Why fixing alpha in the NRTL model might be a bad idea -- Identifiability analysis of a binary Vapor-Liquid equilibrium

Volodymyr Kozachynskyi,Christian Hoffmann,Erik Esche

2024-08-15

Abstract:New vapor-liquid equilibrium (VLE) data are continuously being measured and new parameter values, e.g., for the nonrandom two-liquid (NRTL) model are estimated and published. The parameter $\alpha$, the nonrandomness parameter of NRTL, is often heuristically fixed to a value in the range of 0.1 to 0.47. This can be seen as a manual application of a (subset selection) regularization method. In this work, the identifiability of the NRTL model for describing the VLE is analyzed. It is shown that fixing $\alpha$ is not always a good decision and sometimes leads to worse prediction properties of the final parameter estimates. Popular regularization techniques are compared and Generalized Orthogonalization is proposed as an alternative to this heuristic. In addition, the sequential Optimal Experimental Design and Parameter Estimation (sOED-PE) method is applied to study the influence of the regularization methods on the performance of the sOED-PE loop.

Optimization and Control

What problem does this paper attempt to address?

The paper primarily explores the potential issues that may arise from fixing the parameter \(\alpha\) in the Non-Random Two-Liquid (NRTL) model and proposes an alternative method—Generalized Orthogonalization—to improve the performance of model parameter estimation. Specifically, the paper aims to address the following core issues:

1. **Analyzing the impact of fixing \(\alpha\)**: When describing binary vapor-liquid equilibrium in the NRTL model, it is common practice to empirically fix the parameter \(\alpha\) to a certain value (usually between 0.1 and 0.47). This approach can be seen as a manually applied subset selection regularization method. The paper demonstrates through identifiability analysis that, in some cases, fixing \(\alpha\) may lead to poorer predictive performance of the final parameter estimates.

2. **Comparing regularization techniques**: To overcome the limitations brought by fixing \(\alpha\), the paper compares several popular regularization techniques, including the eigenvalue algorithm, singular value decomposition (SVD) algorithm, forward selection algorithm, etc., and proposes Generalized Orthogonalization as an alternative to fixing \(\alpha\).

3. **Studying the impact of regularization techniques on Sequential Optimal Experimental Design and Parameter Estimation (sOED-PE)**: The paper further utilizes the sOED-PE method to study how these regularization techniques affect the performance of the sOED-PE cycle.

In summary, the core objective of the paper is to conduct an in-depth analysis of the handling of the parameter \(\alpha\) in the NRTL model and its impact on parameter estimation performance, thereby proposing improved strategies and techniques to enhance the accuracy and reliability of vapor-liquid equilibrium predictions in chemical engineering.