An aerospace engineer shares his years of experience simulating composite materials.
Arthur Dubois, director of Aerostructures at Kittyhawk, is no stranger to simulating composite materials for use in aerospace structures. The company he works for has spent over a decade developing electric vertical take-off and landing (e-VTOL) aircraft that can be remotely piloted. The aim is to make air taxis affordable, ubiquitous and green. And in case you’ve been living under a rock for 20 years, composites are the ideal materials to produce lightweight aircraft—which is essential to ensure the feasibility of electric aircraft.
Dubois put it this way: “At a high level, what you generally get out of composites is better properties per mass. So, whether you’re talking stiffness or strength, you typically will be a little bit better, per unit mass, than metallics. The other key piece is that composites allow you to tailor your structure to specific loads. You can play with the orientation of plies and with the thickness to locally add padding and reinforce certain areas of your structure.”
He also notes that over the last 20 years, composite manufacturers have been enhancing the technology to improve material properties (especially at higher temperature and humidity) while simultaneously reducing their costs. Both aspects have made composites important to transportation industries. In other words, not only are composites offering better properties per mass, but they are also offering them at a reasonable price.
Metallics, however, are much easier to design and simulate due to their tendencies towards consistent material properties. “Metals can typically be approximated as being isotropic. What that means is that they’ll have the same properties regardless of the direction you load them,” Dubois clarified. “It’s not 100 percent true, because there are some variations in metals, but they’re slight compared to composites.”
“Composites, on the other hand, are what we call orthotropic materials,” he continued. “That means that if you grab a composite panel and load it in the direction of the fiber, and then you do the same test at 90 degrees to that fiber, you may get vastly different results. You can sometimes get orders of magnitude variations in properties between one direction and the other depending on the material and the weave. Variability is the baseline.”
It is this, and other unpredictable aspects of composites that makes them so hard to simulate. Luckily, Dubois has some tricks up his sleeve to share with the greater engineering community.
How Composite Material Variation Complicates Designs and Simulations
It is the very nature of a composite to have variability as, at their core, they are a non-homogenous mixture of different materials with different properties and a variety of failure modes. “It’s fairly easy to understand if you’ve ever seen a unidirectional ply of carbon,” Dubois said. “If you load in the fiber direction when loading the carbon, that’s really strong. If you load in the transverse direction of the fiber, then what’s reacting to the load is the resin which is effectively plastic and much weaker than a strand of carbon.”
A way around this weakness is to weave the internal fibers to reinforce the structure along different directions within a given plane. However, the plies of composite are still stacked on top of one another, so the weaving does not extend to the vertical, or out of plane, direction. As a result, if you load the laminate in the vertical direction—for instance, with a tension force—then once again you are putting all the load onto the plastic resin.
“With the traditional prepreg materials that we use, the so-called transverse tension shows with thicker, curved laminates. If you load them, you can show that there is interlaminate tension, which tends to effectively pull different layers apart. And because the material is weak to begin with in that direction, sometimes this can lead to really complex failure modes that aren’t thought of from the get-go,” Dubois said.
There are ways around this limitation as well, but they aren’t widely used at this time. For instance, Dubois notes 3D stitching, which connects the laminated fiber sheets in the Z-direction, to hold the plies together. He’s also seen a method for thicker laminates with high curvatures, where fasteners can be added to the laminate to clamp them together and prevent plies from coming apart.
“For parts that need transverse strength, there are 3D woven materials that can be used. You essentially can braid a 3D carbon preform and infuse it with resin,” Dubois added. “That way you can effectively control the direction of each fiber and you can potentially run it not only in the XY-direction but also the Z-direction. There are a lot of interesting projects around that; over time we will see more and more of these, and they will alleviate the current weaknesses in the status quo of the state-of-the-art.”
There are other factors that can affect the physical properties of a composite. For instance, the type of force inflicted on the composite can affect how it reacts. As an example, Dubois said that “a lot of times, composites will behave differently if you load them in tension or compression. That can sometimes differ by a factor of three or four.”
Temperature, humidity and even the slight variations between the manufacturing of different batches of the same material can drastically change the composite’s properties. Since many manufacturing processes for composites still contain humans-in-the-loop—more on that later—this variation can be further amplified.
“You may see a 50 percent variation in properties for the same material,” Dubois clarified. “So obviously, that’s a challenge to engineers. You’d love to have a material where you know exactly what strain it is going to fail at, but that’s not the case for composites.”
“At the end of the day, it comes down to the properties and how they vary in different directions. That’s why designing with composites poses additional challenges, because it’s not a one size fits all answer. You really have to tailor your design to load cases and the different types of failures that a composite material can see.”
The good news is that this variability does not affect every simulation the same way. For instance, Dubois noted that it doesn’t have much of an effect in simulations that assess the stiffness or frequency of a structure made from composites. However, he clarified that “where it becomes a challenge is when you’re trying to predict failure.”
How to Design and Simulate Composites with Variability in Mind
Let’s assume the part you are designing requires you to understand how it might fail. In these scenarios, simulating that failure becomes tricky if it includes composite materials. So, what can be done?
For Dubois, it all starts with the engineering fundamentals. “Start with hand calculations. You should know—generally speaking—what the structure is going to do before you start simulating. I have a rule on my team that you don’t start simulating until you’ve shown a hand calculation. Before you jump in there, think a bit, run some calculations and get a sense of what you think is going to happen.”
He also notes that you may have to adjust the simulation approach based on the failure mode you wish to assess. Engineers will need to think of all the failure modes of interest early in development and adjust their analysis strategy accordingly.
