Evolutionary Algorithms in Modeling Aerodynamic Properties of Spray-Dried Microparticulate Systems

Spray drying is a single step process in which solutions or dispersions are converted into dry particles. It is widely used in pharmaceutical technology to produce inhalable particles. Dry particle behavior during inhalation, described as the emitted dose (ED) and fine particle fraction (FPF), is determined in vitro by standardized procedures. A large number of factors influencing the spray drying process and particle interaction makes it difficult to predict the final product properties in advance. This work presents the development of predictive models based on experimental data obtained by aerodynamic assessment of respirable dry powders. Developed models were tested according to the 10-fold cross-validation procedure and yielded good predictive ability. Both models were characterized by normalized root-mean-square error (NRMSE) below 8.50% and coefficient of determination (R-2) above 0.90. Moreover, models were analyzed to establish a relationship between spray drying process parameters and the final product quality measures. Presented work describes the strategy of implementing the evolutionary algorithms in empirical model's development. Obtained models can be applied as an expert system during pharmaceutical formulation development. The models have the potential for product optimization and a knowledge extraction to improve final quality of the drug.