in the literature as the correlation coefficient (R) and Mean Relative Absolute Error (MRAE) values are greater than 0.999 and less than 0.007, respectively. Moreover, a comparison was carried out between the outputs of the SGBTree model and two intelligent models namely Artificial Neural Network (ANN) and Least Squares Support Vector Machine (LSSVM) schemes as well as a reported simple empirical correlation for estimating specific heat ratio of air. In terms of accuracy, it is found that the SGBTree model is better than other methods for estimation of specific heat ratio of air.

Prediction of nanoscale thermal transport and adsorption of liquid containing surfactant at solid–liquid interface via deep learning

HypothesisRecent advances in deep learning (DL) have enabled high level of real-time prediction of thermophysical properties of materials. On the other hand, molecular dynamics (MD) have been long used as a numerical microscope to observe detailed interfacial conditions but require separate simulations that are computationally costly. Hence, it should be possible to combine MD and DL to obtain high resolution interfacial details at a low computational cost. ExperimentWe proposed a novel DL encoding–decoding convolutional neural network (CNN) coupled with MD to realize the mapping from micro solid–liquid interface geometry to molecular temperature and density distribution of liquid containing surfactant. A multi-nanoscale optimization scheme was further proposed to reduce the uncertainty of DL prediction at the expense of local details to obtain more resilient predictors. FindingsThe statistical results showed that the proposed CNN had high prediction accuracy and could reproduce the heat transfer and adsorption phenomena under the influence of various factors including liquid composition, wettability, and solid surface roughness, while the computational efficiency was greatly improved. Our DL method with the support of multi-nanoscale learning strategies can achieve the fast and accurate visualization and prediction of various interfacial properties of liquid and assist for interfacial material design.

A novel method overcomeing overfitting of artificial neural network for accurate prediction: Application on thermophysical property of natural gas

As a powerful tool to solve nonlinear problems, artificial neural network method (ANN) gets a wide range of applications in data regression. However, the overfitting often occurs during the ANN training process, which results in high accuracy of correlating the training data but poor prediction performance. At the same time, the principle of k-Nearest Neighbor method (kNN) makes it impossible to make an accurate prediction exceeding the range of the training data, but it can confine the overfitting of ANN. In this work, combining ANN and kNN, a new machine learning method called ANN-kNN combination method (AKC) for thermophysical property prediction of material is proposed. To evaluate the performance of AKC, we take the thermophysical properties of natural gas as an example. The inputs of AKC are temperature, pressure and the components of natural gas, the outputs are the compressibility factor, speed of sound and viscosity. AKC not only overcomes the overfitting problem but also needs less training data than ANN. The average absolute relative deviation of AKC for prediction are 2.5%, which are better than ANN (5.9%) and kNN (19.2%).

High pressure viscosity measurements of ternary (methane + propane + heptane) mixtures

Prediction of thermophysical properties for natural gas mixtures close to the mixture's critical point and cricondenbar (the maximum pressure at which two fluid phases can co-exist) is challenging but important as often natural gas processes operate close to these conditions. Here we present a comprehensive study of the viscosity of (CH4 + C3H8 + C7H16) mixtures with heptane fractions up to 15 mol%. Accurate measurements of saturated-phase and compressed-fluid (single phase) viscosities, at temperatures from 200 K to 423 K and pressures up to 35 MPa, are presented. The collected data were compared with the prediction of the extended corresponding states (ECS) model, with relative deviations spanning between (−20 to 11) % from the model. Additionally, pure component parameters needed for the residual entropy scaling (RES) model were fit to literature data and then applied predictively to mixtures viscosity without any additional fit parameters. The relative deviation of the measured viscosities from values calculated with the RES model is generally within 15%. This work demonstrates that no current model satisfactorily predict mixtures viscosity, and that high quality experimental data are needed to anchor the methods used to design and optimize natural gas processing.

Natural gas density measurements and the impact of accuracy on process design

The liquefaction of natural gas is an energy intensive process, requiring at least 5% of the energy associated with methane's lower heating value. Key to estimating and optimizing these energy requirements are process simulations which rely upon calculated thermophysical properties of the natural gas. In particular, the prediction of thermophysical properties of natural gas mixtures at pressure-temperature conditions close to the mixture’s critical point or cricondenbar is challenging but important as often natural gas processes operate close to these conditions. In this work, we present a comprehensive study of two natural gas related systems: (CH4 + C3H8 + CO2) and (CH4 + C3H8 + C7H16) with n-heptane fractions up to 15 mol%. High accuracy measurements of densities, at temperatures from 200 K to 423 K and pressures up to 35 MPa are presented. The extensive experimental data collected for these mixtures were compared with the GERG-2008 equation of state, as implemented in the NIST software REFPROP. The relative deviations of the measured densities from those calculated using the GERG-2008 model range between (−2 to 4)% for all mixtures, presenting a systematic dependent on mixture density and n-heptane content. Finally, a case study is presented that probes the impact of the accuracy of density on the pinch point in a simulated LNG heat exchanger. An uncertainty in the density of 1% is shown to cause significant 30% reduction in the minimum approach temperature difference, suggesting that accurate thermophysical property calculations are key to reducing over-design of processing plant.