“Take laminate failure,” he said. “If, for example, there is a lot of thickness and curvature in the material, it might force you in the direction where you have to use a solid model versus a shell model to capture interlaminate tension effects. Another example is if you have non-linear failure modes, like post buckling and crippling, you will have to adjust your modeling approach.”
“In a lot of cases you’ll use a hybrid method where FEA isn’t the end of the story. You might use FEA to get closer to the problem, but you write your final margin with hand calculations,” Dubois added. “I think the key is adaptability and you have to understand the material variability. You must properly account for the environment that the part is going to see and then what’s going to happen. Those are the three pointers I would go with: hand calculations, failure modes and material variability.”
As for that material variability, how do you include that into the simulation? In this case, Dubois turns to lots of data and statistical analysis.
“When you derive your material limits, like the strain cut-off that I’m going to consider the laminate fails at, you have to derive them from a large number of material tests coming from different batches of materials,” he said. “Multiple coupons, of different types, need to be tested by different technicians at different temperature and humidity. The idea, and I’m simplifying a bit, is that you gather this dataset and from there have the bell curve of material properties. From that you can say, for example, I’m going to choose my average material properties minus three standard deviations. That way you can mathematically ensure that 90 to 95 percent of all the material batches that you’ll get, at all the possible environments, will result in a structure having higher properties than this lowball number.”
Now, this might seem counterintuitive to the whole philosophy behind the use of composites, which is to lightweight designs without compromising on the structure. It doesn’t appear to make much sense when the process to design parts with these materials includes a step that can be summarized to the old engineering adage: ‘when in doubt, build it stout.’
In response to this, Dubois said, “It’s fair to say that you’re sort of negating some of the advantages of the composite by adding these knockdown factors. That is the accepted aerospace approach. Recently, I have seen some new techniques being proposed to assess probability of failure of a structure given the presence of certain flaws, instead of just assuming the flaws are always there. You could, for example, have a more granular approach by running a Monte Carlo simulation with a number of flaws in a number of places to see the probability of an actual catastrophic failure. If you can show that the probability is exceptionally small you can effectively still make the argument that you have a safe structure.”
Another challenge behind simulating composites, according to Dubois, is learning how to read the results of the simulation. For a metal, it is simpler; you plot the Von Mises stress and go from there. For composites, it is a lot more complicated. You need to go into each ply and understand why it is, or isn’t, failing and relate that back to the allowed limits of the material.
“You’re going to have to look through the laminate at each ply. I think the one tip I would give is to make sure, before you jump in, that you understand composite laminate theory. It will get you a sense for what should happen in a composite laminate under a load. Then, when you get the outputs of your simulation, you’ll be able to know what you’re talking about more fluently than if you were looking at pretty pictures of an FEA solution.”
A source that Dubois recommends, and uses, is Practical Analysis of Aircraft Composites by Brian Esp. He notes that you can find a lot of information in many legacy books but that that Esp did a good job collecting all that information thoroughly.
“It’s a really good book,” he said. “I have almost each person on my team read it when they start.”
How the Manufacturing of Composites Can Simplify their Simulation
Since the primary complication of simulating composites is their variability, ensuring their consistent manufacturing will go a long way to reduce that hurdle and ensure that the final design is as lightweight as possible.
“The first thing is to tightly control the process and material,” Dubois said. “The way you cure the carbon, and the way you source it, needs to be tightly controlled such that you are maximizing your chances of getting a consistent material every time. Your layup process and the tolerances of each ply must be defined carefully, and you have to have a solid quality control plan to make sure every time you lay down a ply that it’s in the same location plus or minus a very small percentage.”
To ensure tighter tolerances and lower costs, consider automating the manufacturing process. As previously mentioned, many composite manufacturers still use humans-in-the-loop; however, there are newer technologies, such as automated fiber placement (AFP) and automated tape layout (ATL) which utilize robotics to lay down the plies in a more repeatable way.
“I think in the next 10 to 15 years we can really hope to see some of these approaches reducing costs and variability, allowing us to take even fuller advantage of the performance of composites,” Dubois said
There will still be some variability and unknowns in this automated manufacturing process. As for how that could affect the simulation, Dubois notes, “you have to not only control really tightly what you do, but for the stuff you can’t control, you also have to assume it’s always the worst it could be.”
What Simulation Software to Look for When Working with Composites
As for which simulation software to use, Dubois noted that the only people who can answer that question are the people working on the specific project in question. However, he does offer some guidelines.
“It really depends on what you’re trying to do,” he said. “What I can say for static simulation is that most of these codes do a great job, whether you’re talking NASTRAN, Ansys, Abaqus, you name it. They will do the trick because the simulation of composites in a linear static sense is well understood. If you’re doing more complex analysis, say a dynamic explicit analysis, there are a lot more differences.”
Dubois added, “What I’ve seen, though, is that the accuracy often depends on the way that you model failure and the onset of damage in the laminate.” As a result, he notes the importance of choosing software that includes the ability to inject real-life test data into the damage model and simulation.
As for which software he has used, Dubois said, “I haven’t done a full survey of the market, but I’ve used, or seen others use Abaqus and LS-DYNA fairly successfully to do this sort of dynamic explicit analysis of composite impact events.”
As for any of the software Dubois hasn’t used, as long as it meets the requirements he notes above, they should be good candidates. But it might be a good idea to test and evaluate the software you’re thinking about using for your individual application, even if it’s one he recommends.
As for the software features Dubois recommends, they are:
- The ability to assess and visualize the results on a per-ply basis.
- A user-friendly interface for pre- and post-processing.
- A library of materials data and failure modes.
- The ability to create custom routines to model specific materials and failures.
- A data thread that auto-updates data, changes and results between CAD and simulation.
- Parametric modeling with design space exploration and optimization capabilities.
Resources:
