Predicting the Self-Diffusion Coefficient of Liquids Based on Backpropagation Artificial Neural Network: A Quantitative Structure–Property Relationship Study

Zeng, Fazhan; Ren, Wan; Xiao, Yongjun; Song, Fan; Peng, Changjun; Liu, Honglai

doi:10.1021/acs.iecr.2c03342

Cited by 8 publications

(6 citation statements)

References 64 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The σ-profile binning integral interval length of 0.5 e/nm 2 is standard in the literature, and we have merged all zero sigma fractions into a single fraction (for example: S1, S6, S7, and S12). Similar σ-profile integration intervals were clearly described elsewhere. − The explicit interactions between cations and anions are not calculated due to the high computational cost. We calculated a binned probability of polarized charge at the molecular surface (i.e., the COSMO-RS-derived sigma profile) that we hypothesized is likely to implicitly capture the propensity for certain intermolecular interactions, either among anion or cation molecules or between an anion and cation.…”

Section: Methodsmentioning

confidence: 98%

Quantum Chemistry-Driven Machine Learning Approach for the Prediction of the Surface Tension and Speed of Sound in Ionic Liquids

Mohan

Smith

Demerdash

et al. 2023

ACS Sustainable Chem. Eng.

View full text Add to dashboard Cite

Ionic liquids (ILs) have unique solvent properties and have thus garnered significant interest. However, exhaustive experimental determination of the physicochemical properties of ILs is unrealistic due to the large structural diversity of anions and cations, their high cost, the requirements of elevated temperature and pressure, and the time required. To circumvent these experimental costs, computational approaches to accurately calculate these properties have emerged. In the present study, we present a demonstration of two machine learning (ML) models for the prediction of two critical IL physical properties, the surface tension and the speed of sound, across a wide range of temperatures and pressures. The models make use of molecular descriptors derived from the COSMO-RS, a quantum chemical-based model. The ML models show excellent agreement with experimental observations, with an R 2 value of 0.96–0.99 and RMSE of 1.71 mN/m and 16.12 m/s for the surface tension and speed of sound, respectively. This work paves the way for the development of COSMO-RS-informed ML models for the prediction of IL properties which can help to further optimize and accelerate technology development for ILs.

show abstract

Section: Methodsmentioning

confidence: 98%

Quantum Chemistry-Driven Machine Learning Approach for the Prediction of the Surface Tension and Speed of Sound in Ionic Liquids

Mohan

Smith

Demerdash

et al. 2023

ACS Sustainable Chem. Eng.

View full text Add to dashboard Cite

show abstract

“…To determine the σ-profile input descriptors for the ML models, as exemplified in Figure , the σ-profiles of HBA and HBD constituents were divided into 10 bins (i.e., S1–S10), each obtained by integrating the σ-profile p x (σ) curves over a discrete interval of σ. The σ-profile binning integral interval length of 0.5 e/nm 2 is standard in the literature, and we have merged all zero-σ fractions into a single fraction. ,,, For example, the area of most HBA and HBD regions has a value of 0 from −3.0 to −2.5 e/nm 2 and from 2.5 to 3 e/nm 2 . Therefore, these regions are merged with those of −2.5 to −2.0 e/nm 2 and 2.0 to 2.5 e/nm 2 , respectively, and labeled as S1 and S10, and all other interval lengths are 0.5 e/nm 2 .…”

Section: Methodsmentioning

confidence: 99%

“…The σ-profile binning integral interval length of 0.5 e/nm 2 is standard in the literature, and we have merged all zero-σ fractions into a single fraction. 26,28,43,44 For example, the area of most HBA and HBD regions has a value of 0 from −3.0 to −2.5 e/nm 2 and from 2.5 The molecular polarity is graphically represented by the colors blue and red, where blue is the negative screening charge density (i.e., "hydrogen bond donating capability"), and red is the positive screening charge density (i.e., "hydrogen bond accepting capability"). The green and yellow color regions characterize "neutral or nonpolar" molecular surfaces.…”

Section: Calculation Of Cosmo-rs-derived Input Features Formentioning

confidence: 99%

Accurate Machine Learning for Predicting the Viscosities of Deep Eutectic Solvents

