Modeling & Feedback Control of Batch & Continuous Crystallization Systems - Webinar Presentation EN

Key Concepts

• Quality by Control: Controlling crystallization systems to achieve desired product quality through real-time monitoring and feedback.
• Composite Sensor Array: Combining data from multiple sensors (FBRM, PVM, Raman, UPLC) for a comprehensive understanding of the crystallization process.
• Process Analytical Technology (PAT): Using analytical measurements in real-time to monitor and control manufacturing processes.
• Model Identification: Determining the parameters of a mathematical model based on experimental data.
• Population Balance Model (PBM): A mathematical model that describes the evolution of particle size distribution in a crystallization process.
• Feedback Control: Adjusting process parameters based on real-time measurements to maintain desired product properties.
• Direct Nucleation Control: Controlling the number of nuclei formed during crystallization to influence crystal size distribution.
• Plug Flow Crystallization: A continuous crystallization method where the fluid flows through a reactor in a plug-like manner, minimizing mixing in the axial direction.
• Oscillatory Baffled Crystallizer (OBC): A type of plug flow crystallizer that uses oscillations to enhance mixing and heat transfer.
• Active Impurity Control: A strategy to control crystal shape by manipulating additives that affect the growth rate of specific crystal faces, compensating for the influence of impurities.

Crystallization Control via Composite Sensor Array and Process Informatics

The speaker discusses the use of feedback control and mathematical modeling for the design and control of crystallization systems. The core idea is to use Process Analytical Technology (PAT) not just for monitoring, but for real-time model development and feedback control, aiming for "quality by control."

• Combining Multiple Sensors: The approach involves integrating information from multiple sensors (a composite sensor array) like PVM (Particle Vision and Measurement), FBRM (Focused Beam Reflectance Measurement), Raman spectroscopy, and online UPLC (Ultra Performance Liquid Chromatography).
• Process Informatics System: The data from these sensors is incorporated into a process informatics platform.
• Iterative Model Development: This platform is combined with a modeling platform for rapid model identification and improvement of measurements and experimental design.
• Direct Design Approach: Instead of the traditional Quality by Design (QbD) approach, the speaker advocates for a "direct design" approach. This involves running a few clever experiments and using feedback control to search for optimal operating conditions to achieve desired crystal properties.
• Redundancy and Robustness: The use of multiple sensors provides redundancy, allowing for real-time data reconciliation and robust feedback control strategies.

Examples of Crystallization Control Strategies

The speaker presents several examples of how this approach can be applied to different crystallization scenarios:

• FBRM for Model Identification and Optimization: Using FBRM data for quantitative information to directly develop models and optimize crystallization processes.
• PVM for Feedback Control with Real-Time Image Analysis: Using PVM for quantitative information extraction, model development, and direct feedback control with real-time image analysis.
• Integration with Online UPLC: Integrating PAT with online UPLC for monitoring and controlling crystallization systems in impure media, providing a complete impurity profile.
• Direct Nucleation Control in Integrated Wet Mill-MSMPR Systems: Implementing direct nucleation control, previously developed for batch crystallization, in integrated wet mill-MSMPR (Mixed Suspension Mixed Product Removal) systems.
• Process Intensification via Plug Flow Crystallization: Using FBRM and feedback control in plug flow crystallization systems for process intensification.

Continuous Manufacturing Platform

The speaker's lab has developed a fully integrated continuous manufacturing platform:

• Small-Scale Microfluidics: Using microfluidic platforms to create the production line of an API (Active Pharmaceutical Ingredient) and connecting it to a formulation platform.
• Large-Scale Continuous Reactors: Using continuous Kentrix-type reactors connected with oscillatory crystallizers and continuous filtration, linked to a formulation platform.
• PAT Monitoring and Control: The key idea is to run these systems continuously, using PAT to monitor and control the unit operations and maintain steady-state conditions.

Population Balance Modeling Toolbox

The speaker's group has released a population balance model (PBM) toolbox:

• Model-Based Design: The toolbox allows users to perform model-based design and simulation of two-dimensional population balance models.
• Crystal Shape and Purity Simulation: It can simulate the effects of operating conditions on crystal shape and purity.
• User-Friendly Interface: The software has a graphical user interface for easy model parameter input and simulation generation.
• Real-Time Simulation: The efficient numerical implementation (parallelized on GPUs) enables real-time simulation and optimization.
• Integration with Experimental Platforms: The goal is to integrate the modeling platform with experimental platforms like EasyMax and MATLAB for a comprehensive model-based design system.

