Python-Based Algorithm for Estimating NRTL Model Parameters with UNIFAC Model Simulation Results

• Published in: ACS omega Vol. 10; no. 3; pp. 2949 - 2957
• Main Authors: Jo, Se-Hee, Lee, Jina, Won, Wangyun, Kim, Jun-Woo
• Format: Journal Article
• Language: English
• Published: United States American Chemical Society 28.01.2025
• ISSN: 2470-1343; 2470-1343
• DOI: 10.1021/acsomega.4c09246

A major challenge in bioprocess simulation is the lack of physical and chemical property databases for biochemicals. A Python-based algorithm was developed for estimating the nonrandom two-liquid (NRTL) model parameters of aqueous binary systems in a straightforward manner from simplified molecular-input line-entry specification (SMILES) strings of substances in a system. This algorithm conducts a series of procedures: (1) fragmentation of the molecules into functional groups from SMILES, (2) calculation of activity coefficients under predetermined temperature and mole fraction conditions by employing universal quasi-chemical functional group activity coefficient (UNIFAC) model, and (3) regression of NRTL model parameters by employing UNIFAC model simulation results in the differential evolution algorithm (DEA) and Nelder–Mead method (NMM). The algorithm was applied to aqueous, binary mixture systems composed of 37 common biochemical substances such as amino acids, organic acids, and sugars. The obtained NRTL parameters were compared with those from Aspen Plus, a commercial software, which has an equivalent function for estimating the NRTL parameters. The percentage mean absolute residuals of the activity coefficients obtained using DEA, NMM, and the parameter estimation tool in Aspen Plus were in the ranges of 0.05–16.69, 0.05–16.69, and 0.09–326.77%, respectively. This in-house algorithm will be helpful for obtaining more accurate NRTL parameters in a timely manner and will facilitate the simulation of biochemical processes for process optimization, energy consumption estimation, and life cycle assessment.