A framework for material calibration and deformation predictions applied to additive manufacturing of metals

Abstract

Additive manufacturing of metals offers the possibility of net-shape production of topologies that are infeasible by conventional subtractive methods, e.g. milling. However, additive manufacturing of metals posses a high defect density, which necessitates accounting for the larger statistical material variation than in conventional metals. To prevent time-consuming experimental component testing, computational methods are required that account for the topology and metal as well as that precisely predict component failure even for metals with statistically varying properties. Predictions require an automatic and modular evaluation of the possibly large set of material properties and their scatter by employing calibration experiments. In addition, a module for failure and deformation prediction as well as a module for the aggregation of experimental data are required. In this contribution, such a modular framework is presented and the encapsulated and non-structured storage of statistically varying material properties is highlighted. In addition, the framework includes multiple FEM solvers and a multitude of optimizers, which are compared and the objective function is addressed. The approach is employed for the third Sandia Fracture Challenge for which two configurations (homogeneous and heterogeneous material properties) are studied and discussed. The blind-predictions of the verification geometry are used to identify the benefits and drawbacks of the framework.