What is Design for Additive Manufacturing?
If you work in design engineering, then you have probably heard of the design for manufacturing (DFM) and design for assembly (DFA) philosophies.
DFM concerns itself with the best practices to allow for the optimal manufacture of parts with the goal of keeping costs down. It achieves this by addressing manufacturability concerns early in the design phase, so they don’t cause issues further downstream in the manufacturing process.
For example, when manufacturing a metal part with a milling process, using aluminum with just the right balance of material properties compared to steel can save tooling costs. Steel is harder than aluminum, and tools wear out a lot quicker when cutting steel. So this particular manufacturing issue has been addressed early on—money is saved, and engineers (and the finance department) are happy.
Likewise, DFA seeks to reduce costs by factoring ease of assembly into the early design. Picture the aluminum part mentioned in the DFM example. Imagine that it needs to be mechanically mated to another part, thus making it an assembly of some kind. How do you mate it? Maybe you wish to use a series of fasteners to connect the two. Certain fasteners are easier to work with. In any case, fasteners typically make up 5 percent of bill of materials cost yet contribute up to 70 percent of the labor cost. So maybe fasteners aren’t optimal in this case.
These design considerations are the crux of DFM and DFA. And thanks to additive manufacturing (AM), you can now ignore them both. Wait, what?
Yes, with new manufacturing methods and new software capabilities, a new paradigm in design philosophy is here: arise, design for additive manufacturing (DFAM)!
After all, why build a machine part with 200 pieces when you can just print one in metal and forget about the whole assembly stage altogether. 3D printing can remove the need to assemble, as all parts are printed in place. Take, for example, the 3D-printed rocket fuel injector, revealed to the public by NASA engineers in 2015 (see Figure 1). A traditional injector of this type made with traditional methods would normally consist of over 200 parts. The 3D-printed version? Just two parts.
Figure 1. 3D-printed rocket injector. On the left, we see the part as it has emerged from the printer. On the right, we see the part after it has been cleaned and polished. (Image courtesy of NASA/Marshall Space Flight Center.)
Obviously, this did not happen by chance. NASA engineers didn’t just load up a CAD assembly for the old injector, send it to a 3D printer and gasp with surprise when the printer revealed the new part in two halves. No. They had to design it this way from scratch, using a DFAM design philosophy.
So now you know what DFAM is. How do you accomplish it? I will share a few secrets with you.
Forget everything you know about DFM and DFA. As mentioned, these old rules do not apply to AM. Remember in the old days when you designed dies for plastic injection molding? Where you had to design draft angles to allow for part ejection and you had to make allowances for shrinkage? You don’t need to worry about that anymore, because the printer software will compensate for the shrinkage, depending on the material selected. You can even design negative draft angles now and not have to worry about the part getting jammed into the die. Also, you don’t need to worry about designing mating surfaces and bolt holes … because, hopefully, there won’t be too much to mate. And if there are many parts to mate, then maybe you are not using AM to the best of its abilities.
Think fresh. Do not allow yourself to remain stuck to old designs. Yes, they may work, but that old rocket injector was designed in simpler times, and its design was dictated by the design rules of the time. Ask yourself—what does 3D printing allow that wasn’t possible before?
Choose the best material and process for your design. This is perhaps an obvious point and applies to traditional manufacturing as much as it does to AM. Different printers will print different materials and will have different properties and different surface finishes, as well as different post-processing requirements. As a design engineer, you will likely have an idea about which material and printer you will be using before you open your CAD package. A nonfunctional aesthetic prototype can be printed without need for much surface finishing by use of PolyJet or stereolithographic apparatus (SLA). However, a functional fused deposition modeling (FDM) part made from ULTEM or ABS plastics may need sanding, and a metal part may require additional machining. These considerations must be made in the early stages of design.
Design for support. Most types of AM processes require some form of support structure. This is for three main reasons: to allow the printing of overhangs, to anchor the part to the print bed to keep it static during build and to reduce the distortion of the model by anchoring the model material to the bed.
Generally, Magics software will generate different types of support structure geometry according to user requirements for most kinds of AM processes. So what are the considerations that dictate the support geometry?
With FDM, it’s pretty straightforward. You have two options in terms of support material (soluble and breakaway). With highly intricate parts, you may opt for the soluble type and simply immerse the part in an alkaline heat bath until it dissolves. However, the heat in the bath itself can generate eddies that may cause the part to collide with the bath walls. Again, this can be countered by thickening the parts where it is likely to break.
