Application of different models to evaluate the key factors of fluidized bed layering granulation and their influence on granule characteristics

Different modeling approaches were used to understand the key factors affecting the outcomes of pulse sprayed FBLG. A large amount of metformin hydrochloride (similar to 83%) was layered onto Cellets (R) seeds to obtain directly compressible granules. The effect of spray rate, mass flow rate, inlet air temperature, atomization pressure, and coating solution on the five granule characteristics (mean size, relative width, granule porosity, production yield, and aggregation index) was evaluated using a DSD and correlated with RSM, PLS, and ANN models. The cohesive drug was converted into non-hygroscopic, free-flowing, and stable granules which had several benefits such as large particle size, narrow size distribution, lesser granule porosity, high yield, negligible aggregation, and good compactibility. RSM (R-2 > 0.81) and ANN models (R-2 > 0.80) had a better fit with experimental factors compared with PLS model (R-2 > 0.47). Machine-learning algorithms like the ANN as considering multiple factors could give a robust and successful modeling for the FBLG process.