Python-Based Algorithm for Calculating Physical Properties of Aqueous Mixtures Composed of Substances Not Available in Databases

• Published in: ACS omega Vol. 10; no. 16; pp. 16683 - 16694
• Main Authors: Lee, Jina, Jo, Se-Hee, Lee, Chungyup, Kang, Ji Hun, Won, Wangyun, Kim, Jun-Woo
• Format: Journal Article
• Language: English
• Published: United States American Chemical Society 29.04.2025
• ISSN: 2470-1343; 2470-1343
• DOI: 10.1021/acsomega.5c00424

In this study, we developed a Python-based open-source algorithm compatible with the aqueous physical property models provided in the electrolyte templates of AspenTech software. To validate the accuracy of the model, the results obtained from the proposed algorithm were compared to experimental data for 37 binary aqueous mixture systems covering properties such as density, heat capacity, viscosity, and thermal conductivity. The input variables included results from our previous research on pure component property prediction and the nonrandom two-liquid (NRTL) model parameters based on the UNIFAC model simulations. This open-source algorithm is compatible with AspenTech software. The mean absolute percentage errors (MAPE) for density, heat capacity, viscosity, and thermal conductivity were 2.88, 0.355, 12.1, and 10.1%, respectively. In the case of density and viscosity, the actual data trends could not be accurately reflected under high-concentration conditions for certain substances. In addition, it was confirmed that inaccurate predictions of the viscosity and thermal conductivity in the commercial-scale falling-film evaporator simulation for l-valine production led to inaccurate predictions of the overall heat transfer coefficient. Therefore, caution is required when predicting missing property parameters using this approach as significant errors may occur. Nevertheless, this algorithm can provide an initial parameter value for property models that are not included in existing databases without any commercial package.