A team of scientists at the U.S. Department of Energy’s SLAC National Accelerator Laboratory are working to identify these sources of error and help manufacturers avoid them in the future. Their research, which is being conducted at the Stanford Synchrotron Radiation Lightsource, involves 3D printing sample parts inside an x-ray characterization chamber.
Engineering.com had the opportunity to discuss the project with three of its principal investigators: Johanna Nelson Weker, Kevin Stone and Chris Tassone.
Can you explain why you’re using x-ray characterization rather than, say, thermal imaging to identify the sources of flaws in 3D-printed metal parts?
JNW: The real benefit with x-rays is that we can see subsurface defects forming and structural changes happening within the substrate, not just on the surface or on the powder.
Are you looking at other variables—such as melt pool temperature?
JNW: We’re not currently using thermal imaging to track the melt pool specifically, but we do have the capability to add that to the chamber we’re using in the future. We are using x-ray diffraction to track not the melt pool temperature, but the unmelted material directly around it. Once it either re-solidifies or in the heat-affected zone that’s never melted, we can track the cooling rate of that.
CT: This work is being done in collaboration with Lawrence Livermore [National Laboratory], and they’re the leading experts in characterization using thermal imaging and optical imaging at very high framerates to follow the dynamics on the surface, whereas this program has been focused on what extra information we’re getting when we look through the material as it’s being printed. That’s really the power of x-rays.
What alloys are you working with?
CT: There are some we aren’t supposed to talk about, but we started with Ti64 and we’ve looked at some aluminum alloys. We’ve also done some work with Ti5553.
KS: We’re starting to look at how other alloys change the behavior of the melt process, so that’s been Ti5553, some aluminum alloys, and at least one magnesium alloy.
JNW: We’re trying to go to the lighter alloys and steer away from stainless steel. Mostly because we don’t think our x-ray energy can penetrate stainless steel very well, so we can’t get into realistic printing conditions.
CT: One of the high-value applications areas is structural components for aerospace, so when we look at how we can use additive to impact the deployment of novel alloys that are lightweight but still have good mechanical properties, moving toward these alloys that have been underexplored in the additive community is a place where we think we can accelerate the reliability validation. Right now, the hard part is that you have to perform those validations not only on the part, but also on the additive process itself because it’s viewed as being so stochastic.
What’s your ultimate goal with this project?
CT: The goal is to combine those two approaches [SLAC's and Livermore's], so that we’re getting all the dynamic information about what’s happening on the surface and the rate of heat exchange with the information that we’re getting from the x-rays, in terms of the aspect ratio of the melt pool, the conditions that lead to void formation, how deep those voids are, and how the crystal structure evolves as a function of the processing conditions. It’s really through a combination of all those things that we think we can well characterize the process.
JNW: X-ray characterization is adding another tool to the ones that already exist—by providing complementary information to what you would get from thermal imaging or optical imaging. Our long-term goal is to have the x-rays inform what’s happening below the surface. If we can connect those to heat signatures or signatures in visible imaging, then we don’t need everyone to come to a synchrotron to test every new part because they can use the thermal imaging on their systems and know what’s happening by connecting it to the x-rays.
KS: One of the goals of the project is to put ourselves out of business, where we inform the in-line metrology that will be on the AM systems themselves so users can detect subsurface defects or crystal structures using characterization that you can actually implement on these systems.
JNW: We also want to give more information about the physics of what’s going on to the modelers, so that we can have more accurate factor models. Right now, you can either have a really accurate model that takes months, or a very basic, bare-bones model that may not capture everything that you need to know to print your part.
CT: The National Labs sit at this intersection between academic research and industrial R&D, so our goal is to inform both directions. It would certainly be an amazing thing if you had a model that was capable of simulating any process and knowing exactly what the mechanical properties of your part will be.
One of the things we’ve been looking at most recently is how to prevent the formation of voids at known problem areas. For instance, when you have a turn in your raster pattern of the laser, you can proactively tune the power density to avoid the formation of defects at places like that. We’re happy to freely give that information to all manufacturers, since we’re not looking to develop a commercial product, we just want to accelerate the adoption of additive manufacturing.
You’re currently focusing on powder bed fusion, and I understand that you’re planning on looking at directed energy deposition as well, but does what you’re learning have any relevance to wire-based metal additive processes?
CT: Traditionally, it’s gone the other way. When we look at what other people are doing with in situ characterization, they’ve been looking at the wire feed and then trying to extrapolate that information to the powder bed. One of the things that we’re trying to understand right now is the interplay between the feedstock, the energy source and the substrate material.
But one of the fundamental things we’re getting out of this research is how much of the light is coupling into the power, in terms of thermal energy, and whether we can model that exactly so that we know what the melt pool will look like for a given power density and a given shape of the wave packet of the laser. That should be cross-cutting for whatever the feedstock is, so long as we know the geometry.
What about non-metal 3D printing? Do the insights you’re gaining here have any import for additive manufacturing using polymers, ceramics, composites, etc.?
JNW: No, but one could conceivably do ceramics in the system we have, which is to say that we could modify the system to do ceramics.
