Advancing Nitrobenzene Hydrogenation Process Understanding: A Multiscale Modeling Approach using Industrial Pharmaceutical Production Data

• Published in: Industrial & engineering chemistry research Vol. 63; no. 32; pp. 14029 - 14042
• Main Authors: Callewaert, Wout, Pessanha, Bernardo V., Lauwaert, Jeroen, Fernandes del Pozo, David, Nopens, Ingmar, Baldwin, Peter, McNamara, Mairtin, Thybaut, Joris W.
• Format: Journal Article
• Language: English
• Published: American Chemical Society 14.08.2024
• Subjects: Kinetics, Catalysis, and Reaction Engineering
• ISSN: 0888-5885; 1520-5045
• DOI: 10.1021/acs.iecr.4c00318

The pharmaceutical industry is under increasing pressure to reduce production costs and operate within agile supply chains by leveraging the capabilities that come with integrated data infrastructures and increasing access to various forms of modeling. In contrast to bulk chemical production, pharmaceutical productions typically operate at significantly lower volumes and allow for narrower variability in critical parameters due to tightly defined operating ranges. This makes the data acquired from the process challenging to analyze with statistical methods, but the application of engineering and scientific relations as multiscale models offers a more effective way of leveraging the historical data that are available. In this work, a multiscale reaction model is developed for an exothermic liquid phase hydrogenation of a nitrobenzene functionality in the synthesis of an active pharmaceutical ingredient (API) by using available production data. The developed model successfully described the interplay between reaction kinetics, gas–liquid mass transfer, and heat removal present in the process data, as evident from the simulated versus observed temperature evolution with batch time, which was used as an indirect measurement of the reaction conversion. Moreover, the model was also able to reproduce the temperature profiles in the case of a 30% scale-up. Simulated concentration profiles indicate that the end of the reaction occurs within a much shorter time frame than that prescribed by the process recipe, suggesting that the batch time can be reduced by more than 50%. The results demonstrate the flexibility and predictive power of this type of modeling approach because this model was developed using passive data collection of standard process parameters rather than through a dedicated Design of Experiments (DoE).