The adoption of lattice- or hard-sphere-based equations of state (EoS) has become a common and trusted approach for analysing the behavior of polymers and their mixtures, playing a crucial role in the design of their manufacturing, application, and end-of-life processes. While EoS offer reliable and comprehensive descriptions of polymer phases, their parameters must first be retrieved using data from pure polymers. This requirement limits their predictive ability, particularly for polymers that cannot be tested within certain operative ranges or that have not been created yet. Over the last 25 years, various strategies have been developed to overcome these limitations, including the use of group contribution methods, molecular simulations, and machine learning techniques. This paper evaluates the most successful experimental and computational methods for determining parameters in polymer EoS, emphasizing the work of Professor Maurizio Fermeglia and Professor Sabrina Pricl in promoting multiscale methods based on atomistic simulations [1]. The analysis starts with the experimental methods utilized to gather data on pure polymers and progresses to exploring techniques involving Molecular Dynamics (MD), which enable direct parameter estimation through molecular definitions or the generation of synthetic experimental data. Further, it discusses methodologies that derive parameters purely from the chemical structure of the polymer’s monomer units, specifically the Group Contribution (GC) methods. Notably, machine learning has emerged as a powerful asset in these procedures, enabling the computation of EoS parameters directly from the structural features of monomers, as in the case of the recent Machine Learning Based Group Contribution Method For The Non-Equilibrium Lattice Fluid Model (ML-GC-NELF) and a physics-informed neural network for the Perturbed Chain Statistical Associating Fluid Theory (PC-SAFT) EoS, referred to as MLPSAFT.
Ismaeel, H., Ricci, E., De Angelis, M.G. (2026). Parameterisation of equations of state for polymers: From experimental approaches to machine learning. FLUID PHASE EQUILIBRIA, 114754, 1-24 [10.1016/j.fluid.2026.114754].
Parameterisation of equations of state for polymers: From experimental approaches to machine learning
De Angelis, Maria Grazia
2026
Abstract
The adoption of lattice- or hard-sphere-based equations of state (EoS) has become a common and trusted approach for analysing the behavior of polymers and their mixtures, playing a crucial role in the design of their manufacturing, application, and end-of-life processes. While EoS offer reliable and comprehensive descriptions of polymer phases, their parameters must first be retrieved using data from pure polymers. This requirement limits their predictive ability, particularly for polymers that cannot be tested within certain operative ranges or that have not been created yet. Over the last 25 years, various strategies have been developed to overcome these limitations, including the use of group contribution methods, molecular simulations, and machine learning techniques. This paper evaluates the most successful experimental and computational methods for determining parameters in polymer EoS, emphasizing the work of Professor Maurizio Fermeglia and Professor Sabrina Pricl in promoting multiscale methods based on atomistic simulations [1]. The analysis starts with the experimental methods utilized to gather data on pure polymers and progresses to exploring techniques involving Molecular Dynamics (MD), which enable direct parameter estimation through molecular definitions or the generation of synthetic experimental data. Further, it discusses methodologies that derive parameters purely from the chemical structure of the polymer’s monomer units, specifically the Group Contribution (GC) methods. Notably, machine learning has emerged as a powerful asset in these procedures, enabling the computation of EoS parameters directly from the structural features of monomers, as in the case of the recent Machine Learning Based Group Contribution Method For The Non-Equilibrium Lattice Fluid Model (ML-GC-NELF) and a physics-informed neural network for the Perturbed Chain Statistical Associating Fluid Theory (PC-SAFT) EoS, referred to as MLPSAFT.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