KS: Yeah, ceramics is probably the closest analogue. For polymers, the major difference between 3D printing plastics and metals is that, in the case of metals, you’re dumping a huge amount of energy into the system and then allowing it to relax very quickly, so it’s a kind of “far-from-equilibrium” process. Whereas polymer extrusion is generally much more well understood, so there are fewer uncertainties about what your part will look like, given the process conditions. They’re a bit further along on the polymer side because they’re used to modeling from typical plastic extrusion processes.
You’ve been using x-rays to examine the microstructure of each part layer, as well as the metallurgical impact of heating and cooling the powder. Between those two potential sources of error, have you found one to be more common than the other?
JWN: I think we haven’t gotten to a point where we have enough data yet. At this point, we’re trying to simplify the problem to be able to control all of the variables. The control of the powder is one of biggest issues—and the biggest variable. We’re looking at the quality of the powder, and how that changes things, or if it even does.
CT: One of the things we haven’t done yet, that’s on the roadmap, is investigating the link between microstructure and mechanical properties. If we print a part with a given process, we can tell you what the microstructure of that thing will look like, but we can’t tell you what the mechanical performance of a thing with that microstructure is. So, we can give guidance in terms of which sets of process conditions will lead to highly defective parts versus process conditions that will lead to parts that are fully dense, with the right set of microstructural phases to give the mechanical performance one would expect. But we haven’t established that link with data yet.
KS: Void formation is almost universally bad. That will always be viewed as a problematic defect. We’re starting to understand the processing conditions that will lead to increased void formation so we can mitigate it. In terms of microstructure—things like strain, phase partitioning, texture within the grains—we still don’t have quite enough data to say too much about that. Chris is absolutely right: we can definitely characterize that microstructure in great detail after the fact, or even as it’s forming, but linking that to whether or not that makes a high-quality part with the desired mechanical properties is a gap that we haven’t started to really address yet.
CT: One of the things we can say is that the way the light is delivered onto the substrate is incredibly important. So, how that beam of light is shaped and what it looks like really defines a lot of the structural evolution. We’ve also found that how well the powder is spread out on the surface is very important to the quality of the lines that you’re writing. Both of those factors are important in terms of the outcome for the part. Another thing we’re looking at very closely right now is the minimum tolerance for powder quality that will allow you to print good parts.
What advice would you offer to someone who’s struggling with the quality of their metal additive parts?
CT: I’ll start with the plug: we are a user facility, and the goal of building this tool is so that people who are struggling with process development can come and use this tool to get real-time feedback on what their process conditions are resulting in, in terms of the structure of the part that they’re building.
JNW: I’ll add that even if they don’t want to bother with the in situ system, even analyzing fully printed parts after the fact with x-rays can be extremely beneficial to see why they’re failing.
KS: I think it’s very challenging to give guidance across different alloys, but there does seem to be a fairly narrow regime in terms of what process conditions will lead to full melting of all the powder but no defects forming.
JNW: At least in Ti64. Other alloys are very different—even the shape of the melt pool for aluminum alloys, for example, is very different from titanium.
KS: Zeroing in on that narrow set of process conditions can be challenging without these x-ray techniques. The papers we’re writing right now help to define that for Ti64, where you have full melting but no keyholing. The only other piece of advice that I can offer is that spreading an even layer of powder is incredibly important to printing parts with less strain and low surface roughness.
CT: Building on the idea that there’s a small window of optimal processing conditions, the thing that goes unsaid with that is that if you want to be within that window, you need very tight controls on all your processing conditions. You need to know—as well as you possibly can—how uniformly you’ve spread your precursor powder, how consistent that feedstock powder is, and how well-controlled your laser power and raster pattern are.
JNW: If you buy a 3D printer, there are often knobs you can’t turn, like how the printer deals with a turnaround, or what it prints first. These are things that we’ve found are really important, but they aren’t actually controlled by the user, so you pretty much have to hack the system.
Is there anything else you think our audience would be interested to learn about your research?
CT: Well, one of the things that’s been entertaining about this project is that when the process is going wrong, it makes for incredible, nano-scale action movies, with metal evaporating and flying all over the place. I bring that up because it’s very easy to go to power densities that are way too high and will consistently result in terrible parts. This just comes back to the fact that as we move to lighter alloys, the amount of energy we have to put into the system is a lot lower than we were thinking initially.
JNW: I would also add that we’re constantly seeking input from the industry, because we don’t want to be blind to what the industry needs. There’s clearly a lot of fundamental questions that we think we can answer that everyone in the industry could benefit from. So, we would encourage people to reach out to us and tell us what we’re doing wrong and what we should be using.
KS: We want to get the message out that these tools exist and we are a user facility, so access is open to essentially anyone who wants to use them. We encourage that, and we welcome guidance from all interested parties regarding the most interesting and useful materials we should be studying and what big questions are. We’re happy to be guided by what the field wants to know and to work with them to solve those questions.
CT: We’d be really interested to talk with the engineers who are developing processes on commercial machines to understand what their pain points are. A lot of the work that we’re doing is to validate—at the higher level—yes, industry, additive can get you where you need to go, and—at the lower level—to the modelers, yes, we can give you the information you need to make your model better.
Are you considering integrating metal additive manufacturing into your production line?
Here are 10 Questions You Should Ask.