Our two-part series on 3D Printing Simulation Part 1 and Part 2 managed to leave out a key player: Altair. To make up for our omission, we caught up with Ravi Kunju, senior vice president, Strategy and Business Development, Simulation Driven Design at Altair, the best person to fill us in.
Ravi, give us a little background on Altair.
Altair has always been a global technology company, providing software (including cloud-based) solutions in various areas, especially in areas ofproduct design, product development, high performance computing and data analytics. Our vision has always been to transform business decision-making through our advanced simulation, our advanced numerical methods, optimization and data analytics.
Now let’s dive down into additive manufacturing, as well as manufacturing in general, because for many years we’ve invested in manufacturing process simulation tools on the high-end side—whether it’s for metal extrusion, casting, sheet metal forming and so on. To us, additive manufacturing is one more piece in the puzzle. Today, there is tremendous hype for it, but it’s still less than 1 percent of what is being manufactured.
Does Altair consider itself first in areas of generative design?
Altair has pioneered in the area of creating generative design, specifically with designs to fit the performance needs better than anybody else in the industry. We worked in the old days in the U.S. with Boeing and aerospace companies all over the world.
Everybody used our topology optimization tools and our generative design tools for creating designs specific to a purpose. But in the last few years, we’ve identified that the simulation-driven design approach is absolutely critical in driving designers. When designers need to make sure that they have an efficient design for a part that needs lightweighting, that performs with all of the loading conditions—and at the same time, can be manufacturable. So, we are hitting these three targets with one shot, taking all of the performance requirements, the manufacturing requirements, and creating an efficient design with light weight that can be manufactured.
What about up-front simulation for designers?
So, in the past, we had solutions to either validate a manufacturing process once a product was deemed to be right for mold, castor sheet metal, but then we realized that we needed to bring everything up front into the design cycle. And that was the genesis of the Altair Inspire platform. We started making sure that the entire optimization approach was available to designers—not just the high-end analysts who understand meshing and the rules of modeling. We have not only brought it up front in the design process but we have also made it easy for the designers to digest the information.
How manufacturing process specific does Altair get?
The Inspire platform generates the design but at the same time is aware of the desired manufacturing processes, so that the final shape will accommodate the process. For example, if you have an injection molding part, you can specify a minimum or maximum thickness of a wall. If you have a cast part, you may not want to have any cores put into the castings. If you have a sheet metal part, you’ll want to meet a certain frequency target. If you have a part that is made with additive manufacturing, one of the big problems is post-processing, removing supports. We’ve understood the best design with the least amount of support structure. You define these criteria.
That is what I mean by being cognizant of the manufacturing process. When we incorporate that into our generative design and the topology optimization tools, we are able to generate a design that is tuned to the manufacturing process. It’s built into the framework of Inspire.
This is where we started generating the design. At the same time, we’ve added a 3D printing module to address the laser sintering process. You generate a design for that particular process that can also be validated—all in one single environment—with a thermal-mechanical simulation.
We showed this at our Altair Technology Conference and also at formnext last year. We generated designs not only for traditional manufacturing processes and additive manufacturing, but with additive, we broke it down into one for traditional manufacturing with hybrid additive process, where we were doing the sand 3D printing and the sand molds, and then pouring the casting into it. Then we did it for selective laser melting (SLM) without any support structure. We created one design for the fused deposition method, where there is the thickness requirement as well as distortion during the sintering process. We also did one for the binder jetting. All of them have different manufacturing requirements, but we generated a design for each of them.
Right up front when you’re influencing the part the most, doing the design and thinking of the process to make it, we’re providing all the tools within a single Inspire platform.
With distortion of the 3D-printed part, how can you account for that?
We call a “compensated shape” one that after distortion is the desired shape. A compensated shape is needed with any type of additive manufacturing, and is even more critical forthe metal sintering process. The difference between a selective laser melting (SLM) process and the selective laser sintering (SLS) process is with the way to bring the metal powder together.
Either you use a binding agent and you extrude the binding agent along with the metal powder through the injector nozzle and then build it up. That’s one process known as the fused deposition modeling (FDM) process. The powder and the binding agent are together and the part is built layer by layer. Or, a second approach is where you spread a thin powder layer like you do in a selective laser melting process. You have a very fine layer, a few microns thick, and a spray jet that sprays the binder onto where you need to bind the material. You are creating a “green part” with the binder and the metal powder stuck together. If you take a green part after it’s built and if you drop it on the ground, it may break up into powder.
You take the green part, as it’s called, whether it’s made with fused deposition or with binder jetting. Binder jetting is like printing. Imagine you’re printing, putting in ink paper, except instead of ink, you use a binding agent where you need to bind. The rest of the powder is not stuck together. So, they build it in paper. You have a single two dimensions—you’re building one layer at a time. And then at the end, you have a three-dimensionalgreen part.
