GlobalAM | Enabling Laser Powder Bed Fusion for Large Scale Production of Multi-Material Components

Positioning & Substrate fixation: The state-of-the-art of LPBF process involve printing parts on a bare baseplate, where parts can grow directly from the baseplate top surface or be connected to the baseplate thanks to sacrificial structures called “supports”. much more limited attention in the literature has been devoted to the possibility of printing on top of previously manufactured parts and substrates. This capacity is widely investigated in other processes (like directed energy deposition) for hybrid manufacturing or part repairing. In LPBF the main challenge is the accurate alignment of the part / substrate both in the x-y direction (build area) and along the z direction (build direction). The positioning procedure involves the following steps: 1) placement of the substrates into the build area - this step ensures that the substrate is held securely in place and prevents any unwanted movement during the printing process; 2) alignment of the powder recoating system and set of the zero-level along Z (build direction); 3) fine positioning with respect to the reference systems: four reference systems shall be considered: a) machine coordinate system, b) laser coordinate system, c) camera (or other in-situ sensing method) coordinate system and d) CAD coordinate system: the aim is to align the CAD coordinate system to the laser coordinate system, by comparing the laser position in a predefined number of landmarks in the build area with the actual position of the landmarks identified by means of an additional sensor; 4) build file correction to avoid misalignment: based on the deviation between the laser and CAD coordinates measured by the in-situ sensing method, the CAD will be modified in order to compensate the deviation and maximize the alignment accuracy. Step 3) is the most challenging one: Seminal studies investigated this issue. In some case, extra material is added to compensate for an accurate positioning while in other studies attempts to guarantee an accurate positioning were carried out. No similar capacity is still available at industrial level. The position accuracy objective involved in this project imposes the study of novel techniques suitable to outperform seminal methods proposed in the literature while meeting challenging productivity requirements entailed in the present application.

Innovative solutions will be developed and validated for the two most critical aspects, namely a) the placement of the substrates into the build area to ensure that the substrate is held securely in place and prevents any unwanted movement during the printing process, and b) the fine positioning with respect to the machine / laser reference systems. Regarding point a) investigated solutions will be driven by the following requirements: Ease of installation and removal of substrates into the build plate; avoidance of substrate damage / distortions during installation / removal operations; suitability to install as many substrates as possible within the same build to maximize productivity; avoidance or minimization of post-processing operations on the LPBF part (e.g., for fixture feature removal); minimization of initial misalignment in the x-y direction; Minimization of initial misalignment along the z-direction between the top surfaces of the substrates and the baseplate. Regarding point b) innovative sensor-based methods will be developed, starting from a sensor selection stage. Two scenarios will be explored: 1) Single sensor scenario: in this case, the misalignment between the CAD reference and the laser reference in the
x-y directions will be estimated by means of one single in-situ sensor. Example of sensing solutions to be evaluated to this aim include: a) interferometry (interference patterns used to measure surface height or distance; by focusing a very narrow field of view in correspondence of each reference landmark, a spatial resolution in the order of 1 – 5 microns can be achieved), b) off-axis vision (high-resolution camera that can be used to measure the position of the substrate; affected by perspective errors that must be compensated; very high resolution is need, e.g., >35 Mpixels to achieve a positioning accuracy in the order of 20 microns), c) linear vision embedded on the powder re-coater (much higher spatial resolution can be achieved, in the order to 10 micron/pixel or even higher, depending on the field of view, with no perspective error). 2) Multi-sensor scenario: in this case, multiple sensors are used, and their data are fused to achieve a higher positioning accuracy. This may involve multiple off-axis cameras to enhance the compromise between whole field of view and spatial resolution, or stereo cameras (possibly combined with fringe projection) for enhanced reconstruction beyond 2D vision (as stereo cameras allow for accurate height map measurement). The possibility to exploit the same sensing equipment for both fine positioning (before the process) and in-situ monitoring/control (during the process) is of relevant importance to ease the development of an industrial machine concept.

