Data science for thermodynamic modeling: Case study for ionic liquid and hydrofluorocarbon refrigerant mixtures

Befort, Bridgette; Garciadiego, Alejandro R.; Wang, Jialu; Wang, Ke; Franco, Gabriela; Maginn, Edward J.; Dowling, Alexander W.

doi:10.1016/j.fluid.2023.113833

Cited by 7 publications

(2 citation statements)

References 112 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Recent advances in machine learning (ML) have significantly influenced a wide range of scientific and engineering fields with applications to agricultural sciences, thermofluidic processes, − computational fluid dynamics − (CFD), as well as varieties of other complex industrial applications . More recently, physics-informed neural networks (PINNs) are being increasingly investigated due to their capability in building models that obey desired physics constraints, with implementations ranging from solving systems of partial differential equations − to modeling different chemical processes such as photochemical systems, hydrofluorocarbon refrigerant mixtures, biomass pyrolysis process, heat transfer problems, , etc. In the literature cited before, the physics conservation equations have been either incorporated in the objective (loss) function as additional penalty terms to penalize the violation of physics constraint(s) or imposed by imposing the conservation constraints in an additional layer followed by the development of the data-driven model.…”

Section: Introductionmentioning

confidence: 99%

Development of Steady-State and Dynamic Mass and Energy Constrained Neural Networks for Distributed Chemical Systems Using Noisy Transient Data

Mukherjee,

Bhattacharyya

2024

Ind. Eng. Chem. Res.

View full text Add to dashboard Cite

The paper presents the development of algorithms for mass and energy constrained neural network models that can exactly conserve the overall mass and energy of distributed chemical process systems, even though the noisy transient data used for optimal model training violate the same. In contrast to approximately satisfying mass and energy balance constraints of a system by soft penalization of objective function, algorithms have been developed for solving equality-constrained nonlinear optimization problems, thus providing the guarantee of exactly satisfying the system mass and energy conservation laws. For developing dynamic mass-energy constrained network models for distributed systems, hybrid series and parallel dynamic-static neural networks have been leveraged. The developed algorithms for solving both the training and forward problems are validated using both steady-state and dynamic data in the presence of various noise characteristics. The developed data-driven algorithms are flexible to exactly satisfy mass and energy balance constraints for dynamic chemical processes if the system holdup information is available. The proposed network structures and algorithms are applied to the development of data-driven lumped and distributed models of an adiabatic superheater/reheater system, a nonisothermal continuous stirred tank reactor, as well as an electrically heated plug-flow reactor system where one form of energy gets transformed to another. It has been observed that the mass-energy constrained neural networks yield a root mean squared error of <1% with respect to the system truth for the case studies evaluated in this work.

show abstract

Section: Introductionmentioning

confidence: 99%

Development of Steady-State and Dynamic Mass and Energy Constrained Neural Networks for Distributed Chemical Systems Using Noisy Transient Data

Mukherjee,

Bhattacharyya

2024

Ind. Eng. Chem. Res.

View full text Add to dashboard Cite

show abstract

“…MS has emerged as a valuable tool to predict thermodynamic and transport properties for materials of industrial interest and to understand their microscopic origin. MS can aid in the design of novel processes for challenging separations, such as those posed by HFC mixtures. − MS requires accurate representations of interatomic interactions to obtain reliable predictions of properties and elucidation of molecular-level phenomena. These representations, commonly referred to as force fields (FFs), are mathematical models that describe the potential energies and forces between interaction sites (usually atomic nuclei).…”

Section: Introductionmentioning

confidence: 99%

Assessment and Ranking of Difluoromethane (R32) and Pentafluoroethane (R125) Interatomic Potentials Using Several Thermophysical and Transport Properties Across Multiple State Points

Agbodekhe,

Marin-Rimoldi,

Zhang

et al. 2023

J. Chem. Eng. Data

Self Cite

View full text Add to dashboard Cite

Difluoromethane (R32) and pentafluoroethane (R125) are two common hydrofluorocarbon refrigerants, often used in a mixture termed R410A. Many refrigerants, including R32 and especially R125, have high global warming potentials and so are being phased out. There is a desire to develop processes that can separate and recover these materials, which means that there is a need to determine the thermodynamic and transport properties of these fluids. In this work, we evaluate the ability of molecular dynamics simulations to determine the key thermodynamic and transport properties of these two fluids. We test whether classical interatomic force fields (FFs) parametrized against vapor−liquid equilibrium (VLE) data using a machine learning directed (MLD) approach can also yield accurate estimates of other key properties. The top-performing MLD FFs tuned against VLE data were nearly indistinguishable based on VLE results. This work seeks to investigate if these MLD-tuned FFs are transferable to other properties not used in tuning them and if they can be ranked to identify the "best" FFs. Literature FFs, one each for R32 and R125, are included in the study. A total of ten FFs were tested. Thermal conductivity (λ), viscosity (η), self-diffusivity (D), liquid density (ρ), isobaric heat capacity (C P ), isochoric heat capacity (C V ), thermal expansivity (α P ), thermal pressure coefficient (γ ρ ), isothermal compressibility (β T ), speed of sound (c sound ), Joule-Thomson coefficient (μ JT ), and center of mass radial distribution functions (g r ) were computed using molecular dynamics and compared with experiments when possible. Somewhat surprisingly, the MLD-tuned FFs are found to be transferable to a wide range of properties not used in tuning them. The MLD-tuned FFs were ranked. The FFs labeled R32 a and R125 b were found to be the "best" FFs for R32 and R125, respectively, across a broad range of properties. The MLD-tuned FFs were found to be superior to previously developed literature FFs.

show abstract

Sequential optimal experimental design for vapor-liquid equilibrium modeling

Bubel,

Schmid,

Kozachynskyi

et al. 2024

Chemical Engineering Science

View full text Add to dashboard Cite

Data science for thermodynamic modeling: Case study for ionic liquid and hydrofluorocarbon refrigerant mixtures

Cited by 7 publications

References 112 publications

Development of Steady-State and Dynamic Mass and Energy Constrained Neural Networks for Distributed Chemical Systems Using Noisy Transient Data

Development of Steady-State and Dynamic Mass and Energy Constrained Neural Networks for Distributed Chemical Systems Using Noisy Transient Data

Assessment and Ranking of Difluoromethane (R32) and Pentafluoroethane (R125) Interatomic Potentials Using Several Thermophysical and Transport Properties Across Multiple State Points

Sequential optimal experimental design for vapor-liquid equilibrium modeling

Contact Info

Product

Resources

About