The green part has to go into a sintering oven, and what happens next is the biggest difference.The oven is at an elevated temperature and is finely controlled to avoid hitting the melting point of the metal, but it’s hot enough to melt or burn away the binder material. The metal powder starts fusing, there’s diffusion—not melting—between the layers.
What about the melt pool?
With a selective laser melting process, there is a melt pool.
Imagine the same powder, imagine that you have a thin layer of powder on a tray. There are two ways to bind it. One is you can take a laser gun and you can melt the powder, and so now you create a melt pool. Imagine a small laser creating a melt pool wherever it wants to bind it. Then there’s a recoater that goes in, and then the next layer. In the second approach, instead of melting the pool, you are throwing in a binding agent—you’re spraying a binding agent exactly where you want it to coalesce, and then you take it into the sintering oven.
The melt pool is something that comes in a laser melting process, not in the sintering process.
In the sintering process, there is a shrinkage of usually about 15 to 20 percent. So, you have a green part, and because of the diffusion that happens between the layers, there is distortion. And this causes global distortion. Let’s take simple geometry, like a cube. If you sinter it, it’s going to shrink into a smaller cube. But imagine you’re printing a more complicated shape. The diffusion happens unevenly. You get only global shrinkage but also warpage.
Uniform shrinkage—is that an easy problem to solve?
If it was uniform shrinkage, it wouldn't matter. It would be an easy problem to solve.
Is nonuniform shrinkage a problem Altair has solved?
The process that we actually are able to solvetoday is only selective laser melting, where we do a thermo-mechanical simulation and identify what the distortion is.
Sintering is tricky. We don’t do the sintering simulation today. That’s a very challenging simulation. There is one 3D printer manufacturer that claims it can do it using lots of data and machine learning and other techniques. But even if today, if you give them a complex part, it will take them six to eight weeks to figure out what [would be] the best shape. Maybe they have shortened the cycle to less than four weeks. It’s an iterative process. There is no simulation tooltoday on the market that does accurately predict the sintering distortion and the warpage that occurs in the sintering. It’s very complex.
What is happening in the six to eight weeks to figure out the best shape?
Iterations. You have to physically print the part. Compensate. Iterate.
It’s a tough problem. Let me have you reframe two things. I want to break this down into two separate problems. One is to get to a shape that’s easier to manufacture for a process. What is the best shape that you can generate that produces the least amount of distortion? This is where we tweak the manufacturing constraints. We work very closely with these manufacturers to understand the drivers for this distortion. We have some tricks. I don’t want to spell it out exactly, but it involves where the mass distribution should be. Then you have less distortion for the laser sintering process. That doesn’t mean that we are able to completely eliminate the distortion, but we will give you a design that is better, where the distortion is evenly distributed and there is no really big distortion in one place.
Would you say that an engineer would have fewer iterations to create the final part because Altair can solve some of the distortion?
No. So, again, let me put it this way. There are two bins that I talked about. One is coming up with the design, and the second is validating your manufacturing process. Let’s separate those two for a second.
Your starting point should be the best for that process. We are pretty good at coming up with a starting point for a sintered part because we kind of have some tricks to get the manufacturing constraints, which we can tweak, to get a good starting point. On the process validation side, if you have a design. We don’t have the tools today to do a process validation for sintering because it’s complex. We are looking at it and looking at it very seriously. But we can only validate the selective laser melting process right now. So, from a validation point of view, we are only validating the laser-melting process. In the future, we’re looking to see what we can do with the sintering process as well.
Our software will produce the least amount of distortion and the fewest number of support structures in the case of SLM, which we can do better than anybody else. Best to keep those two—the process simulation and the generation process—as two different things.
In summary, we can find the compensated shape for SLM parts. For the sintering process, it’s extremely complex.
We are still researching the problem as well as studying the published literature to create a reduced order model that we can use for simulation. But at this point, we’re exploring and so are others. There may turn out to be multiple solutions.
Does Altair’s simulation approach make for a quicker time to market than for the manufacture?
Yes, absolutely. When we change the starting point and the iterative process goes from ten weeksto maybe one, two weeks or three weeks. But,again, that is for binder jetting only.
Remember, the metal jetting process is brand new. With metal powder in the sintering process, you can’t get 100 percent density because there will always be some gaps.
The density can be high—97 percent to 98.5 percent with additional processes even—but it will never be 100 percent. There will still be some voids inside. We also have methods where voids are considered from a performance point of view, and we can evaluate it based on density degradation.
The density is uniformly degraded throughout the part model?
Yes. We create a reduced order model. We have a tool called Multiscale Designer. It’s also used in composites simulation, where you have to deal with the orientation of fibers. A reduced order model manages density and a lot of other things as well.
Directional strength properties too?
Yes. We calibrate for every process. We create something on a unit cell and propagate that unit cell across the entire model. We do this today for composites, and we are extending it to the sintered process with a certain degraded density for voids could be easily done.
Would that be used for the reduced strength along the Z-axis due to layering in additive manufacturing?
Yes. It all depends on how you’re calibrating your unit cell. The unit cell could have the information of the build direction.
