Spray drying is a promising method in the pharmaceutical industry, enabling continuous manufacturing of amorphous solid solutions. To reduce the dependence on trial-and-error experiments in this work, we performed a detailed computational fluid dynamics (CFD) simulation, which allows us to predict the outcome of the spray drying process of indomethacin-PVP K30-ethanol mixture. The developed methodology includes two steps, where in the first step we simulate atomisation process, characterise the size distribution of formed droplets, and spraying angle. In the second part, we used calculated droplet size distribution and spraying angle as an input in full spray dryer simulation to predict flow and temperature profiles, properties of produced particles, their drying profile and heat losses. Secondary breakup (Wave model) was found to be required to correctly predict size distribution of formed particles, and an influence of turbulence on the particles was also studied. The proposed method is validated by exploring the effects of varying the process parameters and experimentally confirming the predictions of the simulations. Key outcomes that are predicted include powder size distribution (achieving at maximum 23%F relative standard deviation against the experiments for all control and validation datasets), humidity and outlet temperature (predicted with accuracy of 2.5 °C against the experiment, well within the uncertainty of the experimental measurement). This work demonstrates the potential of computational fluid dynamics as a reliable tool to optimise the process of spray drying in pharmaceutical manufacturing. • Computational fluid dynamics was applied to simulate atomization and spray drying process. • Secondary breakup was found to play a significant role in spray drying. • Particle size agreed with experimental results with maximum 23% relative uncertainty. • Temperature prediction had an error of 2.5 K, below the hardware sensitivity of 3 K. • Wave breakup parameters were adjusted based on a sensitivity study.
Martynek et al. (Thu,) studied this question.