High resolution residual stress analysis: Residual stresses and stress gradients are of great importance in LPBF processes, especially in hybrid applications by using sensitive substrate with a ceramic core material. Residual stress after cooling leads to cracks in the ceramic layer of the substrate and to delamination issues, as was observed in former experiments. However, the metallic layer of the substrate was still in a very good condition after the printing process. For this project we consider two main impacts of the residual stress to the substrate. Firstly, we address stress directly in and under the printed component, which is strongly coupled to the component geometry, material and process parameters. Secondly, cooling structures, require a big number of similar components (e.g. pins) printed on one substrate enhancing the global warpage effect. Consequently, damages are introduced. Here, the layout of the cooling structure is the main parameter to influence this behavior.

The high-resolution residual stress measurements are essential to understand the relationship between the printed material, process parameters and design/layout. This is required to manufacture hybrid high performance cooling structures, optimized for high cooling performance and a low residual stress level. The impact of new multimaterial
powder and the laser modification approaches (e.g. beam shaping) can be directly quantified by residual stress measurements. Consequently, the generated knowledge between the residual stresses and the application related parameters is very powerful for further applications of this advanced AM technology.

Process & defect monitoring: The layer-wise production paradigm entail in AM enables the capability to “loo ” at each manufactured layer, gathering a massive amount of information about the process stability and the part quality on a layer-by-layer basis, while the part is being produced. This has motivated a continuously growing number of research studies. One goal is to use in-situ and in-line measured quantities, also known as signatures of the process to quickly detect the onset of anomalies and defects (in-situ monitoring / in-situ defect detection). Another goal consists of adapting process parameters and scanning strategies to prevent the formation of defects and/or mitigate their growth and propagation in the build (process control)12. In case defects cannot be avoided, inline defect correction/removal is another capacity that has been explored in the literature. Regarding in-situ monitoring methods, the current state-of-the-art consists of a very large number of studies devoted to several types of defects (porosity, geometrical distortions, microstructural deviations, cracks, and delamination, etc.), involving different sensing methods (cameras, thermal cameras, stereo vision, co-axial pyrometers, acoustic methods, etc.) and different analytical techniques (statistical process monitoring, machine learning, AI-based, etc.). Nevertheless, there is a major gap between solutions explored in the scientific literature and methods currently implemented in industry. Main industrial barriers are: 1) lack of robustness in the presence of actual industrial settings, 2) excessive false alarm rates caused by several natural variability sources, 3) difficulty to implement most sensitive/informative sensing methods on industrial machines, 4) difficulty to deal with big data streams in computationally efficient ways, 5) big data storage needs. In addition, some types of defects still lack sufficiently effective and robust
detection methods: it is the case of internal pores and cracking. These have several random origination mechanisms that are difficult to monitor with the necessary accuracy/effectiveness. The latter originate under the layer, which makes their detection and localization an open issue.

This project foresees the development of innovative solutions to push the boundaries of AM capacities, tackle the industrial challenges and barriers mentioned above. The aim is to combine the innovative and challenging features of the industrial application (multi-material, high-productivity, multi-laser, stringer dimensional accuracy requirements, etc.) with a novel intelligent system paradigm suitable to overcome the limits of methods proposed in the state-of-the-art. Innovative in-line monitoring methods integrated with novel adaptive control strategies will be developed and tested. In context of process & defect monitoring, new solutions for robust insitu/ in-line detection of cracks and thermal stress-induced distortions are foreseen: one specific challenge of the proposed application lies in the tendency to crack formation and thermal stress accumulation of new alloys printed on dissimilar substrates, leading to multi-material properties that are prone to these defects. Robust solutions for the detection of these kinds of defects in compliance with stringent dimensional accuracy requirements is still missing in industry. Innovative solutions that combine cutting-edge sensing capabilities with AI-based techniques will be developed and validated to bridge this gap. More specific, two innovative solutions are foreseen: 1) Robust detection of crack formation under the current layer (in the part and/or in the substrate) through acoustic monitoring and multi-sensor data fusion. The aim is to detect when and where a crack has originated in the build. 2) Layer wise high-resolution imaging to detect geometrical distortions caused by thermal stress accumulation. Novel sensing solutions including very high resolution in the order of 10 micron-pixel or even higher (depending on the field of view) will be investigated, exploiting linear imaging sensors on the powder re-coater and/or one/multiple high-resolution cameras. A key aspect is the possibility to exploit the same in-situ imaging approach for both precise alignment of the substrate (before printing) and in-line defect detection (during printing).

