How to analyze a 3D printed part: a matter of meshing
When I was working for the automotive industry, I heard the phrase “additive manufacturing” thrown out in prototyping. Later on, I was briefly introduced to the philosophy behind the practice. My impression by then was that AM was fun, for hobbyists mostly, for automotive prototyping maybe, for leisure all in all, but it might not be the serious competitor to tooling. Even up to last year, I was thinking that 3D printing could have a future but only within the automotive industry. Boy was I stunned when one of the giant aeronautics builder seriously engaged in it! And not only for prototyping, mind you!
Aeronautics and aerospace are getting serious about additive manufacturing
Aeronautics is the field working on advancing AM methodology so it does more, and it is currently doing more. Building actual nacelle parts and other subparts within airplanes, for example! Airbus and NASA were leading the movement, not because of trend or on a leap of faith, But because those guys are extremely serious about their next step.
And since it’s all about decreasing fuel consumption (aka decreasing weight, aka exploring every path in R&D), additive manufacturing was given strong consideration when thinking about how to reduce weight. So, when you see aerospace, aeronautics and defense jumping on the practice, you should give it a good thought.
And in my case, you should brace yourself to analyze that part, and that is a whole fresh but big challenge ahead.
On FEA applied on 3D printed Parts: it’s not standardized
Finite element analysis might have flourished over the past decades, and most might be able to run a 3D printing machine after some classes, but rallying the two is still not a beaten track. In my case at least, the process wasn’t clear since nothing has been officially done and approved, let alone considered a certified practice. Sure, my client couldn’t stop using the word “composite” as if it was just a matter of treating the part through composite theory and methodology. I was looking into the procedure itself because it will deliver typical strengths and weaknesses to the product. When you ask around, fellow consultants rack their brains and offer suggestions. But all in all, it’s just not standardized and ready.
Through my personal experience, I will share my own track. Be free to criticize it however you like! I just hope it will be something fresh that I haven’t already heard from my witty project manager who takes me for a qualified noob and therefore let’s me “investigate” my way through iterations, or my lost client who takes me for an analysis wizard and thinks I detect void defects within his printed parts with my third eye.
Get your input data straight: know your part
Here is the first thing I understood: I might have “meshing the part” as a first step written on the requirement file with some numbers to meet regarding elements quality and whatnot. But I have been in the mechanical analysis and design long enough to know that understanding your input data is the crucial step. In the case of a 3D printed part, it is imperative that I understand the machine’s capacity and limitations, the parameters –as much as possible since confidentiality is a trending issue- involved in the manufacturing.
The material used, along with its properties is a must as well, and the final CAD of the part (because meshing might be applied to a “cleaned” CAD with no rough edges or holes) along with the error ranges obtained after manufacturing should be required. I’m sparing you the failed iterations that you would go through if you trust your client and “just treat it as a laminate/composite and that would give basic results.”
In such cases, I keep in mind an advice that my project manager gave me on one occasion: “The client would certainly not hire you if it was as simple as putting it in a sentence and doing it in three clicks. Get your work right and take every bit he can offer then proceed to hustling the rest from him if you can, or from your station if you can’t.”
Meshing your part…get your input data straight
The second rule about analyzing 3D printed parts is to get your input data straight. You’d think making it the first rule with a detailed reasoning would suffice but it didn’t help Durden and he knew emphasis might be better….so, I might just use his method.
I’m stating it and I can’t emphasize enough on this point: when you say “additive manufacturing,” you basically bring up a versatile range of possibilities in shapes and efficiency. We have previously spoken in the GrabCAD blog about how analysis could be greatly accelerated if 3D CAD had 2D CAD basics.
If you’re going for additive manufacturing, your shape will most likely not be very symmetric or axisymmetric. It might have parallel surfaces but there might be holes within one, or tunnels for wiring and/or lubrication or strange twists within the other…etc. Therefore, not only does the analysis get harder, but you will have to tackle a 3D part with shapes that won’t just let you go hexa/penta through most of them. Plus of course, your customer will compulsorily place a maximal number of nodes, a certain percentage of pyramid elements to not go beyond or, worse, tell you to do it your way then, once you’re done, start to complain about radiuses with not enough elements and regions with too much tetra and whatever he looks at and considers it “an ugly mesh.”
What helps therefore in meshing is to “know your part.” I can never put enough emphasis on that. If you don’t know your part, don’t know where it works, how it works, the limitations of its shapes or its materials, you can’t get an efficient meshing going on. Drill your customer about it if necessary. Ask him/her to put the actual designer on the phone. DO SOMETHING ABOUT IT, but don’t you dare just take a 3D CAD, look at it, then start meshing….nobody is interested in “your” methodology!