Experimental Study on Thermophysical Properties of Propylene Glycol-Based Graphene Nanofluids

This study determined the optimal dispersion conditions of nanofluids by orthogonal experiments. Stable graphene/propylene glycol nanofluids are successfully prepared by a "two-step method" using a 60 wt % propylene glycol deionized water solution as the base fluid, and its thermal conductivity and viscosity are evaluated. Additionally, the effects of the graphene mass concentration and nanofluid temperature on thermal conductivity and viscosity are expounded, and a model for the prediction of thermophysical properties is established. The combined effect of thermal conductivity and viscosity on the heat transfer enhancement of nanofluids is analyzed theoretically, and the optimum mass concentration of graphene is determined. The results show that the thermal conductivity of nanofluids increases with the increase of the graphene concentration and nanofluid temperature. Also, the viscosity of nanofluid increases with the increase of the graphene concentration and decreases significantly with the increase of the nanofluid temperature. When the mass concentration of graphene is higher than 0.2 wt %, the increase in thermal conductivity caused by the increase of the graphene mass concentration decreases significantly, while the increase of viscosity increases significantly, and the effect of the heat transfer enhancement is almost not improved.

Benchmark assessment of molecular geometries and energies from small molecule force fields

Background: Force fields are used in a wide variety of contexts for classical molecular simulation, including studies on protein-ligand binding, membrane permeation, and thermophysical property prediction. The quality of these studies relies on the quality of the force fields used to represent the systems. Methods: Focusing on small molecules of fewer than 50 heavy atoms, our aim in this work is to compare nine force fields: GAFF, GAFF2, MMFF94, MMFF94S, OPLS3e, SMIRNOFF99Frosst, and the Open Force Field Parsley, versions 1.0, 1.1, and 1.2. On a dataset comprising 22,675 molecular structures of 3,271 molecules, we analyzed force field-optimized geometries and conformer energies compared to reference quantum mechanical (QM) data. Results: We show that while OPLS3e performs best, the latest Open Force Field Parsley release is approaching a comparable level of accuracy in reproducing QM geometries and energetics for this set of molecules. Meanwhile, the performance of established force fields such as MMFF94S and GAFF2 is generally somewhat worse. We also find that the series of recent Open Force Field versions provide significant increases in accuracy. Conclusions: This study provides an extensive test of the performance of different molecular mechanics force fields on a diverse molecule set, and highlights two (OPLS3e and OpenFF 1.2) that perform better than the others tested on the present comparison. Our molecule set and results are available for other researchers to use in testing.

Benchmark assessment of molecular geometries and energies from small molecule force fields.

Background: Force fields are used in a wide variety of contexts for classical molecular simulation, including studies on protein-ligand binding, membrane permeation, and thermophysical property prediction. The quality of these studies relies on the quality of the force fields used to represent the systems. Methods: Focusing on small molecules of fewer than 50 heavy atoms, our aim in this work is to compare nine force fields: GAFF, GAFF2, MMFF94, MMFF94S, OPLS3e, SMIRNOFF99Frosst, and the Open Force Field Parsley, versions 1.0, 1.1, and 1.2. On a dataset comprising 22,675 molecular structures of 3,271 molecules, we analyzed force field-optimized geometries and conformer energies compared to reference quantum mechanical (QM) data. Results: We show that while OPLS3e performs best, the latest Open Force Field Parsley release is approaching a comparable level of accuracy in reproducing QM geometries and energetics for this set of molecules. Meanwhile, the performance of established force fields such as MMFF94S and GAFF2 is generally somewhat worse. We also find that the series of recent Open Force Field versions provide significant increases in accuracy. Conclusions: This study provides an extensive test of the performance of different molecular mechanics force fields on a diverse molecule set, and highlights two (OPLS3e and OpenFF 1.2) that perform better than the others tested on the present comparison. Our molecule set and results are available for other researchers to use in testing.

Prediction of thermo-physical properties of 1-Butyl-3-methylimidazolium hexafluorophosphate for CO2 capture using machine learning models

Physical and thermodynamic properties of physical or chemical solvents are of utmost importance for mass and heat transfer calculations, process design and solvent regeneration. In recent times, machine learning has attracted interest for applications in several fields of engineering sciences. The ionic liquid 1-Butyl-3-methylimidazolium hexafluorophosphate [Bmim][PF6] is an emerging solvent for CO2 capture. In this study, three Gaussian process regression (GPR) models - the Matern 5/2 GPR model, rational quadratic GPR model, squared exponential GPR model - and one support vector machine (SVM) model (the nonlinear SVM)– are developed for predicting CO2 solubility, density, viscosity and molar heat capacity of [Bmim][PF6]. Detailed statistics of each model and comparative analyses between the models and their predicted results with experimental results is highlighted.