Also, be aware that, for thin grooves, the fluid may not permeate properly into said groove, and the support may not be flushed out properly, resulting in the need for mechanical removal. As a personal anecdote, I printed a solar cell mounting recently that featured small slots for the installation of the cell (0.254 mm thick). The fluid did not flow into these slots. So I tried leaving the part in the bath for longer. After six hours in the bath, the support material within the slot was still very much undissolved. I had to redesign the part with a slot 0.58 mm thick to allow the fluid to flow inside—and it still took three hours to dissolve! (Notice how my slot thicknesses are multiples of 0.254 mm? More on that later.)
Also, the heat bath method may not be suitable for certain parts containing embedded electronics (which is becoming a popular trend, especially in research institutes). Fluids and electronics tend not to mix, so, in that case, you may opt for a breakaway support structure.
If you do opt for a breakaway structure, then you need to be wary of accessing hard-to-reach places and the risk of breaking protruding features. Breakaway structures are most easily removed from flat surfaces … they just peel off like tape. More complex geometries are a lot trickier. As with the heat bath method, designing thicker features that may otherwise break during the removal process can save tears later on.
Of course, you can always use the self-supporting method. Generally, features designed at a 45-degree angle can support themselves, although this can add material to the model itself. To combat this, 45-degree angles can be combined with lattice type fills in order to create a less dense internal structure while retaining a solid skin.
For metal printing systems such as electron beam melting/selective laser melting (EBM/SLM), it gets even more difficult. Although Magics can autogenerate the structure itself, the type of structure should be selected by the designer based on the needs of the part.
Unlike FDM systems, which can print supports in different (soluble) materials, EBM and SLM must, by their very nature, use the same material as the model material. That is to say, when designing a part for these processes, you are effectively welding the support to the model. And that means a lot of effort when removing. In general, the minimal amount of material should be used that will prevent overhangs from collapsing and parts from distorting, all while remaining easy to remove.
Design for finishing. As mentioned, the post-processing requirements will be dictated by the design requirements and the material itself and will need to be determined early on. Materials like ULTEM have high tensile and compressive strength and are incredibly sturdy. ULTEM can be shot blasted, dry sanded or wet sanded or can even be mass finished by use of centrifugal and vibratory systems. Vibratory systems can be fairly aggressive, which increases the risk of damage to thin or finely detailed parts. If you are planning to use a vibratory method with an abrasive medium, then it may be worth strengthening the weak parts. Or it may be worth using a centrifugal method, which is safer but slower. The abrasive media itself can determine the risk and speed of the process. Heavyweight media such as ceramics can decrease processing speed while increasing risk of damage; lightweight materials such as synthetic media can have the opposite effect. In general, such decisions should be considered for mass production of AM parts. If you want to think about it in terms of process optimization, where finishing time for many parts is the output, then you really need to consider this in the early stage.
As always, it is a tradeoff.
Design in multiples of layer heights to maintain accuracy. Remember how I mentioned that I designed my slots using a multiple of a certain number? Well that certain number was the layer height/z-axis resolution. My preferred 3D printer of choice is the Fortus 450mc. It can print with layer heights of 0.330 and 0.254 mm given my material of choice, ULTEM 9085.
Say for example I wanted to print a surface that is 2.4 mm tall. With a layer height of 0.254 mm, this would not be possible, and there would be a difference between the CAD model and the final part (because 2.4 is not a multiple of 0.254). If I were to design the part without wanting any surprises, I would design my feature height for 2.286 or 2.54 mm, which would provide me an accurate part.
Make strong walls. How strong is strong? Strong enough to not collapse or distort is the quick and easy answer. As with the previous point on layer height and accuracy, walls designed in FDM should be at least two filaments thick and should be designed to multiples of the filament track width to maintain accuracy. For example, using a Fortus T20 extruder tip will yield a track width of 0.48 mm. Therefore, the best practice to maintain accuracy in the x/y direction would be to design feature thicknesses to multiples of 0.48 mm within the CAD design phase. Catering to these perturbations of accuracy in the early design stage can save broken hearts later on in the build process.
Design and optimize topology. Topology optimization allows a designer to take a traditionally shaped part, simulate where the loads will run through that part and remove the superfluous material, leaving only the load paths remaining. This removes weight while retaining the outer shape of the design (see Figure 2 as an example). There has been a lot of academic research about making topology optimization a reality, so you don’t have to worry about the math so much (but feel free to try doing it by hand if you are feeling brave/insane). New software is coming out all the time that automates this process. Speaking of which …
Make full use of new software. And here is some of that software. CAD has been around for decades. So has 3D printing. But finally, new CAD capabilities are being developed to meet the new demands of 3D printing.