FBRM Modeling and Real-Time Optimization

The speaker emphasizes the importance of accurately relating model outputs to sensor measurements:

• FBRM Measurement Model: They derived a model for the FBRM measurement, which measures chord length distribution, not size.
• Chord Length Calculation: The model computes chord lengths based on assumed crystal shapes, accounting for the fact that a single crystal will produce an entire chord length distribution due to its shape.
• Model Identification with FBRM Data: The model allows for direct model identification by fitting concentration and chord length distribution data to kinetic parameters.
• Real-Time Optimization: The model is used for real-time optimization of crystallization conditions, such as the cooling profile, to achieve a desired final chord length distribution.
• Simultaneous Model Identification and Adaptation: The approach provides simultaneous model identification and adaptation of operating conditions, even under varying conditions.
• Single Batch Experiment: A single batch experiment can provide kinetic parameters and an optimal temperature profile for a desired chord length distribution.

PVM and Aspect Ratio Measurement

The speaker addresses the discrepancy between ideal aspect ratio (from PBM) and measured aspect ratio (from PVM):

• Random Crystal Orientation: PVM measures the two-dimensional projection of a 3D crystal in a random orientation, leading to an underestimation of the aspect ratio.
• Aspect Ratio Distribution: A single crystal will produce an aspect ratio distribution due to different random orientations.
• Correction Factor: They derived a correction factor based on mean width and length to correlate model predictions with PVM measurements.
• Improved Shape Measurement: Applying the correction factor allows for more accurate real-time shape measurement and correlation with PBM outputs.

Online UPLC for Purity Monitoring and Control

The speaker describes the integration of online UPLC for real-time purity monitoring:

• Impurity Measurement: UPLC measures the concentration of impurities in the liquid phase.
• Mass Balance Calculation: Combining UPLC data with a mass balance allows for real-time calculation of crystal purity.
• Crystal Purity Estimation: The estimated crystal purity based on liquid concentration measurements correlates well with offline UPLC analysis of solid samples.
• Active Impurity Control: This information can be used to create a map of how impurities influence crystal aspect ratio and to implement feedforward-feedback control strategies to control crystal shape by manipulating additives.

Continuous Crystallization and Wet Milling

The speaker discusses different approaches for controlling crystal size distribution in continuous crystallization with wet milling:

• Direct Temperature Control: Manipulating the jacket temperature in the crystallizer to control the number of counts (related to crystal size distribution).
• Wet Mill Recycle: Recycling part of the crystal stream through a wet mill to control the crystal size distribution.
• Wet Mill as Upstream Nucleator: Using the wet mill as an upstream unit operation for seed generation, decoupling nucleation from growth in the main crystallizer.
• Feedback Control of Wet Mill Temperature: Manipulating the jacket temperature in the wet mill to control supersaturation and nucleation rate.
• Benefits of Upstream Wet Mill: Decreased startup duration and improved crystal quality due to controlled nucleation in a small volume, high shear device.

Plug Flow Crystallization for Spherical Particles

The speaker presents a plug flow crystallization system for creating spherical particles:

• Spherical Particle Formation: The goal is to create spherical particles to avoid processability problems associated with fragile or needle-shaped crystals.
• Zone Control: The plug flow system has different zones for nucleation (controlled by anti-solvent addition), growth (controlled by anti-solvent or cooling), and agglomeration (controlled by binder addition).
• Oscillatory Baffled Crystallizer: An oscillatory baffled crystallizer (OBC) is used to create a plug flow system with good mixing and long residence times.
• Feedback Control of Nucleation Zone: FBRM is used to control the size of primary particles in the nucleation zone by manipulating anti-solvent addition or temperature.
• Future Integration of Imaging: The speaker mentions plans to incorporate imaging measurements in the growth and agglomeration phases.

Conclusion

The speaker concludes by emphasizing the economic and technological drivers for better control of crystallization processes. They highlight the benefits of using PAT not only for monitoring but also as a quantitative tool for model identification, model-based design, and direct feedback control. This approach can significantly optimize development time and performance, whether considering a single unit operation or an integrated particle formulation process.