Recent trends on nanofluid heat transfer machine learning research applied to renewable energy

Nanofluids have received increasing attention in research and development in the area of renewable and sustainable energy systems. The addition of a small amount of high thermal conductivity solid nanoparticles could improve the thermophysical properties of a base fluid and lead to heat transfer augmentation. Various enhancement mechanisms and flow conditions result in nonlinear effects on the thermodynamics, heat transfer, fluid flow, and thermo-optical performance of nanofluids. A large amount of research data have been reported in the literature, yet some contradictory results exist. Many affecting factors as well as the nonlinearity and refutations make nanofluid research very complicated and impede its potentially practical applications. Nonetheless, machine learning methods would be essentially useful in nanofluid research concerning the prediction of thermophysical properties, the evaluation of thermo-hydrodynamic performance, and the radiative-optical performance applied to heat exchangers and solar energy systems. The present review aims at revealing the recent trends of machine learning research in nanofluids and scrutinizing the features and applicability of various machine learning methods. The potentials and challenges of machine learning approaches for nanofluid heat transfer research in renewable and sustainable energy systems are discussed. According to the Web of Science database, about 3% of nanofluid research papers published in 2019 involved in machine learning and such a tendency is increasing.

Influence of equations of state and mixture models on the design of a refrigeration process

In refrigeration technology, there is a need for alternative refrigerants that need to fulfill several requirements, such as low (or no) global warming potential, flammability, ozone depletion potential, and toxicity. While there are many substances available that could potentially be used as refrigerants, only relatively few meet all of the requirements. Pure alternative refrigerants have been studied quite extensively in the past. However, mixtures (blends) of refrigerants are not that well-studied. One of the reasons for this is that the prerequisite for an extensive screening of refrigerant mixtures is the availability of reliable predictive mixture models. In this work, we investigate typically used mixture models for a simple refrigeration cycle (split air conditioner with internal heat exchanger) and compare their predictive performance with the results of reference multi-fluid mixture models for three binary mixtures with different mixing behavior, i.e., R-410a, R-421a, and R-507a. We found that regarding coefficient of performance (COP), heat exchanger performance parameter, and volumetric cooling capacity, the largest differences between the models have been observed in the prediction of the heat exchanger performance parameter. For the studied systems, the COP and heat exchanger performance parameter are mostly influenced by the pure fluid equations of state used, while the volumetric cooling capacity depends more on the choice of the mixture model. Furthermore, some models yield results for some of the studied performance parameters that are in very good agreement with the reference model, although the prediction of thermophysical properties deviates significantly from the reference model.

Generating a Machine-Learned Equation of State for Fluid Properties.

Equations of state (EoS) for fluids have been a staple of engineering design and practice for over a century. Available EoS are based on the fitting of a closed-form analytical expression to suitable experimental data. The mathematical structure and the underlying physical model significantly restrain the applicability and accuracy of the resulting EoS. This contribution explores the issues surrounding the substitution of machine-learned models for analytical EoS. In particular, we describe, as a proof of concept, the effectiveness of a machine-learned model to replicate the statistical associating fluid theory (SAFT-VR Mie) EoS for pure fluids. To quantify the effectiveness of machine-learning techniques, a large set of pseudodata is obtained from the EoS and used to train the machine-learning models. We employ artificial neural networks and Gaussian process regression to correlate and predict thermodynamic properties such as critical pressure and temperature, vapor pressures, and densities of pure model fluids; these are performed on the basis of molecular descriptors. The comparisons between the machine-learned EoS and the surrogate data set suggest that the proposed approach shows promise as a viable technique for the correlation and prediction of thermophysical properties of fluids.

Prediction of Thermophysical Properties of Binary Liquid Mixtures of Propiophenone and Ethyl Acetate at 303.15, 308.15, and 313.15 K

Density, viscosity, ultrasonic velocity and refractive index values of the binary liquid mixtures of propiophenone (PP) and ethyl acetate (EA) have been measured for the whole range of compositions at temperatures 303.15, 308.15, and 313.15 K. The variation of excess properties (VE, Δη, Δu, Δn, ΔZ, ΔβS, ΔLf, and G*E) with composition and temperature is discussed in the paper in terms of molecular interaction of components in the mixture. Redlich–Kister type polynomial has been used to fit the excess properties. A standard RSM design based on central composite design (CCD) has been applied to study the thermo physical properties of binary liquid mixtures as function of mole fraction and temperature. The molecular interaction between PP and EA has been analyzed by the excess properties along with the acoustic properties, and the developed equation can be used to calculate the thermophysical properties at various temperatures and compositions.