Over the past years, I have gotten deeper and deeper into 3d-printing. One
thing that fascinates me a lot is how different it is from other manufacturing
methods. And how this, in turn, means a completely different design philosophy
is needed to create good designs for 3d-printing.

As such, I have been collecting the little tricks and rules for designing
well-printable parts. And of course I am always on the hunt for more. In this
blog post I want to share all that I have learned.

Introduction

While a lot can be found online about the basics of design for 3d-printing,
there is not that much material that dives into it in-depth. And even less
common are good overviews that collect all the little tricks that people have
discovered.

I believe the best format for teaching the rules of well-printable design is a
format that is built on rules of thumb and many practical examples. While such
heuristic rules will never encompass the whole complexity of the real world, I
think they can provide a tangible guide towards the best solution in each
situation. They also provide a starting point for building your own intuition
about the topic.

This approach is close in spirit to the german book Konstruktionspraxis im
Maschinenbau by Gerhard Hoenow and Thomas Meißner (Hanser
Fachbuch, ISBN 978-3-446-46485-8). The book follows
the same philosophy, but for traditional manufacturing methods. If you like
this style, that book might interest you as well.

Scope

Before diving in, a few words about the scope of this blog post: First, I am
explicitly focusing on FDM/FFF printing (Wikipedia). A lot of
the rules are specifically tied to the restrictions of this process and do not
apply to other additive manufacturing methods.

Second, I am focusing on designing functional parts with optimal mechanical
properties. These parts should be easy to print, without requiring much
printer fine-tuning. They should be easy to produce in higher quantity, which
means they should require as little post-processing and waste as little
material as possible. Aesthetics do not play a primary role, although I firmly
believe that great mechanical design just looks incredibly beautiful :)

Goals of Design Engineering

Any mechanical design ultimately must be optimized for a large number of
objectives and constraints. Your job as the designer is to create a part that
fits best between them all.

The most universal objectives are:

• Design according to force — Parts should be designed such that their
  geometry is optimal for the forces it has to transmit and withstand.
• Design according to manufacturing method — Often called DFM
  (Design For Manufacturing), parts should be designed such that they are
  easy to produce. Little changes, which usually don’t affect functionality,
  can often make a part much simpler to manufacture.
• Design according to cost — Parts should be as cheap as possible.
  With 3d-printing, this means minimizing material use and print time.

Of course, any real world project will have many more objectives and a lot of
other constraints that the design must comply with.

In this blog post, I will focus on topics related to design for manufacturing
with a 3d-printer. In general, the rules for other objectives are the same,
regardless of manufacturing process. However, there are some places where
3d-printing does influence these other objectives and I will also highlight
such cases.

An important distinction that I need to make at this point is how the approach
of design engineering attacks the problem from the completely opposite side
compared to, for example, slicer developers and 3d-printer manufacturers.
While they aim for improving the manufacturing process itself, to better print
any possible part geometry, a design engineer instead aims for adapting the
part geometry to best work with the current state of the the manufacturing
processes.

To give an example, a design engineer aims to use the most coarse tolerance on
each dimension that still allows the part to function. A 3d-printer
manufacturer wants their machine to print with the most tight tolerance they
can achieve. While the design engineer could try making use of this most tight
tolerance, this will make the part impossible to manufacture using cheaper
printers and thus drive up cost and limit possible suppliers for a part.

This can be summarized as aiming for portable design. A portably designed
part can be printed by anyone on any printer with ease. If you intend to share
your designs online, this should definitely be a priority.

Finally, it is also important to keep in mind that due to the rapid innovation
in the 3d-printing scene, we are working with a moving target. As the
3d-printers and slicer software get improved, some of the design rules will
become less important.

Terminology

To make sure everyone is on the same page, let’s quickly review the most
important terminology of FFF 3d-printing that will be used in the following
chapters:

• Layers — The horizontal cross-sections of a part that are stacked
  on top of one another to create the part (Layer height — Prusa
  Knowledge Base).
• Perimeters — On each layer, the 3d-printer will first print the
  outline of the cross-section. This outline consists of a number of perimeter
  lines (Perimeters — Prusa Knowledge Base).
• Shell — If you think of only the perimeters on each layer, you get
  a completely hollow object: The shell of the part.
• Infill — Inside the perimeters, the 3d-printer “fills” the space
  using a partially hollow pattern. This is the infill pattern (Infill
  — Prusa Knowledge Base).
• Infill Percentage — How much of the infill pattern is solid vs.
  empty space (Infill — Prusa Knowledge Base).
• Overhang — When the cross-section of a layer gets larger than the
  previous layer, a part of the perimeter lines will not be supported by
  material from below. This is called an overhang and is actually possible
  as long as the overhang is not too large. Mostly, people talk about the
  angle of an overhang when looking at the layers from the side.
• Bridges — While overhangs can become problematic, one thing
  3d-printers are surprisingly good at is bridging gaps between two supporting
  surfaces. This is called a bridge.
• Seam — The seam is the point where the printer starts and stops
  laying a perimeter. This point is usually well visible as a perfectly smooth
  transition is very hard to achieve (Seam position — Prusa Knowledge
  Base)

