Materials will be the biggest challenge in this project. Last time I built a laser sintering machine, I spent about two weeks on hardware development and about two months on material development. I learned a lot from that experience and am applying those lessons now in how I explore processing options and testing strategies.
Ball-milling is a common method for producing powder form brittle materials such as ceramic minerals and crystalline chemicals. They are mechanically very simple systems: a drum filled with dense, hard balls and the material to be pulverized is rotated steadily, causing the heavy balls to continuously fall against the drum walls, progressively crushing the substrate material into finer and finer particles. In the spirit of repurposing, I purchased a rock tumbler and loaded it with stainless steel balls and isomalt grains. Early experiments have been promising and I am scaling up to a 12lb capacity tumbler to produce enough material to run in the SLS machine.
My experiments with cryo-milling two years ago provided a method for powdering paraffin and I am now trying to adapt the process to produce PCL and possibly PLA powders. Lowering the temperature is essential for these materials because they have such a propensity to absorb and dissipate mechanical energy through plastic deformation rather than brittle fracturing. By dropping the whole milling system to well below freezing using liquid nitrogen or dry ice (more available), you can embrittle the material and use the ball-milling process to fracture and pulverize it. At least in theory. In humid environments, such as a lab in Houston, condensation on the drum creates a slick surface that stalls the tumbling. The new, bigger tumbler is constructed in such a way that I don’t think this will remain a problem. However, the safety considerations necessary for cryo-milling will require serious attention.
A brief test of the powder distribution mechanism, which uses a belt drive and a pair of tensioned nylon lines to translate an anodized Aluminum drum across the feed piston to the print piston. The nylon line applies a torque to the drum, causing the counter-rotation necessary for even powder distribution.
More progress on hardware assembly yesterday. I modeled and printed the acme nut mounting plate (a really hacky part, but effective) and began testing the powder distributor while the rest of the piston assemblies printed. I designed the distributor drum to counter-rotate as it distributes powder (as in Z-Corp printers) and to stow away out of the motion plane of the laser toolhead. The drive belt only provides lateral motion to the rod via rotary bushings and counter-rotation is established via a once-wound length of tensioned nylon monofilament. The nylon enforces counter-rotation by frictionally binding to printed pulleys at either end of the drum and apply a torque as the drum translates laterally. I like this design because it allows the two motions of powder distribution to theoretically be achieved with one mechanical drive. At the end of each distribution movement, the drum is stowed at the end of its travel distance in a notch in the body, clearing the motion envelope for the laser head. I used two motors to drive the distributor mainly out of conveience— I have an abundance of stepper motors, but no short GT2 belts wit which to turn a drive rod. Two motors also has the benefit of extra torque, which has proven useful when un-stowing the distributor drum.
I began a basic evaluation the laser’s repeatability and power linearity this afternoon using a Coherent PM150-50C air-cooled thermopile sensor with a broadband coating and a FieldMax-II TOP power meter. For each measurement, I ran the laser at the target power for ten seconds to let the initial surge of power decay away and then took a five second power average. The laser power is specified via either the control software or the manual interface as a percentage of total power and the laser begins continuous emission at 13% total power. I took four measurements at each integer power level between 13% and 40% to get a read of the repeatability of the laser as well as the linearity (if any) of laser power.
The results are interesting: with one glaring exception, standard deviations are relatively low (below one Watt), meaning that most of these power levels are somewhat repeatable, though for lower power levels, the relative magnitude of this figure is significant. The most interesting trend is the repeatable surge in power that manifests at 30% total power. In addition to this power spike, it is at this power level that the laser tube stops making a quiet hiss while emitting. All power levels higher than 30% emit silently while those below generate a noticeable sound. When this region is excised from the data and trend lines are fitted, it becomes clear that there are two more-or-less linear (matlab curve-fitting coming soon) regions of different slopes one either side of the power spike. I wonder if the power spike and change in emission sound reflect a change in emission mechanism? Or maybe there is some sort of resonant constructive interference in the cavity? Both of our lasers demonstrate this behavior.
Next steps are establishing the working power ranges for test powders, modifying firmware to control the laser over these power ranges, and conducting sintering experiments.
I have some preliminary results concerning milling media selection for low cost ball-milling of crystalline isomalt grains. Using a low-cost rock tumbler and two sizes of steel balls for grinding media, I milled each experimental batch for 3.5hrs and then separated the result in a sieve stack on an oscillatory shaker run for 2hrs. By weighing the powder caught in each stage of the sieve stack (meshes 10, 35, 60, 120, and 230), I was able to measure the particle size distribution. I explored the following combinations:
100g isomalt grains milled for 3.5hrs with 60 1/2” stainless steel balls and 250 1/4” low-carbon steel balls (770g total)
As can be seen in the bar graph above, ball-milling with 1/2” steel balls is most effective at reducing particle size of crystalline isomalt, but still leaves much to be desired in terms of the distribution profile. Additionally, my methods are pretty fast and loose— I did not control for mass or number of balls, which really calls into question the comparability of the results for different balls sizes. However, the data illustrate a decrease in particle size with increased milling media mass and/or milling media size. This makes sense in the context of the milling mechanics— higher mass media yield higher energy impacts with the drum walls, increasing the probability of substrate particle fracturing. So there is some valuable targeting information here.
Again, this is a quick experiment more for targeting purposes than fine optimization. Other variables left to explore are the ratio between milling media mass and substrate masses, rotation speed, milling duration, as well as a more thorough and well-controlled study of milling media size.
I’d like to share my final presentation as a fellow at the Advanced Manufacturing Research Institute (AMRI). This video was taken on August 23rd after about two and a half weeks of work on the first laser sintering prototype, but it covers the broader goals of this research, namely the fabrication of vascular structures from sugar glasses for a rapid hydrogel casting process to create vascularized synthetic tissue.
Integration and piston testing underway. The pistons are loaded with bakers sugar in these pictures solely for testing the hardware— I will be ball milling isomalt glass this weekend so that I can start traversing its speed-power space to find optimal sintering parameters.
As of the 16th, I have robust control of both the X- and Y- axes of the laser cutter and am now working on control of the laser. I spliced in leads to the laser control lines and hooked up my pocket oscilloscope to get a read on the control protocol. As can be seen in the trace above (and confirmed from this laser power supply manufacturer’s website), it the laser cutter’s microcontroller uses a PWM signal to control the power level of the laser (between 0 and 80 Watts). I’m working on increasing the PWM frequency on an Arduino to over 20kHz (as spec’ed in documentation) to get basic laser control. After testing with Arduino, I will migrate control to the RAMBo board.
The state of the hardware the night of the 17th. Printing out the remaining pieces now. I assembled the body back wall and chamber side walls with acrylic adhesive and alternately used clamps and weights to keep things square while the adhesive cured. I used a lot of belt and nut traps (as can be seen above) in my parts to make assembly easier.