The point is that you get to have a primary understanding of the regions that matter, the regions that don’t, the zones where there is contact, where there is a kind of transfer, the zones that were tricky to draw, or that the customer has a preference to…if you can gather such data, then at least your priorities are set for a first iteration and you can see meshing as a logical process and not just an ongoing series of iterations until “it’s beautiful” or “it’s the budget’s limit.”
When you know such surfaces and shapes, you know where your meshing can be refined and to what extent. You will also understand where your main efforts will be located to annihilate pyramid/tetra as much as you can. You will finally see that the whole part isn’t frightening and you will be able to mesh a lot without worrying about criterion.
A 2D surface to insure surface analysis
Even if the part is 3D and obviously required volume meshing, creating a “skin” is still an easy 2D feature to introduce and will prove useful in analysis. The exterior layer meshed in second degree elements allows you to assess interaction with the exterior with superb accuracy: you therefore ensure the most thorough boundary conditions/environment interaction modeling and results observation.
Finally, surface analysis shows the significant and predictable regions where volume analysis should be performed. It narrows them to a great deal since you can’t possibly go through every bit of volume. It’s easy to create 2D surfaces from a volumic mesh: the option is available on default on most software that I’ve worked with. Ansys Workbench goes at it straight and just requires that you add it as a comment so it generates it. Plus, you can simply isolate all the surfaces and proceed to meshing them with constraints to follow the volume mesh.
3D printed parts and 3D meshing: a tricky compromise
If it was up to my customer, he’d love it if I meshed every thin layer with nice hexa for most of them, then pile them and model rigid contacts between them. But even my customer has enough wits to not dare ask for 100micron dimensions in elements and to understand that he doesn’t have the money or the power to run billion nodes worth of data that 6 people would have slaved for days to model with all their conditions and theories.
Because that is the thing I faced with my printed part: 3D printing allows us to go wild, so wild with manufacturing! But FEA hasn’t caught up yet with the wildness. Sure, they’re working on dynamic meshing, but until we get a set of nodes that can be as groovy as the shapes we give to our 3D printed parts, we must be extremely patient and learn to iterate, review, and reiterate. Why you say? To get the best possible meshing? And what is the best possible meshing for a wild 3D printed part? It’s one that gives you a good compromise between calculation time, results convergence and bulk of nodes and elements in there. And how do you assess compromise?
Well, personally, I inquire about delivery time and days allocated to the project. If I know I can analyze the results and present them within X days, I will decide to give the rest to meshing and boundary conditioning. Depending how intricate boundary conditions can be, I might end up with 3 – 5 days for meshing and this is when I present to my client the allocated time for meshing, agree with him on a time limit and go through my iterations, register the results, seek convergence and limited server calculation time, and most importantly keep the bulk data manageable, because I won’t get a 64RAM station with an awesome Graphic card and triple sets of proc just because I’m over a billion nodes.
I will most likely get a lecture on how to not suck at setting up FE models and to suck it up a little bit.
Give the external layer your best!
So yes, you think you have five days to work on your meshing? Give the external layer your best! Refine on important regions and pay close attention to radiuses and to specific spots that would bear significant loads. Then run a first mesh with not too many constraints, and gradually increase its quality and look for a homogeneous pattern…you’ll know you hit it when your mockup iteration will give you similar results in quality or basic stress to the previous ones.
Once you hold that mesh, you can discuss the official approval with your customer (if you have one of course, if not, ask for two more opinions! Even your partner can intuitively state if a pattern is “beautiful” or “weird in those spots”. Then move to the next steps: applying the actual boundary conditions and go cuckoo over the results…but I’m not spoiling those parts, it will be for the next time, if this didn’t beat up your analytical spirit!
We're making professional 3D printing easy and accessible
There's a new solution coming your way this summer. Click here or on the graphic to the left to request updates and beta access
About the author: (Khadija Ouajjani)
Mechanical design engineer working for the aeronautics industry, obsessed with improving CAD to improve analysis, doesn't accept numbers without a precision scale, bookworm and aspiring writer, obsessed with garlic and Mandelbrot, loves tea with mint and coffee with cinnamon. When not writing for GrabCAD, she's probably drawing fancy propellers and coloring cube trees in a coffeeshop.
All posts by Khadija Ouajjani