Current Practices and Continuing Needs in Thermophysical Properties for the Chemical Industry

Abstract

The status of thermophysical property needs of the chemical industry is reviewed and updated relative to similar observations from 20 years ago. The paper is informed by a series of symposia held over several years in conjunction with the American Institute of Chemical Engineers (AIChE) national meetings. Experiences of the authors are also incorporated, including a discussion of the state of the art in this area, as well as references to several of the articles included in a recent special issue of Ind. Eng. Chem. Res. (2022, volume 61, issue 42) devoted to the subject. In general, the trend is toward more rigorous molecular methods but ingrained empirical methods tend to hold on for extended times by adding increasingly sophisticated multiparameter correlations. There is also a tendency for research in newer methods to end prematurely with anecdotal proofs of principle, undermining their ability to supersede tried and true methods. Significant gaps exist in experimental data, for the development of estimation methods and validation of models besides a general need for technical knowledge development. Although progress is clear, some of the goals articulated 20 years ago remain to be achieved, even as new needs are identified in estimation. modeling, and measurements. One possible solution, to close experimental data gaps and to provide a continuous stream of trained personnel, is to establish multidisciplinary research centers of excellence in this important methodology.

Multifidelity Gaussian Processes for Predicting Shear Viscosity over Wide Ranges of Liquid State Points Based on Molecular Dynamics Simulations

“…14−17 Other approaches, e.g., those based on corresponding states 18−20 are often limited to simple fluids of only moderate polarity or require accurate experimental data for a reference compound. 21 Here, we show that the need for an ansatz function can be avoided if simulation results of reasonable accuracy are available and if regression is carried out by Gaussian processes (GPs), a method often used in machine learning for classification and regression tasks. The almost linear relationship between the reduced shear viscosity and the reduced residual entropy obtained from MD simulation in conjunction with the PC-SAFT equation of state serves as low-fidelity data in a multifidelity model 22−26 based on GP regression 27 and carries over to the high-fidelity level such that a linear multifidelity model as introduced by Kennedy and O'Hagan 28 can be used in such a way that it does not suffer from the limited extrapolation ability of stochastic processes.…”

“…An issue with the entropy scaling approach, when correlated onto experimental data, is the extrapolation quality of the model function used to describe the univariate relation between the reduced transport property and reduced residual entropy. The often used third-order polynomial function shows good interpolation quality but insufficient extrapolation behavior. − Other approaches, e.g., those based on corresponding states − are often limited to simple fluids of only moderate polarity or require accurate experimental data for a reference compound …”

Reaction mechanism – explored with the unified reaction valley approach

“…Valuable overall insights into chemical reactions can be obtained from measured and/or computed kinetic data such as reaction rates or yields, or thermo-physical data including thermodynamic and transport properties. [1][2][3] The efficiency and accuracy of experimental instrumentation and computer hardware/software have improved significantly over time, offering today data with increasing accuracy, often collected in large databases, 4,5 along with improved statistical methodologies and advanced software packages for data analysis. [6][7][8] Also artificial intelligence (AI) has entered the scene, where algorithms learn to create and predict reaction outcomes, such as reaction rates, intermediates, products, and yields.…”

Predicting PC‐SAFT parameters based on COSMO‐RS

“…Thermodynamic properties influence process design and optimization, which are important for chemical engineering. [1][2][3][4] There are two ways to obtain thermodynamic properties, that is, experimental determinations and theoretical predictions. Considering that experiments are time-consuming and costly, it is desirable to develop theoretical models to predict thermodynamic properties in a wide temperature and pressure range.…”

“…Thermodynamic properties influence process design and optimization, which are important for chemical engineering 1–4 . There are two ways to obtain thermodynamic properties, that is, experimental determinations and theoretical predictions.…”