Mohan,

Jetti,

Smith

et al. 2024

J. Chem. Theory Comput.

View full text Add to dashboard Cite

Deep eutectic solvents (DESs) are emerging as environmentally friendly designer solvents for mass transport and heat transfer processes in industrial applications; however, the lack of accurate tools to predict and thus control their viscosities under both a range of environmental factors and formulations hinders their general application. While DESs may serve as designer solvents, with nearly unlimited combinations, this unfortunately makes it experimentally infeasible to comprehensively measure the viscosities of all DESs of potential industrial interest. To assist in the design of DESs, we have developed several new machine learning (ML) models that accurately and rapidly predict the viscosities of a diverse group of DESs at different temperatures and molar ratios using, to date, one of the most comprehensive data sets containing the properties of over 670 DESs over a wide range of temperatures (278.15−385.25 K). Three ML models, including support vector regression (SVR), feed forward neural networks (FFNNs), and categorical boosting (CatBoost), were developed to predict DES viscosity as a function of temperature and molar ratio and contrasted with multilinear and two-factor polynomial regression baselines. Quantum chemistry-based, COSMO-RS-derived sigma profile (σ-profile) features were used as inputs for the ML models. The CatBoost model is excellent at externally predicting DES viscosity, as indicated by high R 2 (0.99) and low root-mean-square-error (RMSE) and average absolute relative deviations (AARD) (5.22%) values for the testing data sets, and 98% of the data points lie within the 15% of AARD deviations. Furthermore, SHapley additive explanation (SHAP) analysis was employed to interpret the ML results and rationalize the viscosity predictions. The result is an ML approach that accurately predicts viscosity and will aid in accelerating the design of appropriate DESs for industrial applications.

show abstract

“…However, if one of the conductivities is significantly different from the others, it will be excluded. The exact approach can be found in our previous work . The resulting refined data set for electrical conductivity contained 3102 data points of 254 ILs, covering 139 distinct cations and 46 anions.…”

Section: Dataset and Methodologymentioning

confidence: 99%

QSPR Models Based on Multiple Linear Regression and Back-Propagation Artificial Neural Network for Predicting Electrical Conductivity of Ionic Liquids

Shi,

Song,

et al. 2024

Ind. Eng. Chem. Res.

Self Cite

View full text Add to dashboard Cite

Electrical conductivity of ionic liquids (ILs), a crucial transfer property, has been investigated using various methods, including experimental measurements, semiempirical models, and molecular simulations. Among these methods, the quantitative structure−property relationship (QSPR) model is extensively utilized in diverse applications. However, constructing an effective QSPR model depends on the selection of the appropriate molecular descriptors. In this study, linear and nonlinear QSPR models were developed to predict the conductivity of the ILs. These models incorporated temperature as well as the molecular descriptors of cavity volume (V COSMO ) and charge-density-distribution area at specific intervals (S σi ) obtained from the conductor-like screening model for segment activity coefficient (COSMO-SAC). The evaluated data set comprised 254 ILs with 3102 data points; the ILs were considered to consist of "pure ions" or "ion pairs". Three distinct models were developed in this study: model I represented the pure ion model combined with multiple linear regression (MLR), model II referred to the ion pair model combined with MLR, and model III was constructed by considering ILs as ion pairs and using the back-propagation artificial neural network (BP-ANN). For the entire data set, models I and II exhibited coefficients of determination (R 2 ) of 0.9478 and 0.9642, respectively. Furthermore, the calculated average absolute relative deviations (AARDs) were 25.38 and 21.37% with root mean square errors (RMSEs) of 0.3291 and 0.2724, respectively. By contrast, model III demonstrated R 2 values of 0.9939, 0.9911, and 0.9887 for the training, validation, and testing sets, respectively. The corresponding AARD values were 6.96, 8.13, and 8.97%, while the RMSE values were 0.1132, 0.1214, and 0.1569, respectively. Application domain (AD) analysis revealed that 91.1% of the data points lie within the AD of model II and only 0.16% of the data showed response outliers. In contrast, 93.42% of the data were bound within the AD of model III, with no response outliers. These findings indicate that the QSPR model, which treats ILs as "ion pairs" and combines them with the BP-ANN, can predict the conductivity of diverse ILs at various temperatures with a high accuracy.

show abstract

Predicting the Self-Diffusion Coefficient of Liquids Based on Backpropagation Artificial Neural Network: A Quantitative Structure–Property Relationship Study

Cited by 8 publications

References 64 publications

Quantum Chemistry-Driven Machine Learning Approach for the Prediction of the Surface Tension and Speed of Sound in Ionic Liquids

Quantum Chemistry-Driven Machine Learning Approach for the Prediction of the Surface Tension and Speed of Sound in Ionic Liquids

Accurate Machine Learning for Predicting the Viscosities of Deep Eutectic Solvents

QSPR Models Based on Multiple Linear Regression and Back-Propagation Artificial Neural Network for Predicting Electrical Conductivity of Ionic Liquids

Contact Info

Product

Resources

About