Figure 1 Shape comparison between CAD (solid colour/white) and additive manufactured components (transparent green). The CAD colour coding indicates the surface deviation in [mm]. Left: Original component. Right: Shape-compensated component.
Directed energy deposition (DED) is a powerful additive manufacturing technique that combines a high degree of freedom in design with relatively high deposition rates that can increase throughput. DED structures are created by repeatedly depositing weld beads from powder or wire feedstock, which also makes it an attractive technique for hybrid manufacturing because features can be added to a base component. However, the numerous heating and cooling cycles involved in the deposition process result in a complex thermal history that can entail distort the entire component.
Today, trial-and-error is the common approach to correct for these inadmissible distortions and the cost and time required has reduced momentum in the use of DED. The AGENT-3D project, IMProVe, brought together plasma and laser technology provider OSCAR PLT, the Chair of Material Engineering at TU Dresden and manufacturing simulation specialists from Simufact (part of Hexagon’s Manufacturing Intelligence division) to demonstrate smarter ways of addressing the distortions that arise during DED processes. IMProVe stands for “Innovative materials, systems and processes by overcoming the process limits in additive manufacturing”.
The IMProVe team decided the best approach was to feed the weld robot with a pre-deformed geometry that compensates for the distortions that arise during the DED process, so that it will then adjust the final part to match the intended – and undeformed – shape. The key to this procedure, termed distortion-compensation, is predicting the ideal pre-deformed geometry. As the numerical process simulation specialist, Simufact’s remit was to predict the compensated geometry without the need of expensive experimental builds.
Uncompensated benchmark
The project used sample geometry provided by OSCAR PLT. The thin-walled structure consisted of 61 weld tracks, stacked to form a tube with several different corner radii that would be used to probe the effect of sharp turns and smooth curvatures. This was first manufactured without distortion compensation to provide a reference geometry.
To manufacture this part, the start and end points were shifted with each layer to prevent the accumulation of recurring irregularities at these points. A 316 LSi (1.4430) wire feedstock material with a diameter of 1.0 mm was deposited by a coaxial direct-diode laser system at 960 W and a weld velocity of 12 mm/s. The coaxial wire feed ensured high quality deposits in all weld positions.
After the deposition process, the final shape was evaluated using a high-resolution 3D-scan at TU Dresden. Comparison to the CAD data (Figure 1) revealed a shape deviation of up to 1.4 mm had been introduced during manufacturing that needed to be reduced.
Simulation strategy
Figure 2 Model set up in Simufact Welding. Yellow lines indicate the 61 weld tracks with red disks visualising the weld source at the respective start positions. Black arrows represent the weld orientation. For clarity, the meshed geometry is reduced to transparent blue.
While using simulation for shape-compensation is already a state-of-the-art procedure for powder bed fusion processes, the same desirable characteristics that give DED technology additional degrees-of-freedom and high deposition rates also pose new challenges for the setup of numerical process simulations. Unlike powder bed fusion processes, simplified mechanical approaches or techniques based on the effects of cumulative layer heating do not apply. The DED process shares more characteristics with welding, and as such the Simufact Welding simulation environment was used to set up a transient thermo-mechanically coupled model.
To setup the simulation, the 61 weld tracks were directly imported from the robot G-Code (Figure 2). The track data was also used to create the required deposit geometry in the MSC Apex CAE environment (Figure 3), where a congruent mesh of hexahedral elements that aligned with the individual weld beads was constructed. The 316 LSi material properties were selected from the Simufact Welding material database and the heat source properties were set according to OSCAR PLT’s actual DED process parameters. To model the DED process, an advanced element activation scheme was introduced that can provide both stable deposit element handling and more accurate results. The simulation results were the temperature distribution, stresses and strains, and of course the final distortion of the component.
Figure 3 DED sample geometry positioned on baseplate. The Parasolid was directly generated from the weld tracks defined in the robot G-Code. The volume associated with one of the weld tracks is highlighted (orange).
Print with distortion compensation
Shape comparison between the simulated and real parts revealed that the simulation predicted the part distortion accurately, with only a slight overestimation of the distortion magnitude. Both the simulated and real structures bulge inward in the corners of the tube, while the straight edges are less affected. The simulation results were applied to generate the compensated geometry by inverting the calculated distortions. To feed this correction into manufacture, the path planning for the robot was revised for the compensated geometry, and the updated weld paths were utilized to repeat the DED process both virtually and experimentally.
Shape comparison between the 3D-scan of the compensated component and the original CAD data confirms the numerical prediction (Figure 1): with a maximum shape deviation of less than 0.5 mm, the distortion-compensated additive manufactured part (Figure 4) is significantly closer to the CAD geometry than the uncompensated test component.
Figure 4 Final distortion-compensated component. (Credit: OSCAR Plasma-Laser-Technologie)
Residual distortions of the component can be attributed to the following two causes. The first lies in the nonlinear relation between the distortion and changes to the geometry. Experience with numerical shape compensation of powder bed fusion processes has taught us that this issue can be effectively addressed via iterative compensation schemes. The second reason lies in the finite offset between the initial simulation and the real part. Because the simulation slightly overestimated the distortion, the inversion of the predicted deformation also resulted in over-compensation, as expected. Further calibration of the model would have reduced this effect.
This project aimed to evaluate the benefits arising from a model based on default material data and parameters without the need for experimental input other than the robot G-Code. With more than 60% of the distortion eliminated, this ambitious target was achieved. Even in cases for which no initial experimental test-build is available, numerical distortion compensation makes it possible to manufacture an optimized part right with the first build.
Having demonstrated successful distortion compensation for a DED process, the project partners OSCAR PLT, the Chair of Mechanical Engineering at TU Dresden, and Simufact moved forward industry efforts towards the industrial-scale implementation of DED manufacturing and is looking forward to further collaboration.
Acknowledgement
The authors gratefully acknowledge funding from Bundesministerium für Bildung und Forschung and support from Projektträger Jülich (PTJ) within the Agent-3D project ImProVe, Fkz. 03ZZ0210.
