PHARMACEUTICALS operations effectively and to lower the ' activation energy ' for our chemists and customers . The benefit of a toolbox with predictive models to scale up chemical reactions and physical operations justifies the additional effort at lab-scale . Scale-up performance and effective process control are improved in the long term . Quality by Design ( QbD ) strategies can be applied to process development . Nevertheless , we have to provide our chemists and customers easy-touse tools to study and establish these models . Like in a toolbox , the chemist can then choose the best tools for his task in process development . To avoid overloading , we have established different toolboxes for different unit operations – some very specialised , some more general . Figure 1 illustrates what is needed to generate and establish a reliable
predictive model for scale-up and process control . We have combined all of the pieces into a toolbox . The general idea is to scale up directly to manufacturing-scale based solely on dedicated smart lab experiments . Predictive computerbased models of the chemical reactions and unit operations are used to minimise the risks and are part of an improved control strategy . The heart of a typical toolbox therefore consists of standard lab programmes to generate all of the necessary quality and integrity data in order to establish a reliable predictive model for the intended final scale . Figure 2 illustrates the lifecycle of such a predictive model . Downscaling from plant to lab equipment is the first step . The geometrical comparability of the reactors , and their performance in heat and mass flow or mixing properties , have to be considered in such a downscaling . Digital twins of the reactors of the targeted manufacturing train , as well as the used laboratory equipment , are needed to design meaning full experiments on lab scale . The same IT-based models generally used for scale-up may also be used to scale down . Experiments performed in wellcontrolled downscale environment generate integer data of value , which can be processed later by the software . Lab automation generates time-resolved data points of the process performance . Analytics provide kinetic data for the reactions and transformations . Inprocess controls ( IPC ), in-line process analytical technologies ( PAT ) and heat flow provide us with a huge amount of significant laboratory data . Using the digital twin approach , this data is transformed to scale-independent mathematical equations and models . The models can then be used to predict how the process will perform in large-scale equipment or what the influence of process parameter variation will be . Sensors for product and jacket temperature and pressure provide data at large scale , which can be compared to the predicted values . This data can then be reintroduced to the software to adjust , optimise and fine-tune the model . Comparing prediction with real data can be used to qualify the model . The advantage of this toolbox approach is demonstrated in two examples below .
Hydrogenation toolbox
In collaboration with the University of Applied Sciences & Arts of Northwestern Switzerland ( FHNW ), we have established a novel newscale strategy for hydrogenations . 2 This work was part-funded by the Research Fund of Aargau , a Swiss canton . Heterogeneous catalyst-driven hydrogenations are one of the most common types of chemical transformation in the synthesis
15