This paper describes a practical method for granulation scale-up by me
ans of a neural network. Wet granulation was conducted using an agitat
ion fluidized bed, and the scale-up characteristics were investigated
using a neural network with back-propagation learning. Granule propert
ies obtained by production-scale granulation under various operating c
onditions were predicted. Extremely good correlation was obtained betw
een the predicted data and the experimental data of agitation fluidize
d bed granulation. It was found that granulation scale-up could be con
ducted with high accuracy by a neural network without constructing a m
athematical model with a complicated non-linear relationship using a v
ast amount of experimental scale-up data.