Accuracy and repeatability are key for product features measuring 30μm or less
Selective laser sintering (SLS) is a broad topic with many different implementations. In the most general sense, it is the act of selectively turning a powdered material into a solid structure without first turning that powder into a liquid. In practice, the phase transition from powdered solid to liquid mass to solid mass does happen, but very quickly, because of the speed with which an industrial laser allows for the application and dissipation of large concentrations of thermal energy. To leverage this technology for the purpose of manufacturing, a system must be able to accurately deliver both the metal powder and the laser power to the same locations over a three-dimensional (3D) structure. Positioning the powder and the laser require high motion accuracy as feature sizes continue to decrease to make more precise products (FIGURE 1).
|FIGURE 1. A complex 3D object produced by selective laser sintering.|
As product features move toward a limit of 30μm or less, all of the elements of a SLS system are critical to allow those features to be realized in an accurate and repeatable fashion. For example, there are two main types of powder metal delivery used in SLS. The first is a powder bed that continues to roll more powder material over a part that is being processed. The second is a nozzle head with powder sprays and the laser delivery head mounted together and synchronized for the sintering process. For either one of these processes to produce laser sintering on the micro level, the powder must be sized to accomplish this, the laser spot size must be small enough to create a small-enough heat-affected zone, and the motion system must be accurate and repeatable enough to be in the right location for each and every sintering event. For the purpose of this article, we will discuss the considerations one should make in creating a motion system capable of providing just that type of motion for micro-selective laser sintering.
Good motion system design begins with the end in mind, meaning that the types of moves required must be known. What acceleration, velocity, accuracy, repeatability, and tracking error are required? For this stage of the evaluation, the mechanical solution, whether gantry, XY galvo scanning head, simple stacked-stage solution, or any combination, need not yet be considered, as the move specifications will provide guidance into the optimal mechanics. Let's imagine that this micro-selective laser sintering process requires single-digit micron accuracy at the work point. Let's also imagine that this is a powder-bed application, not a powder-nozzle application. This will help limit the scope of our discussion.
Machine base assembly
We will start from the bottom up, beginning with the machine-base assembly. The first consideration is to design a machine base that is adequately stiff, heavy, and capable of rejecting ground disturbances that can cause small oscillations at the work point (FIGURE 2).
|FIGURE 2. Granite is the performance leader for SLS base plates due to its superior flatness-having a flatter base increases the performance of a motion system.|
Considerations include proper leveling feet, a stiff metal structure, and a proper type of isolation system between the machine base and the base plate. For some precision applications, it may be necessary to characterize the ground vibrations for the location at which the machine will ultimately reside. Knowing this gives a machine designer insight into which frequencies of oscillation their machine must ultimately reject. This, however, is not often required, although sometimes machine builders give recommendations for ideal ground floor characteristics. Most often, machine builders will design a system for the most common situation. Isolation systems vary in complexity, from simple passive dampening material, passive air isolation, active air isolation, or isolation through some custom system employing another fluid. Whatever the choice may be, the consideration here is the rejection of unwanted frequencies.
Once a preliminary base design is completed, the next design decision is the type of mechanics for the actual motion system. For this example, imagine a 300 × 300 × 150mm XYZ motion envelope over a powder bed with a special powder addition and leveling system below. The work envelope requires three dimensions of motion and must span the powder bed. The system design will consist of linear stages in all three directions (X, Y, and Z) and will be a traditional gantry configuration (FIGURE 3)—meaning dual axes carrying a single bridge axis. It is important to note that a controller can only position a stage to be within some accuracy specification limited by the encoder's resolution, the mechanical resolution of the bearings, and the location of the encoder with respect to the work point. The first limitation, the encoder resolution, is typically overcome by using well-tuned sine wave encoder feedback. A good controller will have the ability to tune the sine and cosine signals by applying offsets and scaling factors to the typical 1Vpp signals. This corrected signal will be sampled and interpolated by the controller well into the nanometer level. The choice of bearings for the gantry stages will limit the ability for that very fine electrical resolution to be realized—the less friction in the bearings, the closer to the electrical resolution that a motion system can approach. For this reason, many high-accuracy machine builders consider air bearings. However, air bearings will not work in a vacuum environment, can be subject to contamination by the powder, and do not have the same stiffness as mechanical bearing solutions. Finally, the location of the encoder relative to the work point is important due to angular errors (pitch, roll, and yaw) of any linear stage. The further the work point is from the bearing and encoder system, the more greatly exaggerated these angular errors will be.
