How to maintain control in a flow process, through a smart feedback control system

The Problem

For an established flow acylation process, no online analysis or reaction conversion measurement had been identified. The lack of real-time measurement meant that it was difficult to optimize conversion and this, in turn, hindered the viability of the process at manufacturing scale. 

The Breakthrough

By developing a chemometric model, APC scientists were able to utilize an ATR-FTIR to monitor the process. In addition to establishing PAT to monitor the process in real-time, the information was used to inform a feedback control model to identify and control the process at optimum conditions for maximum conversion. 

The Impact

The relative abundance (predicted by APC’s chemometric model) of the three species (product, impurity and starting material) was within 3 area% of the offline analysis. The feedback control system designed was thus deemed successful in being capable of identifying and controlling the process unsupervised.

At first, manual tuning was required to find the optimum operating region, based upon analysis of offline data. While this process was sufficient to demonstrate the potential for high conversion in a laboratory setting, the inability to consistently achieve the optimum conversion (based on a defined set of operating conditions) challenged the viability of the flow process to suffice for manufacturing. The existing practice of manual tuning adopted for this process represented a significant barrier to the adoption of the flow process in manufacturing, where inconsistent feed blends and in-process disturbances are inevitable.

To mitigate the time spent at non-ideal conditions, real-time measurement, and response in the form of a process feedback control loop was required. The overall project objective was therefore to develop an in-flow feedback control system and demonstrate proof-of-concept via control of the flow peptide acylation reaction. Suitable process analytical technology (PAT) for online measurement was identified that could predict the reaction conversion. A chemometric model was developed for application to the flow acylation process to predict Product, Starting Material and By-Product conversion. The chemometric model was developed using a PLS model basis, with the first derivative and standard normal variate applied to the data.

During online application of the model to the spectral data generated by the ATR-FTIR, the relative abundance of the three species was within 3 area% for the data tested. Through the chemometric model developed, the identification of the Product, Starting Material and By-Product species was possible through ATR-FTIR spectroscopy for real-time accurate prediction. The success of the feedback loop proved that through using online PAT, the reaction variables could be controlled in real-time and without operator interaction.