In-line defect compensation: Regarding process control, the number of successful studies in the literature is much smaller compared to the wide literature devoted to process monitoring. In this field, seminal solutions focused on either feedforward control strategies (where local variations of process parameters are embedded in the job file before starting the process, relying on model-based predictions of the process thermal history in every location of the part) or feedback control, where process parameters are adaptively modified in real time based on in-line sensor reading. Feedback control can be implemented track-wise (i.e., in real time during the production of each layer) or layer-wise (adapting process parameters in the i-th layer based on measurements performed in the (i-1)-th layer. Process control is still at a lower TRL than process monitoring. Main industrial barriers are: 1) closeness of industrial controllers, 2) difficulty of computationally efficient data elaboration in real-time, 3) limited reliability/resolution of process modelling and simulation tools.

Concerning defect compensation, a novel optimal process control methodology to mitigate residual stresses and prevent distortions is foreseen: in order to meet ambitious targets in terms of process productivity and part quality, robust defect detection shall be augmented by novel and robust capacities in terms of defect prevention and avoidance. To this aim, a novel approach that combines the in-line mapping of the thermal history in every layer (through pyrometers, thermal cameras, and multi-sensor data fusion) with a layer wise adaptation of process parameters will be developed. The method entails the capacity to adapt key process parameters and scan strategies in the next layer to mitigate the thermal stress onset and accumulation observed in the previous layer. The result is a thermally stable process, with a homogeneous heat accumulation and dissipation, aiming to prevent residual stresses that may finally lead to cracks, delamination, and distortion. In addition, a novel method to augment data driven monitoring and control by means of hybrid modelling/meta-modelling and transfer learning is foreseen: One approach consists of tuning and augmenting the robustness and effectiveness of process monitoring and control by combining the sensor-based data-driven framework with physical and thermo-mechanical models. A hybrid/meta-modelling capacity is therefore achieved using as complementary inputs both in-situ thermal measurements (used as a proxy of thermal stress accumulation) and thermal stress predictions through process
simulation. Having different fidelity and resolution, the two sources of information can be combined to enhance the capability to detect and avoid defects in the part and/or in the substrate. One important aspect of the project is to maximize the industrial impact of the innovations developed. To this aim, a transfer learning approach will be developed and adopted to enable the capacity to generalize and transfer the models developed for one product shape and one material type to other shapes, other materials, and other process settings.

Multi-scale & Hybrid modelling: Trial-and-error-based process optimization requires manufacturing and processing of a huge number of samples which would in turn require an additional long analysis time as well. Such sole experimentation-only method does not help the investigator to find the underlying reason behind a certain phenomenon or defect. Whereas, with the recent progress in the field of computational sciences, it is now possible to develop advanced numerical models to study the detailed impact of process parameters on the final samples’ quality in a very time-efficient and reliable manner while excluding the uncertainty clouds which exist due to experimental uncertainty. So far special focus has been on how primary input parameters i.e. laser power, scanning speed, beam size and hatch spacing would affect the samples’ quality. But the impact of beam shaping as an emerging method to further leverage the process is rather unknown or at best unexplored especially when it comes to deterministic analysis tools such as simulations. The description of effects resulting from the use of multi-material powders is also currently only in a prototype status. In addition, despite the described progress in the field of computational sciences, current simulation solutions are still relatively slow and comprehensive screening of highdimensional parameter space is often still not feasible on economic time scales.

In this project we aim to develop a deposition-scale (melt pool scale) multi-physics simulation to model the impact of different beam shaping scenarios on the melt pool, extended by analyses on surface roughness, bulk density, and mesoscopic residual stresses. To capture the effect of such sophisticated beam shapes on the entire
part’s geometry, we propose developing a part-scale model based on the hierarchical inherent strain method which receives its strains from a thermo-mechanical model at a lower scale i.e. deposition scale. By means of this hybrid model, we will be able to find the most suitable beam shape profile for a specific purpose, i.e. minimized density, controlled cooling, minimized residual stress etc. Furthermore, the developed model is extended to describe multi-material LPBF. In order to enable efficient use in industrial application scenarios, hybrid approaches are also being developed in which reduced order process models are combined with data-based approaches. At the end of the project, a simulation tool tailored to industrial applications will be available, which for the first time can take multi-material effects and different beam shapes into account and delivers results on economically reasonable time scales.