The old-school CAD companies are leading the charge in this. SOLIDWORKS 2016 has 3D printing capabilities that can help optimize the workflow for AM products. SOLIDWORKS, Autodesk and PTC are all moving toward greater integration within their respective software. Gone are the days where you have to design in one package, convert to a different format, simulate in another package and send back to the original software for changes (ad nauseam) before sending to a third slicer package. Software exists to aid with the DFAM process—so use it.
Optimize STL files. Your beautiful design is finished in CAD, and it’s time to convert and export your file for printing. STL files, for the time being, are the most popular file type used in 3D printing. There are several parameters within the STL file that can affect surface finish, upload times and processing times. An STL file is made up of facets (small two-dimensional triangles) that represent the polygonal geometry of the CAD model. In curved surfaces, this can be a little problematic. The difference between the curves in the original CAD geometry and the faceted STL file is referred to as the chord height, and this parameter is generally controlled with some form of chord height tolerance option. In SOLIDWORKS, this chord height tolerance option is accessed via Save As > STL and by clicking the Options button on the File Export window. It is referred to in SOLIDWORKS as “resolution,” and you are offered three options: course, fine and custom (see Figure 3). Other CAD packages use a different name for this function, but the function is the same—it controls the resolution by altering the chord height, which, in turn, alters the number of facets. In AutoCAD, it can be adjusted with the FACETRES value. In IronCAD, it is changed via the Set Facet Surface Smoothing parameter and so on.
Figure 2. Screenshot of SOLIDWORKS resolution options menu.
A small chord height tolerance will reduce the delta between the CAD part and the true STL part curvature (and will enable a smoother finish); however, it will also increase the number of facets. More facets equal larger file sizes and, hence, more processing time. Conversely, fewer, larger facets will reduce the file size and processing time but will create a part that deviates from your original part geometry.
Figure 3. Donut A is the CAD model, donut B is the coarse resolution STL and donut C STL has fine resolution.
As a comparison, see Figure 4 for an example. I modeled a ring torus in SOLIDWORKS (technically known as a “donut”) with the main radius, R, being 40.3 mm and the cross sectional radius, r, which provides the thickness, being 12.3 mm. Donut A is the model as it appears in the CAD software, all smooth and shiny. When exported as an STL with coarse resolution (donut B), it has a chord height of 0.18 mm, yielding 2,808 facets and a file size of roughly 140 kB. Donut C is exported with a custom (fine) resolution and a chord height 10 times less than donut B (0.018 mm), yielding 10 times as many facets (28,220) and a file size 10 times greater at 1.4 MB. Naturally, the smoothest, curviest part will be the one manufactured from donut C. The optimal design will depend on your requirements but will generally lie in between both extremes.
Processing time and file size can also be affected by multibody parts and assemblies. It is advised to split up assemblies and multibody parts into separate STLs rather than exporting all the parts in an STL. Not only does this speed up processing time and reduce file size, but it gives you greater freedom when orienting separate parts on your build platform.
Orient Parts for Strength
On the subject of part orientation, when it comes to laying out your parts in the printer software, you should be aware (particularly with FDM parts) that there is a huge difference in part strength depending on how they are printed. Currently, STL parts only carry surface geometry information, so you must orient your parts manually.
Unlike injection-molded parts, FDM parts are anisotropic. That means that the material strength is not uniform in all directions, and the difference according to build orientation is pretty drastic. If you want stronger parts, then it is imperative to print them in a manner that will allow any mechanical forces to be dispersed longitudinally along the length of the filaments to maximize strength. Figure 5 shows the orientation nomenclature and Figure 6 shows the stress/strain curves for each of the respective orientations. As you can see, parts printed in the X-direction are significantly stronger.
Figure 4. Part orientation of a tensile test bar. (Image courtesy of DMRC.)
Figure 5. Stress-strain curves for the aforementioned tensile test bars, according to each build orientation. (Image courtesy of DMRC.)
So there, in a nutshell, are a few tips to get you started on the path to DFAM.
Just to recap: Forget everything you have learned, because it’s useless. Don’t get hung up on old designs, because they were based on old design rules. Buy some new software that will allow for some topology optimization. And make sure that your parts are rugged enough from the beginning to facilitate the removal of support material.
Here’s a takeaway quotation from 3D Systems’ former CEO Avi Reichental that summarizes the entire philosophy of DFAM:
“In AM, complexity is free. So step back away from the constraints of old. Just let your mind soar, design what you want and let the software do the rest.”
About the Author
Phillip Keane is currently studying his PhD at the School of Mechanical and Aerospace Engineering at Nanyang Technological University, Singapore. His background is in aerospace engineering, and his current studies are focused on the use of 3D-printed components in spaceflight. He previously worked at Rolls-Royce and Airbus Military and served as an intern for Made In Space and the European Southern Observatory.