Abstarct: In this work, a modified Drucker-Prager/Cap (DPC) constitutive model was used to describe the densification behavior of alumina KMS-92 powder. The model parameters were obtained through some simple tests. For determining the parameters of shear failure surface, diametral and axial compression tests were used. An instrumented die was employed to obtain the parameters of cap surface, elastic constants and friction coefficient between powder particles and die wall. Finally, the obtained parameters were implemented in Abaqus for finite element (FE) simulation of powder densification. User-defined subroutine USDFLD was used to describe material behavior as a function of relative density. The FE results showed a very good correlation with the experimental ones. In addition, a cost effective method for obtaining the constitutive model parameters for similar powder material from a simple uniaxial compaction test and artificial neural networks (ANN) algorithm was proposed. In this approach, ANN is trained to predict a set of parameters for modified DPC model such that the differences between experimental an FE analysis curve for uniaxial powder compaction was minimized. The input parameters of ANN were paprameters describimng the load-displacement curves. The results obtained from this approach for load-displacement curves of uniaxial powder compaction show a very good agreement with the experimental counterparts.