Ultrasonic Levitation as a Handling Tool for In-Space Manufacturing Processes

In recent years, plans for long-term lunar and Mars missions have regained international importance. In-space manufacturing (ISM) represents a key competence for these missions, offering a time- and cost-efficient alternative to the otherwise required, logistically demanding deliveries of components [1]. ISM encompasses production processes in transit and orbit as well as on the surface of extraterrestrial astronomical bodies. In conjunction with in situ resource utilization, this results in even greater independence and sustainability [2]. Due to the reduced amount of raw material required, additive manufacturing (AM) plays a particularly important role in this context. However, under conditions, 3D printing techniques like powder-based selective laser melting (SLM) exhibit the problem that neither the feed material nor the printed parts are held in place by gravity. Consequently, the necessity arises for the development of a method that will enable the restriction of both by other means.

In regard to this problem, we propose the use of acoustic levitation as a tool for containing and manipulating materials and objects in microgravity (μg)-production processes. Even though we find this technique suitable for a whole range of handling applications, in this article, we focus on SLM as an exemplary process. Acoustic levitation is a technique that utilizes sound waves to manipulate powders, liquids, and solid objects in a contactless manner. The manipulated matter can thereby be suspended in mid-air or other fluid media. It is particularly useful when handling delicate objects or if surface contamination is undesirable. In the most basic form, with objects that have a higher density than the surrounding wave propagation medium, the objects are trapped within the nodes of a standing pressure wave [3]. However, the size of particles that can be manipulated in this way is limited to about half the wavelength (⁠$λ$⁠) of the sound wave [4].

Initial preliminary tests on the feasibility of the use of this technique have been carried out and were successful. The objective of this project is, therefore, to set up an experimental platform with which we can further investigate the proposed method and demonstrate its functionality. To test the process under μg conditions, the setup will be placed in our third generation drop tower facility [5–8], which allows for about 4 s of weightlessness per flight. In order to manipulate the target matter in all three translational, and, where possible, also the three rotational degrees-of-freedom (DOFs), a system with a configuration of two opposing planar transducer arrays was chosen for the setup. The ultrasound waves emitted by these arrays can create different types of sonic fields. By using phased arrays, in which each transducer can be controlled individually, these fields can be focused and steered by phase shifting the different transducer signals. A serial manipulator positions the test objects in the correct location within the acoustic field of the ultrasonic levitator at the outset of the μg phase during each flight.

In order to implement this process, the acoustic pressure gradients and their effect on the spherical particles must first be studied and understood. This can be done by modeling the resulting ultrasound field from all the transducers using continuum methods like finite element method (FEM) or computational fluid dynamics (CFD). These are then coupled with a discrete element method (DEM) to study the effects of different fields on individual particle trajectories [9,10]. Subsequently, by integrating a control strategy between field influence and particle position, the particle motion and, thus, the resulting position should be specifically influenced. The application of AI methods in particle technology is already established and will be used as intelligent control by integrating a deep reinforcement learning (deep RL) approach [11,12].

The method has a significant potential for a range of applications in space and on Earth. The use of noncontact, minimally invasive, and flexibly controllable handling processes under μg and, in the future, under all possible gravity conditions will enable the development of new research and manufacturing processes. As an alternative to ultrasound, magnetic [13] or electric fields [14], as well as air currents [15], can be employed for levitation. Ultrasound offers the simplest setup with numerous small, cost-effective, and flexibly controllable actuators, rendering it the optimal choice for this application. Moreover, it offers the greatest possible material compatibility [4] (liquids, dusts, solids) regardless of their electrical or magnetic properties and additionally offers sufficient dynamics.

Ultrasonic levitation is not a novel concept in the field of science. Various techniques for creating acoustic traps have been developed, including the use of delay-lines or metamaterials [16–18]. When manipulation of the levitated object in more than one dimension is necessary, some kind of controllable phased array is required. For that, there are well-described do it yourself (DIY) setups, open-access simulation software, and control programs [19–22]. The limitations here lie in the required DOFs, and that there is, as far as we know, no provision for feedback with external sensors yet. Ezcurdia et al. used such open-source levitators to contactless fabricate structures with UV-curable glue and elongated parts [23]. However, the limitations of Earth’s gravity meant that only small and lightweight structures could be fabricated. There have also been early experiments using acoustic levitation as a containerless handling tool in μg [24–26]. These experiments, however, were limited to specimens below the wavelength limitations and were focused on droplet shapes and high-temperature specimen handling rather than on the fabrication of bigger objects in an AM approach. To achieve higher accuracy and lower latencies, there have also been efforts to precompute acoustic fields and interpolate between those discrete points in space [27,28]. Of course, there are also various other approaches for powder-based AM in space [29–31]. The physics and governing forces for different particle size regimes have been extensively discussed in numerous papers, including Refs. [3,32–37], and will therefore only be briefly addressed in this publication.

