Copertina - Population modeling cristallization
Mer 15 Mag
Seminari e Convegni

Population balance modeling for pharmaceutical crystallization processes: industrial case studies and simulation-based knowledge generation

The seminar titled "Population balance modeling for pharmaceutical crystallization processes: industrial case studies and simulation-based knowledge generation" will feature the presence of professor Botond Szilagyi from Budapest University of Technology and Economics, in Hungary.

Population balance modeling (PBM) has been around for decades to describe crystallization and other particle processes. The model’s heart is the PB equation, which considers the temporal and spatial evolution of the population density function under the action of primary and secondary crystallization mechanisms.
The PBE can be accompanied by energy, material, and momentum balances, which can simulate numerous processes from the crystal size and shape distribution dynamics through multi-population systems to micromixing-governed crystallization. Sufficiently detailed models, solved sufficiently accurately, identified with sufficient precision, and applied reasonably quickly can benefit science and technology. However, this is sometimes more art than science. In industrial environments, the models must answer questions rapidly, with minimal effort.
The first case study presents a PBM identification and optimal design space (DS) determination based on mostly historically available laboratory and plant-scale batch cooling crystallization data. The optimized temperature cycling process halves the currently applied batch time at a 4m3 scale and considerably reduces the heating and cooling demand. The process was validated successfully in laboratory experiments. Since the same API can also be obtained in salting-out crystallization with improved productivity and yield, a follow-up project aimed to adapt the afore-described concepts to this fed-batch process. Off-the-shelf lab and plant scale data was available from the industrial partner again for a rapid model re-calibration. Then, with the identified model, we determined the critical process parameters (CPPs) using a global sensitivity analysis at the lab and industrial scale - which surprisingly happened to be different in the aspects of stirring rate. Finally, we determined the DS in three different crystallizers using Monte-Carlo mapping. Despite time pressure (2 months overall), the calculated DS was validated successfully in the lab scales. The models involved in these case studies were simple 1D nucleation-growth-dissolution PBMs assuming perfect mixing, which can be viewed as a sacrifice of mathematical complexity to be able to deliver results within aggressive time frames.
PBMs are also handy for testing novel hypotheses, especially considering that the simulations bring more value as the process complexity increases. Such can be the case of temperature cycle-aided deracemization (TCID). TCID is a crystallization and racemization-enabled technique that can deliver enantiopure crystal products with good yield. Choosing the right process conditions is not trivial or simple, though, and most studies neglected the product particle sizes. We hypothesize that Quality-by-Control through direct nucleation control (DNC) using standard particle monitoring techniques can tremendously simplify the design and allows for simultaneous particle size control. To test this, we implemented the PBM-based simulation of the TCID, which involved two populations for the two conglomerates. Simulating the DNC, which is a simple, relative particle number-based feedback control strategy confirmed this hypothesis. Experimental validation of these findings is in the pipeline. Still, TCID has a high energy demand, so we moved further and analyzed integrated milling as a process intensification platform. A parametric optimization was performed for the integrated crystallization-wet milling system, parameterizing the key kinetic and process constants to generate a synthetic database.
This was followed by data mining, whose efforts can be divided into three groups: classification of typical operation modes, explainable regression for product property estimation, and time series analysis and clustering to find qualitatively similar dynamic operating profiles.
By this fully in-silico approach, we identified several known and a few yet unknown features of the process, which can serve as guidance for rapid experimental design of highly performing, milling-intensified TCID.

Speaker: Botond Szilagyi - Budapest University of Technology and Economics (Hungary)

Botond Szilagyi
earned his PhD from the Babes-Bolyai University in 2016. After the graduation, he worked in the group of Dr. Zoltan Nagy at Purdue University (US) as a post-doctoral research associate for more than four years. He moved back to Europe in 2021 and he is now Associate Professor at Budapest University of Technology and Economics (Hungary).
His research interest is the process engineering, especially the population balance based modelling, optimization and control of cooling crystallizers. He co-authored a book, a book chapter and over 30 research papers.
Botond received awards and scholarships from EFCE, Hungarian Chemical Society, World Federation of Scientists and the European Commission (MSCA Individual Fellowship). He is also co-funder of the CRYSYST LLC start up, which provides software development and consultancy services for crystallization processes.

For more information contact:
  • Daniele Marchisio, professor of Department of Applied Science and Technology-DISAT, at the following e-mail:
  • Elena Simone, professor of Department of Applied Science and Technology-DISAT, at the following e-mail: