Predicting Diffusion Coefficients of Binary and Ternary Supercritical Water Mixtures via Machine and Transfer Learning with Deep Neural Network

Abstract

Prediction for diffusion coefficients of multicomponent supercritical water (SCW) mixtures is crucial for the system design and industrial application of SCW-related technologies, such as SCW gasification and oxidation. In this work, machine learning (ML) and transfer learning (TL) techniques with deep neural network (DNN) are explored to predict diffusion coefficients of binary and ternary SCW mixtures. Initially, diffusion coefficients are calculated through molecular dynamics (MD) simulations. Then, the structure of DNN is found and examined with a cross-validation method so that an accurate predictive ML model is trained with our database for diffusion coefficients of binary SCW mixtures. Finally, TL is used to train a DNN model for diffusion coefficients of ternary mixtures where the knowledge learned from the pretrained DNN model for binary mixtures is transferred to improve the model performance with minimal training data. Compared to the model trained from scratch (non-TL), the TL model reduces the mean squared error by 89% and proves its possibility for industrial and engineering applications. It indicates the advantage of ML and TL methods over empirical equations when the priority is to predict diffusion coefficients of multicomponent mixtures without complex and tedious MD simulations.

Review of Machine Learning for Hydrodynamics, Transport, and Reactions in Multiphase Flows and Reactors

“…In addition to the submodel closures discussed above, some other closure coefficients or parameters such as the lift force coefficients, , diffusion coefficients, interfacial area, − curvature, , and droplet/bubble size , should also be informed in advance in Eulerian–Eulerian (E-E), E-L, or VOF simulations of gas–liquid, gas–solid, and gas–liquid–solid flows. Despite the importance of the lift coefficient calculation in TFM simulations of gas–liquid flows, few reports have used ML to aid in direct data-driven modeling of lift force in such flows.…”

Machine Learning Predictions of Simulated Self-Diffusion Coefficients for Bulk and Confined Pure Liquids

“…48 Transfer learning has been used to predict diffusion coefficients in binary and ternary water mixtures. 49 We seek to build on these efforts by developing a generalized predictive model, which can be applicable to a wide range of liquids, pore sizes, and pore interactions.…”

Adversarial Transferred Data-Assisted Soft Sensor for Enhanced Multigrade Quality Prediction

“…As a branch of machine learning, transfer learning (TL) aims to tackle the problem of insufficient training data between domains. − By leveraging knowledge from the related domain having relatively sufficient annotated samples, the predicted target domain having limited labeled samples is expanded. Recently, with good feature extraction capability, deep transfer learning methods − attempt to transfer knowledge. To address the problem of model construction of multigrade processes with limited samples, deep transfer learning-based soft sensors have been developed.…”