Additive Manufacturing (AM) is growing in importance as a fabrication process for the space industry, enabling weight and cost savings through optimized designs for components. The use of AM gives aerospace engineers an alternative to more traditional manufacturing processes, but also retains the challenge of producing parts without defects. These problems can be approached using nondestructive methods such as X-ray Computed Tomography (CT) and Finite Element Modelling (FEM) to inspect geometries and quantify the impact of defects on the mechanical properties of a part, taking into account factors such as internal stress from metal cooling during fabrication.

Figure 1. SOGECLAIR CAD Model

Researchers at ELEMCA and CNES have found success with this method by focusing on the quality control of an aluminium AM part for space applications. This study is part of a larger project, ALMIA (Additive Layer Manufacturing for Industrial Application)[1], led by SOGECLAIR Aerospace with CNES, FUSIA, RATIER-FIGEAC and ICA; the project aims to define and validate a new AM process for space applications. In this particular study, a single part intended for the TARANIS satellite, and made from the aluminium alloy, AS7G06, reduces weight by 16% and integrates eleven functional parts into one.

X-ray CT was used to identify the location of porosities in the material, and to generate Finite Element (FE) models in Simpleware software (Synopsys, Mountain View, CA) for simulation of actual part response, including design optimization and fabrication validation. A random vibration model was considered, and comparison made between the results from a theoretical geometry and the manufactured component. The goal was to confirm the quality of the AM process, and to test a new method of validating CT scanned parts. Simulations using these models represent the actual part behaviour in real conditions, enabling analysis of the effect of each defect, and comparison to the theoretical simulations carried out using an idealised Computer-aided Design (CAD) model.


Table 1. Random vibration input [©SOGECLAIR]
Table 2. Point Masses [©SOGECLAIR]

SOGECLAIR carried out simulations from a CAD (Figure 1) model to validate the design evolution of the part, in particular the result from topological optimization. The model considers random vibrations in the X, Y and Z directions (Table 1), with the critical axis chosen for the real test, and vibration tests performed in this direction on the real part to adjust the model. In this case, the FE model is based on the CAD design, rather than the X-ray CT dataset. In the simulations, the part is considered to be fixed to the base, and two point masses represent the SAS equipment and connectors (Table 2); the complete model includes titanium screws and permaglass rings, containing about 800,000 elements. Simulation results showed that the dimensioning axis is X, and the higher stresses were found when applying a vibration across the X axis. The value of maximum stress obtained by the modelling on CAD data was compared to the results from modelling the actual data obtained from CT scanning.

CT Reconstruction & Simulation

Figure 2. Image-based Segmentation

X-ray tomography was conducted on a V Tome X system, developed by General Electric (Phoenix), with a copper filter placed just after the X-ray source to reduce the reconstruction artefacts and increase the power of the scan[3]. The volume and internal features were visualised for inspection, and did not show any voids, cracks, inclusions, or other material health defects larger than the resolution of the voxels; this proved that the manufacturing process for the part is well-controlled.

Simpleware software was used to process the image data and to generate a computer model for simulation. Image-based meshing algorithms in the soft-ware generate Finite Element (FE) meshes for topologies of arbitrary complexity, offering advantages over traditional CAD models by representing ‘as-manufactured geometries’ that capture slight deviations from AM designs. With these techniques, the accuracy of the meshes is only limited by the quality of the image acquisition and segmentation, whereby automatic and semi-automatic tools are used to capture regions of interest within the image data. The mesh generation algorithms ensure that multi-part models have perfectly conforming interfaces without gaps or overlaps[2].

To obtain the FE models, Simpleware ScanIP was used to segment the main body of the component from the image space. Threshold-based segmentation was used to create a mask of the structure, before the mask geometry was refined by disconnecting all regions of the mask from the main body. Manual segmentation techniques such as painting and threshold-assisted painting were used to ensure accurate segmentation of important details such as the screw holes, and to reduce the influence of metal artefacts. Light Recursive Gaussian-based smoothing was applied to the segmented geometry to increase the surface smoothness prior to meshing (Figure 2).