Machine Learning-Derived Correlations for Scale-Up and Technology Transfer of Primary Nucleation Kinetics.

Scaling up and technology transfer of crystallization processes have been and continue to be a challenge. This is often due to the stochastic nature of primary nucleation, various scale dependencies of nucleation mechanisms, and the multitude of scale-up approaches. To better understand these dependencies, a series of isothermal induction time studies were performed across a range of vessel volumes, impeller types, and impeller speeds. From these measurements, the nucleation rate and growth time were estimated as parameters of an induction time distribution model. Then using machine learning techniques, correlations between the vessel hydrodynamic features, calculated from computational flow dynamic simulations, and nucleation kinetic parameters were analyzed. Of the 18 machine learning models trained, two models for the nucleation rate were found to have the best performance (in terms of % of predictions within experimental variance): a nonlinear random Forest model and a nonlinear gradient boosting model. For growth time, a nonlinear gradient boosting model was found to outperform the other models tested. These models were then ensembled to directly predict the probability of nucleation, at a given time, solely from hydrodynamic features with an overall root mean square error of 0.16. This work shows how machine learning approaches can be used to analyze limited datasets of induction times to provide insights into what hydrodynamic parameters should be considered in the scale-up of an unseeded crystallization process.

Linked experimental and modelling approaches for tablet property predictions.

Recent years have seen the advent of Quality-by-Design (QbD) as a philosophy to ensure the quality, safety, and efficiency of pharmaceutical production. The key pharmaceutical processing methodology of Direct Compression to produce tablets is also the focus of some research. The traditional Design-of-Experiments and purely experimental approach to achieve such quality and process development goals can have significant time and resource requirements. The present work evaluates potential for using combined modelling and experimental approach, which may reduce this burden by predicting the properties of multicomponent tablets from pure component compression and compaction model parameters. Additionally, it evaluates the use of extrapolation from binary tablet data to determine theoretical pure component model parameters for materials that cannot be compacted in the pure form. It was found that extrapolation using binary tablet data - where one known component can be compacted in pure form and the other is a challenging material which cannot be - is possible. Various mixing rules have been evaluated to assess which are suitable for multicomponent tablet property prediction, and in the present work linear averaging using pre-compression volume fractions has been found to be the most suitable for compression model parameters, while for compaction it has been found that averaging using a power law equation form produced the best agreement with experimental data. Different approaches for estimating component volume fractions have also been evaluated, and using estimations based on theoretical relative rates of compression of the pure components has been found to perform slightly better than using constant volume fractions (that assume a fully compressed mixture). The approach presented in this work (extrapolation of, where necessary, binary tablet data combined with mixing rules using volume fractions) provides a framework and path for predictions for multicomponent tablets without the need for any additional fitting based on the multicomponent formulation composition. It allows the knowledge space of the tablet to be rapidly evaluated, and key regions of interest to be identified for follow-up, targeted experiments that that could lead to an establishment of a design and control space and forgo a laborious initial Design-of-Experiments.

A Structured Approach To Cope with Impurities during Industrial Crystallization Development.

The perfect separation with optimal productivity, yield, and purity is very difficult to achieve. Despite its high selectivity, in crystallization unwanted impurities routinely contaminate a crystallization product. Awareness of the mechanism by which the impurity incorporates is key to understanding how to achieve crystals of higher purity. Here, we present a general workflow which can rapidly identify the mechanism of impurity incorporation responsible for poor impurity rejection during a crystallization. A series of four general experiments using standard laboratory instrumentation is required for successful discrimination between incorporation mechanisms. The workflow is demonstrated using four examples of active pharmaceutical ingredients contaminated with structurally related organic impurities. Application of this workflow allows a targeted problem-solving approach to the management of impurities during industrial crystallization development, while also decreasing resources expended on process development.