Simpleware Case Study: Microstructure Features of Metal Additive Manufactured Parts

Introduction

Metal powder-based additive manufacturing (PAM) processes typically produce microstructures with different grain sizes and aspect ratios, which affects the mechanical properties of the material. This research looked to better represent the local granular stress fields within the microstructure in a workflow using Simpleware software. The study built on previous constitutive models of a microstructure-informed model based on crystal plasticity. This work was based on a synthetic microstructure representative volume element (RVE) that was numerically generated.

When cubic hexahedral elements are used for discretizing the sample RVE (structured mesh), there is insufficient resolution to capture the grain boundary due to the stair stepped surfaces. The new model developed with Simpleware software tackled this problem by generating a smooth, watertight and multi-domain volumetric Finite Element (FE) mesh, which ensures that the grain boundary is accurately captured.

The results of the new unstructured mesh RVE were then compared to the previous results that used a structured mesh RVE while using simulation time and predicted RVE mechanical properties to determine the efficacy of the new method. This new method allows the local stress intensification to be better captured in the vicinity of the grain boundary, helps with the prediction of defects and void formation in the material and informs future insights into better understanding the performance of complex parts produced using PAM.

Synthetic Microstructure to FE Mesh

An RVE of a synthetic microstructure was generated to simulate grain evolution within 3D domains. The matrix data was converted to a bitmapped voxel image containing grey scale values equal to the grain labels, before being imported to Simpleware ScanIP for processing. The synthetic data was segmented to create a surface representation of each grain, and then exported as a surface and volume mesh. One problem of structured hexahedral meshes are their “stair stepping” effect, which produces a degradation in model accuracy. This was solved in the new model through Simpleware’s anti-alias and smoothing tools, ensuring smooth contours along grain boundaries.

The Simpleware mesh generated with Simpleware FE module was better able to capture the size of small grains at a suitable resolution, compared to the structured mesh. Furthermore, Simpleware’s meshing algorithm was used to convert the segmented boundaries of the mesh into a triangulated surface representation, before applying a multi-part surface decimation algorithm and Delaunay tessellation with tetrahedral elements. The non-uniform meshing scheme results in a more accurate representation of grain geometry, while also reducing total degrees of freedom in the model. This has the effect of decreasing computational cost without significantly affecting result quality.

FE Simulation and Results

Specific boundary conditions and constitutive model definitions were added to the mesh, based on a microstructure-informed constitutive model. Simulation was carried out in Abaqus/Standard to subject the RVE sample to a uniaxial tensile load, using properties from the literature. The properties of the constitutive model were defined using different equations to represent grain shape, while periodic boundary conditions were enforced on all external faces of the RVE.

The new workflow was tested by running a taste case with an RVE of a single grain using the structured and hexahedral mesh. The same material properties, grain orientation, loading and boundary conditions, and dimensions were used. Results showed that the homogeneous normal stress-strain behavior in the loading direction produced identical results. The accuracy of the unstructured mesh was then tested using the RVE of a polycrystal material.

Simulations showed differences in results, based on both grain orientation in the constitutive model, and more detailed capturing of stress concentration effects in the unstructured mesh. These results are due to the hexahedral mesh using equal element sizes that omit some small regions in the model to save on computational cost, leading to a misrepresentation of grain characteristics such as volume, size, and shape. In addition, this mesh introduces artificial stress intensification. By contrast, the tetrahedral meshes show the highest stresses in the grain due to being able to capture sharp corners in the geometry.

Conclusions

The use of Simpleware software in this project to create a more accurate RVE mesh of the PAM results in a better understanding of local stress fields within grains, compared to previous structured models. While in agreement on homogenized behavior, the unstructured mesh showed a significant difference in results for local stress determination, which can be determined much more accurately using a well-refined unstructured mesh. Future work will focus on improving analysis of differences in grain orientation assignment, as well as testing of as-built AM microstructures to validate shape effects, studies of aspect ratio, grain boundary effects, load direction, and material point implementation.

Learn More

• Watch Robert Saunders’ presentation: Mechanical Property Predictions of Additive Manufactured Microstructures via Surrogate Modeling, recorded October 7, 2020.
• Download the presentation slides (pdf)
• Read paper on npj Computational Materials (2021)