Is Multiscale a separate product?
Yes. With Multiscale Designer, you are representing a model on a unit cell level. You can use this approach to any kind of a problem where a unit cell can be replicated. For example, a lattice structure.
Today, if you want to model a lattice structure purely with finite elements, it will create a very fine model. One of the ways to handle that is to fill it up with tet, quad or hex elements. Or you could say that the individual elements act like another shape—like a diamond shape—after calibration with a particular process, whether it is for distortion or structural analysis or some other mechanical simulation. Then you can propagate that across the entire model.
Can the lattice be selected from a menu?
That’s more on the design side. Once the lattice structure is defined on the design side, we can tell you where you need to have a material density to be 100 percent.
Does Altairsoftware have the ability to design with lattices?
Yes, we can tell you what kind of lattice to put where from a design perspective. That’s been there for ages. But from an additive manufacturing perceptive, can we model this?
Then you can take the lattice structure as designed and simulate it with a uniform volume that has a property derived from the Multiscale?
Yes, you can with MD, our Multiscale Designer.
That makes the problem solvable; otherwise, you’d have so many elements you couldn’t possibly solve it.
Correct. Personally, I’m not a huge fan of lattices, especially for structures that are loaded. The variation in the manufacturing process is so large that a lattice runs into a lot of issues and inconsistency in the printed model. It’s just the way that the machines are today. You can try to get consistent lattices, but if you fail, how do you know? You can’t do an optical inspection easily.
So, if lattices look nice and pretty on thescreen, 3D printing machines probably can’t make them. And you can’t inspect them, so you won’t know?
They do; they can. A 3D print can be fine-tuned. But if you’re going for large-scale production, consistency becomes a problem. For a space expedition—a one-off part—you need to reduce the weight. You could get it right. But can you make multiple parts like it? I’m not saying that we cannot do it, but as we scale this to larger volumes, there could be inconsistencies. Would I be comfortable in a plane that has its landing gear made out of lattice? Today with additive manufacturing, I would not be okay with that.
I wouldn’t be on that plane. Or if the lattice part was in the wings. You said that you are conscious of manufacturing processes. How specific can it get? For example, can the software be machine specific?
We can look to adjust the process, but not the specific machine. Whether it’s SLM, binder jetting, casting, etc. So, from that perspective, it is not conscious of what exactly the machine itself is. But it’s conscious of what the process is.
For the process simulation, we can go into detail. Thermal mechanical simulation, for example. Other parameters can be included: the bed size of the printer being used, the wattage of the laser. You could put that as an input. But you can’t give the make and model number and have all the simulation parameters defined for simulation in one shot.
What about supports? You say you can help in making a part design that needs fewer or no supports?
Yes. We have something called an “overhang constraint.” What happens, especially with selective laser melting, is that you’re printing and you’re going up. If the angle of the part is below 45 degrees, you need to have a support structure. If there is nothing underneath it, the bottom side will be distorted.
Every time you have a horizontal section, you need to have a support. If that angle is over 45 degrees, you can live with that.
Let me give you a quick idea of what we are talking about to help you understand. We have a complete ideation tool. It’s called Inspire Studio and it’s part of the Inspire platform. You can draw a sketch, you can import a sketch, you can scan it, and you can start drawing lines. This is a revolutionary product in terms of the initial designers for them. It has a construction history—you can create polylines, you can create solids very, very quickly.
Everything that you’ve done here, it is only from a design perspective. You’re designing it in such a way that even the angles that you saw for the lattice inside were all within 45 degrees, so there’s no need for any support. We do simulation during this process. Take, for example, you are printing a pen. You can extract it and it’s ready to go. We just created this pen, and the beauty of it is that we have absolutely no support structure.
Another example is what BMW has done, which helped them win the Altair Enlighten Award. We talk about performance, manufacturability and efficiency coming all together. Take, for example, a bracket for an assembly. There could be lots of loads coming in. You want to extract those loads. So, what we do is we take all of those loads and that becomes the basis for your performance. Not everybody can take all the loads like we do.
Do you mean you can get all the loads at once and apply them simultaneously when others may have to apply them sequentially?
Yes. Others do loads one at a time.
What do you mean by manufacturability?
We may have the same design requirement, but if the part is to be machined, it has one shape. If it is to be casted, it will be a different shape. If it will be 3D printed, it will be a different shape. We have extended this beyond 3D printing, as mentioned earlier, with distinctions between 3D printing processes, like whether it’s SLM, whether it’s a binder jetting process or whether it is a fuse deposition method. All of them, if you look at them, the performance is the same, but the shapes are all different.
That’s why I keep these builds separately, even while what is driving the design stays the same.
Let’s say you decided to make a casting component. Irrespective of what the manufacturing process is, we generate a specific design for the purpose of the component. The design process is the same. In the end, you have to choose the manufacturing process. We can do a casting simulation to identify where the porosity is going to be. That will tell you where you need to put the gates. That’s what we mean by understanding all of the process parameters to capture.