Additive manufacturing gives us limitless geometric freedom—theoretically. So you’d expect to find exciting computer aided design (CAD) software that capitalized on that. But when the hype around the mantra “if you have a CAD file you can print it” runs into reality, users find the utility in some form of computer aided manufacturing (CAM).
On the other hand, observed Lauren Heinrich, a master’s degree researcher in hybrid machining at Georgia Tech and intern at Mazak in Florence, Ky., some CAM software simply reverses a subtractive tool path—going from the bottom up rather than the top down. That won’t work in many situations. So we’re here to help you understand the issues and explore what’s available from some of the best players in the field.
In generative design, the human sets functional requirements and constraints, and the software automatically generates multiple designs that fulfill those conditions. It’s led to a number of creative solutions, and increased our admiration of mother nature, since so many optimum designs tend to look biologic. But not every design can actually be manufactured, and what can be manufactured by one additive technique may not be possible with another. The better CAD/CAM packages account for this.
For example, Mathieu Pérennou, global business development director for additive manufacturing at Hexagon’s Manufacturing Intelligence division (U.S. headquarters in North Kingstown, R.I.) said their MSC Apex Generative Design software does all its topology optimization while taking manufacturing constraints into account. For example, the topology optimization process can use a “cinematic multi-body dynamics simulation to calculate what loads the part is going to see. From that we can derive a number of load cases, and the input will be those load cases, plus the design space.” The software then proposes geometries that meet the load criteria.
“We know the part will not fail, but that doesn’t really tell us how good this part is from a manufacturability standpoint, so we also link the topology optimization to our process simulation software. We get feedback on the manufacturability of a product and can then modify or improve our generative design based on that feedback.” The software would offer options for orienting the part in the build area and adding support structures, if needed. Pérennou said the support structures could come from outside or be built along with the part if it’s a powder bed fusion (PBF) machine (using Hexagon’s ESPRIT software).
Besides arriving at the “topology optimized geometry,” Hexagon’s software can also produce what Pérennou referred to as a “second geometry.” That is, “the product we want to make, plus the stock we add where we know we need to machine. Because what we produce in additive manufacturing is usually not the finished product, but the product plus additional stock where we need to machine. Our software can automatically add that material.”
The simulation process also shows the part being built virtually, layer by layer, and shows the user production issues like hot and cold spots, areas with a risk of cracks or distortion, and re-coater interference. This manufacturability assessment generally eliminates a number of design candidates, at which point the software also offers cost estimates of each design to further refine the decision process.
Generative design goals need not be limited to optimizing for strength, weight and material usage, explained Ashley Eckhoff, marketing manager for the manufacturing engineering group concentrating on AM at Siemens Digital Industries Software, Plano, Texas. Siemens also offers the ability to optimize for fluid flow. “It could be air, or a liquid coolant,” Eckhoff said. “The software optimizes to achieve a maximum or target volume through a system. We’ve found applications like ducts to improve airflow in cars or aircraft cabins. And we’ve explored things like electric vehicle battery cooling.”
Eckhoff pointed out that heat exchangers present a somewhat different problem, because you would not want laminar flow if you were trying to maximize heat dissipation. “You want as much coolant as possible to touch the hot zone, in order to draw more heat away. So you want that coolant to be turbulent, swirling around inside those channels to get maximum coverage between the coolant and the area you want to cool.” Ulf Lindhe, general manager for industrial additive manufacturing at Oqton, Ghent, Belgium, agreed that this is a growing application for AM. “Manufacturers have started to print copper alloys, for example. These are used for electrical components, but also for heat sinks and for heat management in electronics and electric drives in the automotive industry and many other areas.”
The more you examine the challenge of applying AM to industrial applications, the more you recognize the need for accurate simulation in your CAM software. The need to prevent re-coater interference in PBF is one reason, explained Eckhoff.
Without accurate simulation, the sintering process can cause the metal to distort, “due to heat buildup between the layers, or the laser dwelling too long in one spot. If the material bends up, it can ruin your re-coater. Not only does it destroy your print, which is now a $50,000 paperweight, but you have to fix your re-coater before you can start printing again.” Eckhoff said their patented PBF simulation technology identifies these areas of concern in advance, and “any simulation that helps a user save that $50,000, even if it takes an hour or two, is well worth it to them.”
Gene Granata, director of product management at CGTech, Irvine, Calif., argued that “the vast majority” of CAD/CAM packages are not simulating the posted output that actually runs on the machine. Thus he sees a role for their VERICUT product, which does exactly that.
