Spray drying techniques are widely used in the pharmaceutical industry to produce fine drug powders with different properties depending on the route of administration. Process parameters play a vital role in the critical quality attributes of the final product. This review highlights the progress and challenges in modeling the spray-drying process, with a focus on pharmaceutical applications. Computational fluid dynamics (CFD) is a well-known method used for the modeling and numerical simulation of spray drying processes. However, owing to their limitations, including high computational costs, experimental validation, and limited accuracy under complex spray drying conditions. Machine learning (ML) models have recently emerged as integral tools for modeling/optimizing the spray drying process. Despite promising accuracy, ML models depend on high-quality data and may fail to predict the influence of new formulation or process parameters on the properties of the dried powder. This review outlines the shortcomings of CFD modeling in the spray drying process. A hybrid model combining ML and CFD models, emerging techniques such as the digital twin approach, transfer learning, and explainable AI (XAI) are also discussed. A hybrid model combining ML and CFD models is also discussed. ML is considered an emerging technique that could assist the spray drying process, and most importantly, the utilization of this method in pharmaceutical spray drying.

The spray drying technique is suitable for different kinds of liquid dispersions and can be easily optimized to produce solid particles with tailored properties. The spray drying technique is a complex process. As an example, it is difficult to track drying kinetics, shape, and morphological changes on the scale of a single droplet. To better understand the effect of drying process variables on dried particle formation, it is useful to observe the drying of single droplets. Fundamental processes, such as mass and heat transfer, can then be easily monitored and compared with theoretical models. Acoustic levitation enables droplet/particle suspension in the air without any mechanical contact. Experiments in the acoustic levitator can be used to mimic the drying process in the spray dryer. The drying kinetics of single droplets of PEG6000 into solid particles was studied. Droplets with an initial polymer concentration (PEG6000 aqueous solution of 5%, 10%, and 15% (w/w)) were investigated at different gas drying temperatures. The size of the droplet, moisture content, and the shape evolution of the droplet/particle during the drying process were studied. The experimental drying curves were compared with the Reaction Engineering Approach (REA). The REA models were shown to provide a very good agreement for drying behavior, with a relative error of about ±3% between the initial and predicted droplet mass. This model can be implemented into the large-scale modeling of spray drying using Computational Fluid Dynamics (CFD).