A Transfer Learning Study of Gas Adsorption in Metal-Organic Frameworks

Ma, Ruimin; Colón, Yamil J.; Luo, Tengfei

doi:10.26434/chemrxiv.12111552

Cited by 12 publications

(11 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Machine learning plays an important role in the discovery and deployment of NPMs [26][27][28][29][30][31][32]. Supervised machine learning models have been widely used to predict the adsorption properties of NPMs [33][34][35][36][37][38][39][40][41] from vectors of hand-crafted structural features [42,43] or from a graph representation [44]. Unsupervised machine learning methods have been used to embed NPMs into a lowdimensional "material space" [45] and cluster together NPMs with similar structures [46][47][48].…”

Section: Review Of Previous Workmentioning

confidence: 99%

Recommendation System to Predict Missing Adsorption Properties of Nanoporous Materials

et al. 2021

View full text Add to dashboard Cite

Nanoporous materials (NPMs) selectively adsorb and concentrate gases into their pores, and thus could be used to store, capture, and sense many different gases. Modularly synthesized classes of NPMs, such as covalent organic frameworks (COFs), offer a large number of candidate structures for each adsorption task. A complete NPM-property table, containing measurements of the relevant adsorption properties in the candidate NPMs, would enable the matching of NPMs with adsorption tasks. However, in practice the NPM-property matrix is only partially observed (incomplete); (i) many properties of any given NPM have not been measured and (ii) any given property has not been measured for all NPMs.The idea in this work is to leverage the observed (NPM, property) values to impute the missing ones. Similarly, commercial recommendation systems impute missing entries in an incomplete product-customer ratings matrix to recommend products to customers. We demonstrate a COF recommendation system to match COFs with adsorption tasks by training a low rank model of an incomplete COF-adsorption-property matrix. A low rank model, trained on the observed (COF, adsorption property) values, provides (i) predictions of the missing (COF, adsorption property) values and (ii) a "map" of COFs, wherein COFs, represented as points, with similar (dissimilar) adsorption properties congregate (separate). We find the performance of the COF recommendation system varies for different adsorption tasks and diminishes precipitously when the fraction of missing entries exceeds 60 %. The concepts in our COF recommendation system can be applied broadly to many different materials and properties.

show abstract

Section: Review Of Previous Workmentioning

confidence: 99%

Recommendation System to Predict Missing Adsorption Properties of Nanoporous Materials

et al. 2021

View full text Add to dashboard Cite

show abstract

“…They also introduced the TL technique to improve the accuracy and applicability of ML models trained with a small amount of gas absorption data in metal−organic frameworks. 64 In this work, we first build a database (1220 data) of selfdiffusion and Fick diffusion coefficients of binary and ternary SCW mixtures calculated from equilibrium MD simulations. Details of mole fractions and compositions about mixtures are listed in the Supporting Information.…”

Section: Introductionmentioning

confidence: 99%

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

Zhao

Luo

Jin

2022

Ind. Eng. Chem. Res.

Self Cite

View full text Add to dashboard Cite

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.

show abstract

“…Thus, most recently, studies using "transfer learning" and developing "multipurpose" models are emerging. [19][20][21][22] Anderson et al 19 showed that an ANN could predict H2 adsorption in MOFs at 𝑇, 𝑃 values different from those included in the training data. Sun et al demonstrated a similar capability for an ANN to predict sorption of pentanediol/water on zeolite MFI, and also illustrated the reuse of some ANN layers on a new model that predicts sorption when changing either the adsorbent or one of the adsorbates.…”

Section: Introductionmentioning

confidence: 99%

“…22 Ma et al reused some layers of an ANN trained to predict H2 sorption in a new ANN trained to predict CH4 sorption. 20 Using genetic algorithm regression and adsorption data for multiple molecules in a subset of previously synthesized MOFs, Gharagheizi et al obtained an equation capable of predicting isotherms for a diversity of CxHySuOvNw molecules. 21 These authors used this equation to screen MOFs for many chemical separations.…”

Section: Introductionmentioning

confidence: 99%

Deep Learning Combined with IAST to Screen Thermodynamically Feasible MOFs for Adsorption-Based Separation of Multiple Binary Mixtures

Anderson¹,

Gómez-Gualdrón²

2021

Preprint

View full text Add to dashboard Cite

This study demonstrates the coupling of a multipurpose multilayer perceptron (MLP) model that predicts single-component adsorption for a various molecules with ideal adsorption solution theory (IAST). The resulting computational framework predicts MOF separations properties for various binary mixtures at various compositions and pressures. The accuracy of the MLP+IAST framework was sufficiently high so that, for a given separation, MOFs in the 90th percentile from MLP+IAST-based screening contain ~87% of MOFs in the 95th percentile one would obtain from molecular simulation-based screening. Clustering algorithms were shown effective to identify so-called "privileged" MOFs that were high-performing for multiple separations. Free energy calculations were performed to determine privileged MOFs that were likely to be accesses synthetically, at least from a thermodynamic perspective.

show abstract

A Transfer Learning Study of Gas Adsorption in Metal-Organic Frameworks

Cited by 12 publications

References 0 publications

Recommendation System to Predict Missing Adsorption Properties of Nanoporous Materials

Recommendation System to Predict Missing Adsorption Properties of Nanoporous Materials

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

Deep Learning Combined with IAST to Screen Thermodynamically Feasible MOFs for Adsorption-Based Separation of Multiple Binary Mixtures

Contact Info

Product

Resources

About