|FIGURE 3. A gantry-style motion system that incorporates an XY galvo scanner.|
Fortunately, there are ways that high-end motion providers can get around some of these errors. For example, for the gantry system under discussion, parallel gantry base axes or "spars" on either side of the powder bed carry a single gantry bridge (also linear motor-driven) over the top of the powder bed. These parallel base axes may not be perfectly orthogonal to the bridge axis, but the right controller choice can "force" orthogonality by applying an offset to one of the base axes or "spars." This assists in minimizing angular errors at the work point. Additionally, each encoder must be calibrated such that the highest level of accuracy is achieved at the work point. This calibration procedure involves placing reflective optics mounted to the XYZ motion system at the laser work point. A laser interferometer is used as the master position reference as the axis being calibrated is moved along its travel. Calibrating an optical linear encoder with laser interferometer feedback at the work point will increase performance in two ways. First, it corrects for the magnitude of error along the axis being calibrated due to angular error motion of that axis, which is magnified by the distance between the encoder and the work point. It should be noted that the angular element of that error vector is not corrected. However, with symmetric Gaussian or top-hat laser spots, the angular aspect of this error is not an important consideration. Second, it increases the raw linear accuracy imparted by the stage at the work point. Moving to multiple positions along the travel of the stage tracks the differences between the native encoder and the more precise interferometer feedback. The difference in measurements between the two devices is used to generate a calibration file. A controller will use this calibration file to ensure that a commanded motion of, for instance, 10 mm will be as close as possible to 10 mm at the work point, not necessarily at the encoder.
Now that the precise, calibrated XYZ motion system has been designed for use over the powder bed, it is decided that the velocity requirements to meet the process throughput goals cannot be accomplished with the linear motor drive-train of the XY gantry. The scanning speed needs to be a few meters per second, so XY galvo scanning heads are chosen as the proper tool for fast beam steering. The only issue is that the galvo scanning head has a field of view of 100 × 100mm while the part is 250 × 250mm in XY size. Also, the accuracy of the scanning head is ±50μm, and that is after a theoretical f-theta lens correction file is applied.
There are several ways to address the limited field of view of galvo scanning heads. The first is to implement a step-and-scan technique. This technique means that the very high-speed moves are made with the galvo scanning head, followed by indexing steps with the servo axes. During scanning, the servo axes hold the galvo scanning head in position. Using this method, process moves that are longer than the galvo field-of-view (FOV) are "stitched" together at the edges of each FOV. Also, one must consider that the scanning mirrors move the laser beam towards the extents of an f-theta lens, spot size distortion occurs, and higher levels of positioning error also occur. For this reason, it is desirable to find an industrial controller that can blend servo and galvo motion together, which has two advantages. The first is eliminating any stitching errors. The second is maintaining quick processing times while limiting the FOV of the galvo scanning axes to the more accurate and less distorted central region of the lens. Slower servo moves under a galvo scanning head constantly "re-center" the part, which allows for this desirable behavior.
Now, it is time to review the mechanical design. At this time, the system includes a machine base with proper leveling feet and isolation, an XYZ gantry motion system with an XY scanning head, and some of the process requirements have helped in narrowing our controller choices. It is time to determine the base plate material and finally revisit the machine base design. All parts of a system must work together, so it is important not to forget that the machine base and the motion equipment must work seamlessly.
