The discretized population balance model ofHounslow (Houslow et al., 1988) is modified by Adetayo(1993) to obtain a mathematical model of thegranulation process. The resulting sequential two-stagekernel follows the observed granulation mechanismadequately and is able to predict the shape of thegranule size distribution for the full range ofexperimental data. A size-independent kernel is usedfor the first stage of growth while a size-dependentkernel models the second stage. The extent ofgranulation within the first stage, given byk(,1)t(,1), is found to be proportional to thefractional liquid saturation of the granules. Here,k(,1) represents the rate constant for the first stageof growth and t(,1) represents the time required toreach the final equilibrium size distribution for thefirst stage. A critical fractional water saturationS(.crit). is necessary for the second stage ofgranulation to occur leading to further growth. ForS(,sat) < S(crit), the second stage does not exist (k(,2)(...)0) an equilibrium size distribution isreached at time t(,1). The present research adopts Adetayo's model,which is written in FORTRAN 77 tested and debugged.Validation of the component models (drum granulator,screen and crusher) is next carried out along with thecorresponding sensitivity study of each constituentcomponent model. Based on the actual plant data, thegranulator model parameters are estimated for thefertilizer grade 16-16-18. Finally, a guideline on howto estimate the parameters of a industrial scale modelfrom those obtained with a small-scale granulator fromthose of a lab-scale one is explained and simulation ofa industrial-scale granulation process is carried outto see the effect of the liquid phase content on theprocess performance. Simulation results show that therecycle ratio goes through a minimum as the liquidcontent increases.