There are many published models for predicting crystal size distribution (CSD) in the literature. However, none of them have been independently and comprehensively tested, which is important for industrial acceptance and confidence of these models. Therefore, in this study, using solubility and kinetic data from the literature, an in silico tool for predicting the crystallization process performance of a model compound system (paracetamol in ethanol) was developed and challenged by independent experiments at the 50 L pilot scale. The solute concentration was tracked, and the final CSD was quantified using three measurement techniques including a novel image analysis tool. The reported parameter uncertainties were also addressed using Monte Carlo simulations. The results showed that, when the models were used within their validity range (e.g., suspended solids), they were able to describe the observed process trends/dynamics (CSD and solute concentration) under varying experimental conditions (cooling time and seed mass) with R2 ranging from 0.72 and 0.90. Overall, the results indicate that, using Monte Carlo simulations to account for known parametric uncertainties, the models can support model-based approaches for crystallization process development from scale-down to scale-up studies as well as control evaluation.
Substance-tailored testing strategies in toxicology: An in silico methodology based on QSAR modeling of toxicological thresholds and Monte Carlo simulations of toxicological testing