The base plate
Base-plate material can significantly affect machine performance. Typically, the choices are steel, aluminum, and granite. Often, granite is the performance leader due to its superior flatness—having a flatter base increases the performance of a motion system because flatness errors are transferred to the stages that are mounted to the base (FIGURE 3). This is due to base plates typically being designed to be less stiff than the equipment which they carry. Negative points about granite are the cost, difficulty to work with by requiring special inserts vs. drilling and tapping holes, and having a different coefficient of thermal expansion than the mechanics mounted to it. Ultimately, this decision will be performance-based, but can also come down to costs and availability. For example, if procuring a thick aluminum plate with very tight machined surface specifications at a low price is possible, granite might be avoided. Revisiting the machine base design ensures that the proper support structure exists for the motion system on top.
Now, it's time to evaluate motion controllers. Realizing all controller features required to operate this mechanical system has narrowed our choices. Some key considerations include: What type of controller can manage five axes of coordinated motion? Can the controller continuously process a large part and, if so, how difficult is that part to program? How will the controller interface with the industrial laser? Can the laser be pulsed based upon distance and/or constant velocity? How will the pulse width be controlled? How will the controls required for machine functionality interface with those required for processing? Correctly answering these questions with the right controller will determine whether or not the machine will operate as envisioned. For this reason, it is important not only to understand the interaction of the powder material with the laser, but also to understand how that interaction will manifest in 3D space.
In light of that, coordinating galvo scanner motion and servo motion together for the purpose of eliminating stitching, maximizing throughput, and processing in the center of the f-theta lens are all the goals. The laser spot size and pulse must be consistent, and distortion will be hard to tolerate. Stepping and scanning has the undesirable impact of starting and stopping what should be continuous process moves. Also, using a galvo scan head allows for some unique features such as quickly oscillating the laser (wobble) to create "thicker" part paths. The very high acceleration of the galvo scan head allows for more effective cornering operations, which is helpful when parts have many changes in direction. If these changes in direction exist, timing laser pulses based upon constant velocity will be very difficult. A controller that can track distance traveled and fire pulses based upon vector distance may be helpful in implementing this process. The goal is to maintain a consistent structure in this micro-selective laser sintering process.
A final look at the galvo scan head and f-theta lens combination reveals that the accuracy is ±30μm over the central 25 × 25mm region of the lens, even after a theoretical correction table is applied. The motion is very repeatable, just not accurate enough. Is there a way to calibrate this region to achieve more accurate results, based upon real-world measurements? The accuracy of the galvo scan head/f-theta lens system is in addition to the servo system that carries it, so the less error, the better. How can the galvo scanner be calibrated not only for the highest level of accuracy, but also to be aligned with the servo axes that carry it? There are still a lot of questions.
Moving from rapid prototyping and conceptual designs manufactured by SLS and into the 30μm and below micro-selective laser sintering processes requires a new set of considerations. Many of these considerations are straightforward. Smaller powder particles to make smaller features are needed. Smaller laser spot sizes to make smaller features are required. The ability to control the pulsing of the laser to control energy input is another requirement. In concept, these requirements are easy to understand.
In contrast, the motion system proposed in this article has a total of five moving axes, each contributing its own unique errors. Although many items were discussed, many more were not addressed. These include managing the leveling of powder beds between layers; managing the consistency of powder beds, as powder material behaves differently as particle size decreases; using a nozzle-style powder distribution technique; maintaining normalcy with a nozzle-style delivery; and the ability to program a part profile in three dimensions and to execute the profile with five moving axes—the list goes on. For this reason, it is important to consider your motion system requirements so that the process manifests itself properly at each event in 3D space. As these motion requirements become more technical, it is important to realize which of those tasks can remain in-house, and which require a partner that can guide you in the right decisions.
PATRICK WHEELER(email@example.com) is senior applications engineer with Aerotech, Pittsburgh, PA.