Validation of fluid bed granulation utilizing artificial neural network

Abstract

Three innovative components (an annular gap spray system, a booster bottom and an outlet filter) have been developed by Innojet Technologies to improve fluid bed technology and to reduce the common interference factors (clogging of nozzles and outlet filters, spray loss, spray drying and fluidized bed heterogeneity). In a fluid bed granulator, three conventional components have been replaced with these innovative components. Validation of the modified fluid bed granulator has been conducted using a generalized regression neural network (GRNN). Under different operating conditions (by variation of inlet air temperature, liquid-binder spray rate, atomizing air pressure, air velocity, amount and concentration of binder solution and batch size), sucrose was granulated and the properties of size, size distribution, flow rate, repose angle and bulk and tapped volumes of granules were measured. To confirm the method's validity, the trained network has been used to predict new granulation parameters as well as granule properties. These forecasts were then compared with the corresponding experimental results. Good correlation has been obtained between the predicted and the experimental data. From these findings, we conclude that the GRNN may serve as a reliable method to validate the modified fluid bed apparatus.