This article describes our concept for the contactless handling of objects and powders in μg for AM in space. First, we present the preliminary feasibility tests that we have conducted. These tests include the handling and melting of powders under Earth conditions as well as the handling of objects in μg. We then describe our goals in detail and the experiments, hardware, and software solutions that are necessary to achieve those goals. Finally, we provide a brief summary and outlook on potential future experiments and projects.

The conventional SLM process for powder-based materials heavily relies on Earth’s gravity, rendering it inapplicable in μg. However, our preliminary experiments have demonstrated that acoustic levitation can be used as a fixtureless handling mechanism for laser processing SLM material. These experiments and their results are described in greater detail in a separate publication [38]. Here, we will include only the basic setup and the main results as they are essential for the following experiments.

The Gor’kov potential $U$ is primarily influenced by the radius $R$ of the trapped spherical object. The radius is multiplied by a difference that contains the speed of sound $cs$ and density $ρs$ of the sphere as well as the density $ρ0$ and speed of sound $c$ of the propagation medium. The values for $f1$ and $f2$ are approximately 1 if the levitating object has a high density and, therefore, a high speed of sound in relation to the propagation medium (⁠$ρ0≪ρs$⁠). The pressure $p$ and the particle velocity $u$ describe the acoustic field. In this context, the mean square deviation $⟨p2⟩$ and $⟨u2⟩$ are time independent in a harmonic field, yet position dependent.

Therefore, capturing and levitating powder-based material has been shown to be difficult due to its small radius particles and, thus, low Gor’kov potential. To increase the outer surface area and, consequently, the levitation force, water has been used as a carrier medium. The conceptual process flow is depicted schematically in Fig. 1.

To prepare for the melting process, a mixture of SLM powder and water is made. Polyamide 12 (PA 12) was selected as the powder due to its nonreactivity with the carrier material. A pulsed fiber laser with a maximum average output power of 20 W and a wavelength of 1064 nm was used for both the evaporation and melting processes. Initially, the water is evaporated using 4 W laser power. Previous experiments with pure water have shown that the addition of PA 12 significantly accelerates the evaporation process. After the evaporation stage, the melting process begins by increasing the laser power to 20 W. This process results in a pearl of PA 12, as shown in Fig. 2. The pearl has a smooth surface with some unmelted powder particles attached. Its diameter is approximately $600μm$⁠, and its volume is about 0.113 mm³, as determined from the microscopic image.

As the next step, we investigated the behavior of levitated specimens under μg. For this purpose, we modified our levitator based on the TinyLev system from Marzo et al. [20] to withstand the high accelerations of the Einstein-Elevator during the start and breaking phases. A short description of the Einstein-Elevator and its working principle can be found in Sec. 4, in which we discuss the setup and methods of the planned continuative experiments. Inside the levitator, we placed little Polystyrene spheres with an approximate diameter of 4 mm in the nodes near the focal point of the standing wave field. The diameter corresponds to about half of the wavelength (⁠$λ=c/f≈8.5mm$⁠), resulting from our 40 kHz transducer excitation and a speed of sound of about 343 m/s at standard sea level conditions. To be able to place the spheres in the acoustic traps before the launch, we had to orient the levitator with its main axis vertically, i.e., in the direction of flight. Even though the orientation does not matter during the μg phase, the spheres would otherwise not stay inside their traps during the acceleration. This is due to the fact that the velocity gradient forces acting on the levitated object orthogonal to the sound’s propagation axis are much weaker than the acoustic radiation force (ARF) parallel to the propagation axis. The setup with three levitating spheres under μg conditions can be seen in Fig. 3.

