Use of in-Situ Image-Based Sensor for Model-Based Control of Batch Crystallization Processes

Crystallization from solution is a widely practiced unit operation in specialty chemical, pharmaceutical, and agrochemical industries for solid-liquid separation, purification, and production of solid crystals with desired shape and size distribution. The crystal size distribution (CSD) and shape or polymorphic form are the most important product quality variables for any crystallization process as these variables strongly influence the effectiveness of the end-use properties of the crystal products (bioavailability, compressibility, stability, dissolution rate, etc) as well as the efficiency of downstream operations (filtration, drying, storage, handling, etc). The key control variable for a batch crystallization process is the solution supersaturation as it strongly determines the nucleation and growth phenomena and thereby the final crystal shape and size distribution. There are other possibilities also to influence the product CSD. For example, the amount of seed crystals and its size distribution significantly affect the achievable final CSD. However, the control of a batch crystallization process to produce crystals with desired shape and size distribution is a challenging task due its high nonlinearity and high sensitivity to process conditions. Additionally, the lack of understanding of the process and the lack of in-situ sensors to directly measure the product quality have also hindered the implementation of advanced control of crystallization processes in industry [1, 2]. Currently, these processes are mostly operated so as to follow pre-specified trajectories that are supposed to yield desired product quality, without direct online control of quality. This may lead to batch-to-batch variability, unacceptably low product qualities, and overall production inefficiencies.

For most batch crystallization processes, it is often difficult to obtain on-line relevant information about the crystal size distribution and the dissolved solid concentration in the liquid phase. An in-situ ATR-FTIR technique has been successfully applied to monitor the solute concentration in the liquid phase. Crystal size estimation from in-situ images has received attention recently as a means to estimate product properties in real-time. We have recently developed an automated image analysis strategy that combines classical image analysis techniques with multivariate statistics for online analysis of in-situ images from crystallization process [3, 4]. The strategy introduces a novel image segmentation step based on information extracted from multivariate statistical models. Experimental results have shown that the strategy effectively extracts crystal size and shape information from in-situ images and can be used to reliably investigate crystallization kinetics, especially for high-aspect ratio systems. Since the image analysis strategy yields real-time measurements of both particle length and width, it is useful for validating both one-dimensional and multi-dimensional population balance models for crystallization processes.

The aim of the present research is to develop a comprehensive strategy for effective control of batch crystallization processes that incorporates mathematical models based on first-principles, efficient dynamic optimization techniques and model based control algorithms. It integrates various in-situ process analytical tools (PAT) for real-time measurement of the state of the processes. A mathematical model based on population balance equation [5, 6] is developed and validated with help of the online CSD measurements by our image analysis strategy and concentration measurements by in-situ ATR-FTIR. The model is then used for real-time optimal control of the batch process in a receding horizon framework. The closed-loop optimal control approach ensures that the process always operates within feasible conditions and it drives the process to achieve the desired product specification. The approach is applied for the batch crystallization of monosodium glutamate (MSG) and the experimental results demonstrate that the strategy leads to effective control of the crystallization process.

References

[1] R. D. Braatz, Annual Reviews in Control 26 (2002) 87-99.

[2] Z. K. Nagy, ESCAPE 18 (2008).

[3] D. Sarkar, X.T. Doan, Y. Zhou, R. Srinivasan, Chemical Engineering Science 64 (2009) 9-19.

[4] Y. Zhou, R. Srinivasan, L. Samvedham, Computers & Chemical Engineering 33

(2009) 1022-1035.

[5] A. D. Randolph and M. A. Larson, Theory of Particulate Processes, second ed. Academic

Press, New York, 1988.

[6] D. Ramkrishna, Population Balances: Theory and Applications to Particulate Systems

in Engineering, Academic Press, New York, 2000.