A novel approach to calibrate the Drucker-Prager Cap model for Al7075 powder

Determination of the parameters of modified Drucker–Prager Cap (DPC) constitutive model for Al7075 powder is investigated in this work. The parameters are normally identified by experiment which is time consuming, tedious and expensive. In this study, the constants of DPC model are identified by conducting only a simple uniaxial powder compaction test, using finite element (FE) simulations in ABAQUS/standard, and utilizing artificial neural networks (ANN). The relation between the Young’s modulus (E) and relative density of the powder was incorporated in ABAQUS code using a USDFLD user-defined subroutine. In the proposed approach, the neural networks are trained to predict the DPC parameters in a way to minimize the differences between experimental and FE curves of uniaxial powder compaction. The input parameters of the ANN were features of uniaxial powder compaction load–displacement curve. A reasonable agreement was observed between the experimental and numerical load–displacement curves of the powder compaction for the DPC parameters predicted by ANN. Moreover, the accuracy of this DPC model was verified again in compaction of a bush-type sample.