Substrate-free production is our vision of writing components directly into space without the need for support material. To achieve this goal, a precise manipulation of the component’s position beyond the aforementioned wavelength limitation must be realized. When moving larger and heavier components with a characteristic length bigger than half the acoustic wavelength (⁠$l≻λ/2$⁠), there is initial progress. For example, Ueha et al. describe in their work that objects up to 10 kg can be levitated in the acoustic near field [39]. However, this is limited to planar specimens with a low weight-to-base area ratio and is thus not suitable for our application. This is because the print parts can have complex 3D shapes that have, depending on their orientation, very little area close to the emitters. With more complex acoustic traps, the levitation of objects beyond the wavelength limitation has also been demonstrated in the far field [21,23]. Inoue et al. were able to levitate a sphere with a diameter of 30 mm (approximately 3.5 $λ$⁠) and a mass of 600 mg [34]. However, due to the gravity on the Earth and the limited power of the needed transducers, objects with increased mass are difficult to levitate. In space, this poses less of a problem. The use of a μg facility like the Einstein-Elevator at the Leibniz University Hannover should make it possible to manipulate components of higher mass and size. Because the components are already levitating, the levitator is only responsible for the movement of the parts but not for providing the force required for levitation. Therefore, our experimental system will be initially set up under μg and later for less heavy objects under reduced partial gravities. This can be done in the Einstein-Elevator as well since its gondola is linearly actuated over the whole length of the vertical parabola. This allows us to test the substrate-free printing process under lunar and Martian gravity as well. In addition to the handling of individual components, the positioning of smaller objects in the vicinity of a larger object must be considered. The challenge here is to understand the influence of the larger object on the surrounding acoustic field, as it might distort the wave pattern through the scattering of the sound waves on its surface. Furthermore, to achieve substrate-free production, the melting of the raw material and its fusion to the print part in a more or less continuous manner and also sufficient resolution to compete with other AM processes has to be researched. The technical objective is, therefore, the mechanical design of an experimental setup that allows the study of these different aspects. The setup must withstand the challenging environmental conditions of the drop tower system, such as the high acceleration at the beginning and end of the vertical parabolic flight.

A second challenge, that has to be tackled, is the need for the ability to manipulate the position of multiple objects with different sizes and phases in the same field and fuse those in the correct position. These objects are the print part itself and the molten raw material, which is to be added incrementally to the structure. In the case of powder-based AM, the print material is at first trapped in a liquid carrier medium and then molten to a liquid to be able to fuse it to the print part, as described in Sec. 2. The precise handling of the various components in different states and positions at all times during this process is a major challenge and requires a deep understanding of the interaction of ultrasonic fields with the material. Extensive series of tests, which can be costly and time consuming, especially in μg, can be replaced by coupling continuum methods such as CFD or FEM with the DEM for discrete elements [9,10]. The process modeling must also be extended by a control strategy, such as deep RL or nonlinear model predictive control, to be trained in the virtual experiment. This approach automatically executes actions in the simulation environment, observes the results, and performs an evaluation according to the target function based on rewards or punishments to gain the most accurate understanding of the process [12]. The control strategy will be tested and validated in real experiments, with the possibility of a feedback loop for calibration. There, the position and spatial orientation of the objects are detected by a computer vision system, which transmits the information to the control system, which then controls the transducers. The development of such a control system is the second major goal of this project.

Like the objectives in Sec. 3, the experimental procedure can be split into two main topics: the experimental setup and the simulation and controller. The Institute of Transport and Automation Technology (ITA) of the Leibniz University of Hannover (LUH) focuses thereby on the experimental setup, while the Institute for Particle Technology (iPAT) of Technische Universität Braunschweig (TU Braunschweig) focuses on the control strategies.

The movement and orientation of the particles and components in μg within the framework of the project are realized with the help of a planar levitator with a grid of 16-by-16 transducers. Due to its flexibly controllable individual actuators, this enables three-dimensional investigations and the realization of the required novel trap/field geometries. The individual ultrasonic transmitters (actuators) are controlled using a complex circuit developed for this purpose by Morales et al. [22]. Since a majority of experiments in the literature use transducers with a resonance frequency of 40 kHz, we do so as well to maintain good comparability. In contrast to the preliminary experiments, the main axis of the acoustic levitator is not aligned in the direction of flight but horizontally, as can be seen in the schematic of the setup depicted in Fig. 4. This is due to the otherwise possible damage of the transducers by the falling print parts during the braking phase of the Einstein-Elevator. During the laboratory tests in the usual 1 g environment, the arrangement remains vertical, as in the preliminary experiments.

In order to examine the movement and orientation of larger components in the Einstein-Elevator, it is necessary to fix the component in its starting position during the acceleration phase of the drop tower facility. Once the μg phase is reached, the component should be released and the investigation was carried out. A gripper system with three lateral DOFs is being employed for this purpose. This apparatus comprises a robot arm with multiple movable joints and a gripper as an end effector. This serial manipulator is installed on the side of the levitation system. In the breaking phase of the Einstein-Elevator, the before-levitating component will drop to the chamber floor. After a flight has been carried out, the robotic system is used to return the component to its starting position.

