PSI - Issue 38

Alexander Raßloff et al. / Procedia Structural Integrity 38 (2022) 4–11 A. Raßlo ff et al. / Structural Integrity Procedia 00 (2021) 000–000

6

3

SVE Property

Reconstruction

Light microscopy CT

Simulation

Statistics

Structure characterisation

Ti-6Al-4V LPBF specimens

Structure-property relationship

Statistical structure quantification

Data analysis

Fig. 2. Illustration of the approach to access pore microstructure-property relationships based on experimentally observation and numerical simu lation: Specimens are printed, the material microstructure is observed by light microscopy and computed tomography, the structure is statistically quantified and SVEs are in silico reconstructed for simulating fatigue indicator parameters by crystal plasticity simulations to derive the relation ships by data analysis.

conducted to calculate FIPs. Analysing the simulation results by extreme value distribution yields the targeted data of pore structure and fatigue property pairs, enabling the ranking of microstructures. Additionally, a ranking based on the empirical √ area parameter as introduced by Murakami (1989) is comparatively introduced. The experimental and numerical methods are briefly introduced in section 2. In section 3 the results of three illustrative studies are presented and shortly discussed. A conclusion is given in section 4.

2. Methods

2.1. Experimental Microstructure Characterisation

For this study, Ti-6Al-4V powder of Grade 23, which is atomised by argon, is utilised. A laser powder bed system is used to build rectangular specimens of size 10 × 10 × 15 mm as shown in Figure 2. LM images as in Figure 2 of sections parallel to the building direction are taken to measure the width of the visible prior β -grains, yielding an average width of 192 µ m. These values constitute a simplified characterisation of the grain structure. In order to characterise the pores, x-ray CT scans as illustrated in Figure 2 are conducted with a voxel resolution of 9 µ m. The pores are detected and analysed using the software VGSTUDIO Max 3.4. For the porosity analysis, a workflow based on du Plessis et al. (2018) is utilised. To achieve a more sophisticated processing, the data is trans ferred to MATLAB, where additional quantities are calculated to serve as input for the microstructure characterisation as described in the subsequent section. Microstructural descriptors. An adequate and condensed description of the structure is mandatory for deriving PSP relationships and for reconstructing synthetic microstructures. For the present study, the spatial distribution and mor phology of the pores need to be captured. For this purpose, each voxel cluster from the CT that was identified as a pore is approximated by an ellipsoid to allow for a simplified, analytical description of the pores. These pores are characterised by the distributions of two meaningful parameters, the equivalent spherical diameter (ESD) d eq and the shortest distance between the surfaces to the nearest pore δ S . As the pores are to be reconstructed as spheres for this first demonstration, the ESD captures not only the crucial e ff ect of pore size, but also the complete morphology. The latter parameter contributes to the significant e ff ect of high plastification in dense areas. Describing the distribution of these quantities by probability density functions (PDFs) allows for characterising the structure in the desired low dimensional manner by few parameters, i.e. the descriptors. Figure 3 shows the distribution of d eq and δ S derived by 2.2. Numerical Microstructure Characterisation and Reconstruction