Next generation pure component property estimation models: With and without machine learning techniques

Abstract

Physiochemical properties of pure components serve as the basis for the design and simulation of chemical products and processes. Models based on the molecular structural information of chemicals for the following 25 pure component properties are presented in this work: (critical‐) temperature, pressure, volume, acentric factor; (normal‐) boiling point, melting point, auto‐ignition temperature; flash point; (standard‐) enthalpy of formation, Gibbs energy of formation, enthalpy of fusion, enthalpy of vaporization, liquid molar volume; (environmental‐) (lethal dose‐) LC50 and LD50, photo‐chemical oxidation potential, bioconcentration factor, permissible exposure limit; (physicochemical‐) acid dissociation constant, water‐solubility, octanol–water partition coefficient, Hildebrandt solubility parameter, Hansen solubility parameters. Utilizing functional groups for molecular representation, two parallel property estimation models where the group contributions for each property are regressed through traditional regression techniques and machine learning techniques are presented. Both techniques use an a priori data analysis before regression of model parameters. A dataset with more than 24,000 chemicals for the 25 pure component properties has been utilized for the development of the two sets of property models. The efficacy of the developed models and their use are highlighted together with a discussion on the overall performance, application range, and predictive capabilities with implications to product and/or process engineering problem solutions.

Molecular Design of Fuels for Maximum Spark-Ignition Engine Efficiency by Combining Predictive Thermodynamics and Machine Learning

“…The models achieve an accuracy comparable to the models in the literature , (see Table ). The accuracy of the predictions on the test sets measured by the coefficient of determination ( R 2 ) equals on average R 2 = 0.73.…”

“…A similar approach was presented by Li et al for the prediction of sooting tendencies. We follow the approaches by Alshehri et al and Li et al and consider the following EHS indicators through GC-based GPR prediction as constraints for the fuel design: Autoignition temperature (AiT) Bioconcentration factor (BCF) Aqueous toxicity as a lethal concentration for fathead minnow fish (LC 50 (FM)) Oral toxicity as lethal dose for rats (LD 50 ) Permissible exposure limit using the OSHA time-weighted average (PEL OSHA‑TWA ) Chemical tendency to form soot expressed through the unified yield sooting index (uYSI) For integration in the fuel design method, the models by Alshehri et al for AiT, BCF, LC 50 (FM), LD 50 , and PEL OSHA‑TWA are retrained using UNIFAC groups as descriptors and the training and test data from Alshehri et al The uYSI model is developed using the data from McEnally et al Note that the uYSI does not predict engine-out soot emissions but rather the chemical tendency of a fuel to form soot. A more practical measure for engine-out soot emissions would be the Particulate Matter Index (PMI), where the number of double bond equivalents as a proxy for the chemical tendency to form soot is divided by the vapor pressure as a measure for in-cylinder mixture formation quality.…”

A smile is all you need: predicting limiting activity coefficients from SMILES with natural language processing

“… 11 The application of machine learning has also already led to recent advances in thermodynamic property prediction. Alshehri et al 12 developed a data-driven model to predict 25 pure component properties based on a Gaussian process. The developed model surpasses classical group contribution models in accuracy.…”

Machine Learning for Physicochemical Property Prediction of Complex Hydrocarbon Mixtures

“…Over the last years, machine learning has emerged as an alternative tool for predicting molecular properties, because of its speed and application range . First, it has especially been applied to predict molecular quantum chemical properties in theoretical chemistry studies. − In chemical engineering, mostly neural network-based approaches have been developed for a wide range of properties, such as but not limited to enthalpies of formation, − solvation energies, , octane numbers, , boiling points, − and vapor pressure. , Alshehri et al published the most extensive study so far that applies both group contribution and machine learning methods to predict 25 pure compound properties. Their data set contains around 25000 organic molecules up to 30 heavy atoms that can contain 9 heteroatoms.…”