I concluded development and testing of the R2 power module in late December. During 2014 the Miller Lab will conduct research with machine and later in the year we can share more of its capabilities and exciting results.
In January, I started a residency at Autodesk, in their phenomenal Pier 9 facility. I am continuing my research into low cost SLS technology there and am focusing specifically on laser sintering steel. I will be sharing my designs soon.
Early in the week I bolted the module to the laser cutter’s z-stage, focused the laser, and started running my testing scripts. At first I disappointed in the new powder distributor’s performance. I had put a lot of thought into improving upon my previous design. It rides on two linear rails to ensure that it doesn’t deviate from the print plane during distribution and the blade is adjustable to ensure that it deposits powder plane-parallel to the gantry. After carefully observing the distribution cycle for the morning, I began to see a pattern: freshly distributed powder’s pristine surface was marred only when distributor blade returned to its resting position behind the feed piston. Even though I had mounted two slim brushes to wipe the blade clean of any compacted powder both before and after distribution, the powder still was affected by this very light contact with the blade. By changing the powder distribution sequence and docking location of the blade, I was eliminated this problem and was able to get completely repeatable and perfect layers. The old powder change sequence:
Raise feed piston k*[layer height], where k is a multiplier sufficient to account for spill-over losses during powder transfer.
Drop the print piston [layer height].
Transfer powder from the feed piston to the print piston with the distributor blade.
Return the distributor blade to its docked position behind the feed piston.
New powder distribution sequence:
Drop the print piston [layer height].
Move the powder distributor blade from its docked position behind the print piston to behind the feed piston.
Raise the feed piston k*[layer height].
Transfer the powder from the feed piston to the print piston and leave the distributor blade behind the print piston.
I started modeling the next rev of the hardware in late November and wrapped up in early December. I took snapshots along the way to document the process a bit. The new powder module is built around the metal print piston, which supports heating and inert gas shielding. It has been very satisfying to take a couple weeks to design more thoughtfully. I designed the first rev of the powder module in a weekend and its performance and functionality reflect that to some degree. This hardware was composed with openness and accessibility in mind and it will have a very thorough BOM and assembly information.
I have been designing the next revision of the OpenSLS hardware over the past couple weeks and haven’t been doing a good job of posting developments as they occur. In November, I got working sintering parameters for Candelilla wax powder. See below for images of prints fabricated with those settings. While it is possible to sinter this material, due to its brittleness, it is very difficult to remove finished parts from the build platform. Additionally, the microscopic structure of “sintered” wax powder indicates more a fine-scale balling process of sufficient density to begin to bind individual balls to each other. It is a process that likely does fall under the umbrella term “sintering”, but I have observed binding mechanisms that are much closer to those cited in the literature with nylon materials. Consequently, I have shifted my efforts to nylon, which may serve as a better positive control for developing the laser sintering process and hardware as it is a material that has been engineered specifically for laser sintering and will hopefully present fewer fundamental processing problems.
SLS is a much touchier process than either stereolithography or extrusion-based 3D printing technologies. Material cohesion is taken for granted in the latter two– an extruded filament of of plastic is a continuous piece of material, as is a photo-cured region of photoacrylate. In SLS, material cohesion is the fundamental challenge. The parameter space governing cohesion includes beam speed, beam power, trace spacing, layer thickness, material temperature, particle size, laser spot size, and powder packing density. The first four are probably the most important and I have developed scripts to traverse the four-dimensional parameter space they form. My favorite tool so far has been a Python script to traverse the power, speed, and trace-spacing parameter space, as shown in the photos above and the video below. It allows for rapid evaluation of parameter combinations for viable combinations, which can then be tested against several different layer heights to find a robust parameter set for full 3D printing. All of these parameter space exploration scripts are available on my github.
I’ve spent a lot of time carefully observing powder flow during distribution cycles the past couple weeks. The chief problem that I’ve been studying is the surface quality of newly deposited layers in the print piston. To make each new layer, the printer raises the feed piston by some proportional factor to the layer height (I’m using 1.2 currently), the print piston lowers by one layer height, and then the distributor pushes the powder from the feed piston onto the waiting print piston. For a long time I couldn’t get smooth distribution on the print piston even though the surface of the feed piston looked great. By watching the distribution sequence and the powder behavior carefully, I saw that the acrylic wall that divides the two pistons was disrupting the stable tumble pattern that emerges as the distributor pushes the new layer of powder across the pistons. The smooth edge of the acrylic exerts less shear on the bottom edge of the tumbling region of powder, disrupting the somewhat stable dynamics of the region (driven by shear against the static powder plane) and the leading edge of the acrylic may apply a slight compacting force to the powder. The powder tumbling slows as the powder moves (slides) across the acrylic. As the distributor crosses over this acrylic edge, the tumbling powder encounters a slight drop to the surface of the print powder plane, the low density leading edge of the tumbling region falls unevenly to the print plane, exerting uneven shearing forces on the remaining feed powder and preventing even distribution across the print surface. Additionally, the friction and pressure between the sidewalls and the distributor edge cause excess powder to fuse and form large flakes that fall into the feed powder and create large tracks in the powder during distribution.
To address both of these problems, I affixed thin rails to the sidewall edges, reducing contact area between the distributor and the sidewalls and raising the distributor 1.5mm above the acrylic divider wall. I used solid core 22 AWG hookup wire, which I measured to have a diameter variation of 0.01mm, and secured it into the epoxy using weighted flat surfaces. This dramatically improved powder distribution quality and reliability. I really should have followed my original design more closely, in which I solved both problems by using a sharp chamfer on the edges of the pistons.
The counter-rotating distributor that worked so well for sucrose didn’t work at all with the wax powder. The wax clumped and compacted against the anodized aluminum surface of the drum, completely eliminating any chance of getting even layers. I reverted to my old geometry of a 90 degree wedge at 45 degrees to the powder plane. I coated a piece of angle aluminum with Kapton (I used Teflon tape last time) and found the new distributor worked far better than the counter-rotating one, but still distributed pretty uneven layers.
In an effort to consolidate z-control to the actual Z-axis channel on the RAMBO, I swapped the 3mm pitch leadscrew on the print piston for one with 1mm pitch. The idea was to drive the two pistons in parallel, but in opposite directions on the Z-channel to hardware hardcode the “overfill” ratio between the feed and print piston at a factor of 3. I quickly learned that while entirely functional, it is too early for this kind of mechanical optimization. It turns out that the overfill ratio needs to be carefully calibrated based on the layer thickness, material, and size of the previously sintered region. The tumble dynamics of the powder change significantly with the quantity tumbling and change the density and packing of the deposited layer. I scrapped my plans to drive the pistons in parallel, but left the 1mm pitch leadscrew in the print piston. It is important to maintain individual control over both pistons while exploring multiple materials and sintering parameter sets.