PSI - Issue 13

Richard J. Williams et al. / Procedia Structural Integrity 13 (2018) 1353–1358 Richard J. Williams et al./ Structural Integrity Procedia 00 (2018) 000–000

1355

3

The profiles of both components were measured using a shadowgraph. The profiles were then compared with those predicted by the FE model. Photographs displaying the slit cylinders are shown in Fig 2. 3. Finite element modelling methodology The modelling approach follows that previously presented in Williams et al. (2018), where a detailed description can be found. A thermo-mechanical FE analysis was run fully coupled within ABAQUS to reduce model development time. The geometry was sub-divided up into ‘layers’, which represent an amalgamation of several ‘true’ layers of the building process, which are 50µm thick. These are all deactivated initially before being re-added layer by layer using the element birth-death technique. A predefined temperature field is applied to the geometry such that each layer is reactivated with a temperature of 1400 ºC, corresponding to the melting temperature of 316L. All other surfaces are insulated. The mesh used in each of the models can be seen in Fig. 3. Note that the coordinate system used is local to the parts and not global in the build direction.

(a)

(b)

Fig. 3. FE mesh used in the (a) horizontal; (b) vertical analysis

In both orientations the parts were modelled as half cylinders employing symmetry boundary conditions. Continuum three-dimensional, coupled displacement-temperature 8 node solid elements, with reduced integration and hourglass control (C3D8RT) were used to maintain computational efficiency. The horizontal model consists of 5 layers of 1.6 mm thick, representing 160 true layers at 0.05 mm thick. The mesh contains 10,680 elements. The vertical model consists of 50 layers, each 0.83 mm thick, and representing 1200 true layers at 0.05 mm. The mesh contains 26,700 elements. The Johnson-Cook constitutive plasticity model is implemented within ABAQUS to describe the plastic strain behavior of the material (Johnson and Cook (1983)). The yield and hardening parameters are derived from LPBF material properties whereas the temperature dependency is derived using the trends of wrought material data, due to an absence of elevated temperature tensile data available for LPBF 316L. This accurately captured the harder material behavior characteristic LPBF material. The model simulates both the build and the slitting procedure performed experimentally. The component was built-in (encastre) at the bottom face whilst the part is built up, after which the boundary condition is removed to simulate sectioning from the build plate. Following this the symmetry condition on the inside face is removed to simulate the EDM slitting process. 4. Results and discussion 4.1. Model Validation In the case of the vertical cylinder, the surface profile of the inside of the cut faces was averaged for the two sides and compared with the FE prediction, as there was only one inside face in the half model. A plot of the inside face horizontal deflection (in the y axis) against the position on the z axis can been in Fig. 4, accounting for the thickness of the EDM wire.