The particles of the print material are supplied by means of a small actuator that has to be developed in-house. A flow-driven conveyor is not a possibility here since the airflow would interfere with the acoustic field and the print part [40]. In this case, a piston-based feeder is likely to be used. The selection of the print material used and its properties regarding size still have to be determined as part of the project. The supplied plastic particles are fused with a diode laser (1064 nm). To do so, the component is heated selectively on the side facing the particle stream. The power of the laser (30 W) can be adapted to the material and the melt pool size. The settings of the laser system for needs-based melting are determined in laboratory tests before the μg experiments. Due to the handling of the powder via a liquid carrier media, no unfused powder should be produced. However, since anomalies and print failures are unenviable, the inner printer parts have to be shielded from unwanted drifting particles. Furthermore, a process of easy excess powder removal has to be developed.

In order to protect the entire system consisting of the levitator, the gripper system, the laser source, the particle supply, the tracking system, and the camera systems from external influences, they are installed in an experimental chamber, which is mounted inside the experiment carrier of the Einstein-Elevator. To avoid disturbing reflections of sound emitted by the ultrasonic transmitters from the walls, they are covered with sound-absorbing material. Free-flying particles have to be kept within the experimental setup and must not disperse in the experiment carrier or the gondola of the Einstein-Elevator. In addition, the use of the laser system (laser class 4) requires the creation of a radiation-safe environment. The large components that fall during the braking phase of the Einstein-Elevator must be contained as well so that they can be repositioned by the gripper system after each flight. Therefore, the chamber also serves as a container that prevents these various emissions from the setup to the outside.

For the experiments in μg, the experiment carrier with the entire setup is placed inside the gondola of the Einstein-Elevator, which is depicted in Fig. 5. After sealing the gondola hermetically, a vacuum between the gondola and the experiment carrier shell is drawn. This ensures an acoustic decoupling between the surroundings and the experiment. This is necessary to achieve μg qualities of up to $10−6g$⁠. After positioning the probe objects in their designated positions in the acoustic field by the serial manipulator, the linear drive of the elevator accelerates the gondola and, therefore, the experimental setup with about $5g$ over a distance of approximately 5 m. After that, the acceleration is set to zero and the experiment carrier separates from the gondola floor. The experiment is now in free-fall while following its trajectory of a vertical parabola, first 20 m up and then 20 m downward before decelerating again over 5 m to a velocity of zero. A flight diagram of the position, velocity, and acceleration of an ideal flight is depicted in Fig. 6.

As described in the project objectives (Sec. 3), three main experiments have to be conducted: first, the manipulation of objects larger than multiples of the sound’s wavelength; second, the manipulation of smaller objects in close proximity to a much bigger object; and third, the continuous feed and fusion of melted material to a bigger object. As the period of weightlessness in the drop tower is limited to approximately 4 s, the manipulation experiments have to be carried out separately.

Figure 7 illustrates the development of a control strategy based on artificial intelligence using a simulative digital twin of the experiment. The first step in the project is to perform an acoustic field calculation of the respective transducers. For this purpose, already developed tools will be adapted to our digital twin [42]. The acoustic field simulation is intended to be able to model, e.g., the change in amplitude and phase shift, and also be designed for μg. In addition, a DEM simulation environment is used, which fundamentally models the mechanical behavior of a particle numerically [43] and can also take into account particle–particle interactions as well as powder and sinter behavior [44]. The effect of acoustic radiation on small particles is of particular interest for this project, for which a one-way coupling must be implemented. The long-term goal of the general research project would also be to develop a two-way coupling so that the interaction of objects, especially larger particles and components, with the field is modeled. After the development of the digital twin, sensitivity and limit range analyses will be carried out with the aim of developing an automated simulation process. At this point, a control strategy such as reinforcement learning (RL) can be applied. RL is the automation of goal-directed learning where the agent interacts with the environment while being rewarded or punished. The goal of RL is to find a policy, a mapping from state to action, that maximizes the cumulative rewards. Two major approaches are as follows: first, policy based, which directly searches for the optimal policy for the maximum future reward; second, value based, which estimates the optimal value function that represents the maximum value that can be achieved under any policy. Because of the high dimensionality of state-action pairs, storing these functions or policies is a challenge. Therefore, function approximators such as neural networks are used in RL. Deep neural networks, effective in image, speech, and language tasks, approximate value functions or policies, leading to deep RL, which perfectly fits as a control strategy in our process [12]. In our case, the movement of the particles to a desired coordinate by adjusting the field forces is the leading policy. The value function could then consist of minimizing the path length to the coordinate or identifying the most energy-efficient solution possible. The developed deep RL agent is trained and optimized in the virtual environment, so that it can be used for the experiments once it has been calibrated and validated.