“VERICUT’s claim to fame is running the same NC data that you’re going to be running on the CNC machines and showing you in VERICUT’s digital twin world what’s going to happen. To do that, you need to incorporate data right from the machine control. You need the same sub-programs. You need the same way to control the build parameters through variables. You need to read those directly out of a post processed NC file, along with all of the other motion controls and the typical things that have plagued NC programmers from day one.”
“There are CAD/CAM simulations that give you a playback of what you did at the programming level. But once you post-process that and then start introducing build parameters and things that are specific to the material, and/or you bounce back and forth between additive and hybrid subtractive processes, now that real CNC file is what’s going to tell the tale of what’s going to happen. Also, a number of our customers edit the files that come out of the CAD/CAM process, because they’re either inadequate, incomplete, or require some level of modification. So now what you have going to the machine differs from what you had at the CAD/CAM simulation level as well.”
If you do find an error with VERICUT, you can easily link back to the line number in the NC path that created the problem. “The software shows the tool, the feed rate, and other process parameters related to that spot,” added Granata. “It’s a very fast debugging tool for programmers. They know exactly where to go fix the problem at the CAM level.” Another nice feature is a free application that provides a review file to the team on the floor. Without requiring another VERICUT license, an operator can view the full simulation on a tablet or PC near the machine. “They can preview how a part is going to be deposited, how it’s going to be machined, the balance between those sequences, what they can expect to happen,” said Granata.
The directed energy deposition (DED) process has its own challenges. For one thing, the method is often used in the repair business, in which you’re not building on a flat plate, but on a variety of complex surfaces in various orientations. Joe Wilker, Mazak’s advanced multitasking manager, said there are also hard coating applications. Another use case is reducing the number of unique castings one might otherwise stock. “If only a little feature differs from type to type, a customer could instead stock one common casting and just add the feature at any given time.”
As David Bourdages, product manager at OPEN MIND Technologies put it, the key here is giving the user “the ability to deliver the powder and the additive process in five axes, because then you can keep the orientation as normal as possible to your substrate. That is the best thing to do in these cases.” OPEN MIND, with North American headquarters in Boston, specializes in five-axis machining with their hyperMILL product. And Bourdages added that the AM module includes an algorithm that helps maintain a constant deposition velocity, “which helps maintain a constant build thickness. This is really important because we’re talking about a complex system with several axes in simultaneous movement. If you force a quick movement, a quick change in direction, then you will have acceleration, and the acceleration will change your rate of deposition. This will create some overbuild or overheating in the part.”
Brad Rooks, OPEN MIND Technologies USA application engineer, echoed this and contrasted it with subtractive machining. “In the subtractive world, if the real velocity of the head slows down or increases a little bit, you may see a witness mark or marks here and there from the rotaries being in play, but it’s not a deal breaker. Whereas with additive, those little deviations from constant velocity cause overbuild and under-build, and wreaks havoc. You start to lose your focal length.” So hyperMILL’s advanced tool path generation options include technology parameters that govern “all the inputs that go into the build, such as powder flow rate—or wire rate if it’s a wire system—the laser power, shielding gas rate, feed rate and focal distance.”
Rooks also illustrated how an additive path often can’t be the reverse of a subtractive one. Interpolating a cutting tool around a circle in the X-Y plane presents no great challenge. “But if you were to take that same tool path and apply it to a blown powder system that only has delivery from the right side, the build is not going to be consistent in that circle, because it’s going to behave differently as the laser and the powder traverse on the right side of that circle as opposed to the left side.” Conversely, if you are able to interpolate the moves in 5 axes, you can keep the powder and the laser in the same relation to each other as they traverse along the same path.
Another example of how OPEN MIND is using five-axis interpolation to improve an additive process is in building parts that are rotationally symmetrical about a post—a rocket nozzle being the test case. It’s also important to note that not only are such parts axisymmetric, they often have varying wall thicknesses. Managing Director Alan Levine explained that the usual approach of “slicing” the geometry to build multiple layers of concentric circles creates many voids and discontinuities in the build. That’s because the laser head is frequently stopping and starting as it moves from level to level.
HyperMILL instead creates a continuous five-axis path that articulates the rotary axis all the way through the build. “We handle variable thickness, moving the head in and out so we have optimal deliveries. If there’s a ledge, we can lean in 30o, so we’re normal to that surface as we’re building. We spiral up, spiral in and out, and angle control, all in one combined process where the entire build has just an initial start and one stop at the end. The research data on the metallurgy is showing extremely high density, with no pores.” The quality of the build is due in part to the fact that the bead of a multi-pass layer never aligns with the path from a previous layer, explained Bourdages. He also pointed out that despite the complexity of the moves, the tool path planning happens in a straightforward 2D view.
