(649c) Sensor Fusion and Advanced Process Control Strategies for Multistep Synthesis in Flow | AIChE

(649c) Sensor Fusion and Advanced Process Control Strategies for Multistep Synthesis in Flow

Authors 

Bhaskar, A., Perceptive APC
Whitaker, D., Takeda Pharmaceutical
Nazemifard, N., University of Alberta
Yamaguchi, K., Takeda Pharmaceuticals International Co., 40 Landsdowne St
Papageorgiou, C. D., Takeda Pharmaceuticals International Co.
Mitchell, C., Takeda Pharmaceuticals International Co.
Transmetalation and coupling reactions with Grignard reagents are rapid and highly exothermic, that can cause safety and selectivity issues during batch scale-up [1]. It has been well established that continuous flow processes can reduce or eliminate heat and mass transfer limitations, that are otherwise inhibitory in large-scale batch reactions [2]. At Takeda Pharmaceuticals, we have developed a flow chemistry platform comprised of two tubular and one stirred tank reactor for the generation and reaction of a Grignard reagent, affording an advanced intermediate for an active pharmaceutical ingredient currently under development [Fig 1].

The process is highly sensitive to small changes in the stoichiometry of the reagents, resulting in the generation of hard to purge impurities. The process control strategy in manufacturing involved offline HPLC analysis and a multicriteria IPC strategy that ensured conversion was maximized and the levels of 2 key impurities were maintained below a critical level. This resulted in a 5-scenario IPC and a complex calculation table to adjust feed flow rates, that was prone to operator errors and required multiple attempts to successfully pass. Most importantly, it did not provide continuous monitoring of the process to ensure that disturbances were addressed, and product quality maintained.

To measure the state of the process, we have implemented an inline mid-FTIR as well as an online UPLC. Due to our low impurity specifications and the noisy solvent baseline, the FTIR measurement was found to have too low of a precision to control on alone but able to capture the process dynamics. On the other hand, UPLC yields far more reliable measurements at the cost of a low measurement frequency. To manage the low observability, we have implemented a control and monitoring strategy that combines multiple information sources to improve our state estimations and control actions and leveraged digital twins.

References

[1] Kadam, A., Nguyen, M., Kopach, M., Richardson, P., Gallou, F., Wan, Z.-K., & Zhang, W. (2013). Comparative performance evaluation and systematic screening of solvents in a range of Grignard reactions. Green Chemistry, 15(7), 1880. https://doi.org/10.1039/c3gc40702k

[2] Riva, E., Gagliardi, S., Martinelli, M., Passarella, D., Vigo, D., & Rencurosi, A. (2010). Reaction of Grignard reagents with carbonyl compounds under continuous flow conditions. Tetrahedron, 66(17), 3242–3247. https://doi.org/10.1016/j.tet.2010.02.078