The particle positions are transferred via a camera system, and the resulting control is then performed by controlling the transducers. Based on these results, a free three-dimensional movement of particles in space should be made possible, allowing precise placement of the material to be welded for ISM. The camera tracking system is intended to not only provide the positions, velocities, and inertia of the particles to be fed back into the simulation and control model but also for process visualization and analysis. In addition to the metrological evaluation of the component and particle movements, two further camera perspectives are planned. These are used for visual inspection of the process and process logging. Here, we rely on a robust, space-proven, and yet cost-effective camera system. Furthermore, the system will support all other processes, including remote control of the gripper system, functional check of the laser, handling of components and particles outside the acoustic levitator, and control of the particle feed.

The development of this novel system and its evaluation with regard to the realization of movements and orientations of different-sized components/particles represent the core of this part of the project. The exact design of the described setup is also to be developed as part of the project.

Experiments were conducted to test the general feasibility of using ultrasonic levitation as a handling tool for substrate-free AM in μg. These preliminary experiments have shown that it is indeed possible to manipulate polymer powders by means of acoustic fields while melting the levitated material. Further experiments in our drop tower facility showed that we are still able to manipulate particles with acoustic traps under μg conditions. We then described our plans to extend those experiments with a bigger acoustic levitation system that is capable of manipulating objects in three dimensions. For these extended experiments, we want to accomplish the following basic tasks:

1. move and orient large components in the levitator beyond the wavelength limit in μg

2. manipulate granular raw material in the same field as the print part while melting it

3. fuse print material to the print part in a more or less continuous manner

To achieve these objectives, an intelligent control strategy has been presented for application to the experiments. The described strategy is based on the results of a digital twin and allows for the free three-dimensional movement of the objects and particles, enabling the precise fusion of the material for melting during the ISM process.

Should the aforementioned experiments prove successful for basic spherical shapes, a subsequent step could be to further enhance the system in order to extend the manipulation of larger objects to the three rotational DOFs. This would permit a greater degree-of-freedom in the design of print parts and would enable the fabrication of highly anisotropic components. However, the implementation of a bespoke slicer and the completion of further experiments would be necessary. It may also be necessary to conduct the described experiments over a longer period of weightlessness. Consequently, it would be beneficial to conduct further experiments during a parabolic flight. Despite the reduction in μg quality, the extended duration of the flights would allow for the acquisition of new insights. Another potential avenue for future research could be the utilization of magnetic levitation as a tool to orient metal parts in μg in a similar manner. This approach could offer advantages in terms of the forces and moments that can be applied to the print parts, resulting in higher possible print speeds. In conclusion, the advances of these ISM techniques will ultimately benefit future long-term space missions, as they will result in increased autonomy and sustainability.

The authors would also like to thank the DFG and the Lower Saxony state government for their financial support for building the Hannover Institute of Technology (HITec) and the Einstein-Elevator (NI1450004, INST 187/624-1 FUGB) as well as the Institute for Satellite Geodesy and Inertial Sensing of the German Aerospace Center (DLR-SI) for the development and the provision of the experiment carrier system.

German Space Agency within the German Aerospace Center (DLR) and the Federal Ministry for Economic Affairs and Climate Action (BMWK) on the basis of a decision by the German Bundestag (FKZ: 50WM2341A).

There are no conflicts of interest.

The authors attest that all data for this study are included in the paper.

$c$ =

speed of sound

$d$ =

diameter

$l$ =

characteristic length

$p$ =

pressure at a given location

$u$ =

velocity of a propagation medium particle at a given location

$R$ =

radius of the levitating object

$U$ =

acoustic radiation potential (also known as Gor’kov potential)

$λ$ =

wavelength (mm)

$ρ$ =

volumetric mass density

$0$ =

sound wave propagation medium

$s$ =

levitating sphere

g =

Earth’s gravity (⁠$1g=9.81m/s2$⁠). It is written in italics to avoid confusion with the unit gram (”g”, not written in italics).