Eckhoff noted that most additive processes require some sort of pattern to fill in the geometry internal to the part. But “these infill patterns sometimes cause the laser or electron beam to dwell too long in a specific area, which causes clumping of the material inside, and can lead to voids and stress areas.” This is especially true for thin walls and when the machine builder does not provide an optimal utility for this function. So Siemens is refining an algorithm that allows the user to find these areas and adjust the deposition path to alleviate the problem before putting the program on the machine.
Achieving the kind of accurate simulation and process control covered above requires a close partnership between the machine tool builders and the CAD/CAM providers. But both sides acknowledge there is more work needed to make AM more predictable and controllable in real-time. As Heinrich summed it up, “Even pipe welding robots are programmed to join a certain diameter and thickness of pipe. They can’t go across a whole gamut. You have to change the settings. So imagine that from a machining perspective. There are a lot of variables. And if the weather changes, it could be another set of variables.”
For example, said Heinrich, layer height consistency can be problematic, and any error compounds with multiple layers. “For a deep build, those layer heights need to be more and more exact. So a current R&D priority is in-situ layer height monitoring, and the ability to adjust that on the fly. Because no matter if you’re running a wire or a powder system on any machine tool platform, that’s critical.”
Heinrich said many researchers are trying various non-contact measurement systems, such as using coaxial camera to peer through the laser head, and determining how changes in the laser spot diameter relate to changes in layer height. “This would make it possible to estimate your offset and how much you need to correct.” Another idea is asking how the temperature of the part affects layer height and whether such data can be used to adapt the process. Clement Girard, product manager for additive manufacturing and artificial intelligence at Hexagon, said they are “putting laser scanners inside the CNC environment in order to measure the stock just after printing to then re-adapt the tool paths for the subtractive machining side.”
It’s not all future tech. Wilker said Mazak has “a dynamic process control system in place for hot wire, where we are feeding information back from the head to the controller on laser power, wire feed, heat and shield gas. We also are monitoring the temperature of the build process and getting some camera feedback on the build process. But there comes a time, and it depends on the type of material we are dealing with, that we’ll build up to a certain height and come in with a cutter to level it off and create a new starting point.” Wilker added that the machining step also allows the material to cool before resuming the build process.
Clearly AM is a software-intensive endeavor. From Lindhe’s perspective, it needs to be even more so, and CAD/CAM needs to connect with a company’s enterprise resource planning (ERP) package and manufacturing execution system (MES). Oqton’s vision is to “help manufacturers increase innovation and efficiency by intelligently automating AM” with a Manufacturing OS. To do that, Lindhe said, they need to be cloud based and “totally technology agnostic” (not tied to any type or brand of machine). Plus they need to take advantage of artificial intelligence.
To Lindhe, the future of 3D printing is “about being dynamic and connected live to other systems in the company to track and manage customer requests…To be able to query quickly for the best utilization of all equipment, for just-in-time predictions and delivery schedules.” This would enable “personalization, or at least highly flexible manufacturing plants.” Lindhe envisions additive manufacturing bringing us “completely new products and completely new business models, where the number of discrete customer orders and products can go from a few hundred to hundreds of thousands.” It has already happened in the dental industry, he pointed out.
Lindhe also argued that traditional build-preparation software cannot solve all these problems. “Such a solution, even if it’s a super smart one, just becomes another point solution in the manufacturing chain…If you talk about real manufacturing, you need a much higher level of connectivity between your software tools, and also with the machinery on the shop floor.” It also needs AI, he said, so they built it into the platform “from scratch” and use it wherever it makes sense. “Oqton can learn from machine usage habits, provide metrics and optimize how you operate the equipment more efficiently.” It also analyzes big data to find patterns that help predict the future for preventive maintenance.
Oqton also uses AI to “recognize features, shapes, forms, and parts, to suggest the best build orientations for additive. We typically suggest a number of orientations, and if the user doesn’t agree with any of those and keeps adjusting, the system learns from that.” AI can also categorize parts. For example, in dental production, “the system can recognize parts like crowns, bridges, abutments, prosthetic devices and models. If the system sees a crown, it knows that in this factory setup it should route it to a specific printer, and it will be printed in titanium with a layer thickness of 15 μm. That part is automatically routed to that type of machine, nested together with similar parts, and automatically scheduled.” You can review the schedule, Lindhe said. But AI has done the work.
Connect With Us
