SPT-NRTL: A physics-guided machine learning model to predict thermodynamically consistent activity coefficients

Capturing molecular interactions in graph neural networks: a case study in multi-component phase equilibrium

“…To conduct a fair comparison, we applied the same training/validation/testing method described in their research 29 through ensemble learning (bagging), which splits the training/validation data randomly 30 times and averages the predictions. We also used the same batch size (32) and epoch number (200). Although the logarithmic values were used to train and validate the models, Medina et al 29 calculated the evaluation metrics on the unscaled g N values.…”

“…13 Recently, there has been growing interest in applying ML models to study more complex chemical systems that might contain multiple components such as chemical reactions, 14,15 alloys, 16,17 copolymers, [18][19][20] and gas/ liquid mixtures. [21][22][23][24][25][26][27][28][29][30][31][32][33] Among the ML techniques explored, graph neural networks (GNNs) 34,35 have gained special popularity because they can directly incorporate molecular representations (in the form of graphs), which enable the capturing of key structural information while potentially avoiding the need to pre-calculate/pre-dene descriptors using more advanced but computationally-intensive tools such density functional theory (DFT) or molecular dynamics (MD) models.…”

Outlook: How I Learned to Love Machine Learning (A Personal Perspective on Machine Learning in Process Systems Engineering)

“…For instance, work in molecular discovery is driving huge advances in ML algorithms and tools that are being developed by math and computer science communities . What the PSE community needs to do is identify “killer apps” that are of broad interest and that can lead to major ML development by diverse communities; for instance, some recent applications that the PSE community has explored are the prediction of battery lifetimes, prediction of thermodynamic properties, and characterization of plastic waste. − Along these lines, I think that there is significant potential in using ML tools to extract/digitize data from the literature and to assemble databases that can be used for benchmarking . Assembling such databases is nontrivial, because it requires domain-specific knowledge (you need to know what you are looking for).…”

Predicting Temperature‐Dependent Activity Coefficients at Infinite Dilution Using Tensor Completion

“…Besides physical prediction methods, machine learning (ML) methods for predicting thermodynamic properties are gaining more and more attention [8]. In recent years, several ML methods, such as deep neural networks [9], graph neural networks [10][11][12], and transformer models [13,14], have been applied in this field. For the prediction of thermodynamic properties of binary mixtures, matrix completion methods (MCMs) are of particular interest.…”