The Standard Printer Profile

One final topic before we can dive in. To achieve the aforementioned portable
designs, it is useful to define a design target for your work. Namely a rough
outline of the 3d-printer and profile that the design will be printed with.

While most of the following rules are relevant regardless, I will need to make
some assumptions. Especially for all the numbers I will mention, a design
target is needed for context. Otherwise values for expectable tolerances and
empirical dimensional recommendations will have no meaning.

As such, here is my definition of the standard printer profile. Cutting
across the current market of 3d-printers, the vast majority are compatible with
the following expectations. A lot of the expectations are intentionally kept
quite vague — we must not assume too much, to keep the designs
portable.

• The standard 3d-printer is using an 0.4 mm nozzle.
• The standard profile prints at 0.2 mm layer height.
• The standard 3d-printer is calibrated such that no considerable dimensional
  deviation is caused by error in the steps/mm setting. Skew between the axes
  is negligible.
• Print speeds are tuned to an adequate level but we cannot assume the printer
  to be free of artifacts like slight overshoot and ringing.
• Bridging across reasonable distances is reliable.
• Overhangs print fine.
• Bed adhesion is adequate and does not pose an issue for non-extreme geometry.

Table of Contents

1. Designing for Part Strength
2. Manufacturing Tolerance and Part Finish
3. Process Optimization
4. Functional Integration
5. Beyond plastic – Machine Elements
6. Appearance
7. Extra: Vase Mode Design

1. Designing for Part Strength

The first rules I will present are related to designing parts that are as
strong as possible. In 3d-printing, the strength of a part is inherently tied
to the manufacturing process. For one, the parts are partly hollow and thus
behave quite differently than a solid block of plastic. Second, the layered
manufacturing process leads to anisotropic parts — meaning their
mechanical properties differ depending on direction (Wikipedia).

Thus, beyond general design rules for part strength, a few additional
considerations should be made when designing parts for 3d-printing.

Part Orientation

Let’s start with the most obvious: 3d-printed parts are much weaker in the
direction of pulling apart the layers. Thus, before starting on a design, you
should first consider the direction of tensile forces in your part and use them
to decide on a print orientation. All further design should then be done with
this print orientation in mind.

R1.1
—
Tensile forces should be aligned parallel to the print surface.

This is probably the number one rule that everyone talks about in design for
3d-printing. And that is for good reason because the difference it makes is
substantial. Tests performed by My Tech Fun indicate roughly 3 times more strength in
ideal orientation (YouTube). While the exact ratio
will depend heavily on the individual part, I think this is a good indicator of
how important this topic is.

You should also keep in mind that tensile forces are exerted to some cross
section of a part when bending it. This means that you should also align any
bending moments parallel to the print surface. A place where this often fails
is printed clips, which bend slightly to clip into their counterpart. Such
features will break easily when not oriented correctly.

I also want to give a small tip here about sharing your designs on platforms
like Printables: Make sure you upload your
models in the correct print orientation. While the correct orientation may
seem obvious to you, others might not know and they will try printing your part
without reorienting it first. You’ll do them a small favor by uploading
correctly oriented files 🙂

When no orientation works

Especially with complex parts, in some cases, no print
orientation is ideal. While it is often fine to make a trade-off and accept
limited strength in certain locations, one alternative should always be kept in
mind:

R1.2
—
Split a part into multiple pieces when no orientation is ideal for all loads.

As each piece gets printed individually, you are completely free to choose
the optimal orientation for each one. Clever joints can be used to easily
assemble the whole part afterwards. Just to mention one: Dovetail joints have
proven themselves for 3d-printing, as they are well-printable in most orientations.

To infill or not to infill

Contrary to the first rule, a topic that is not talked about enough in my eyes
is the role of infill in part strength. There is a big misconception that you
can magically give your part ultimate strength by just using 100% infill. While you
will certainly see an increase in strength, this method is not efficient at all
— most of the additional material does not contribute to part strength
and is just waste and an unnecessary increase of print time.

Why? It all has to do with the distribution of force inside of a part.
Generally, most parts are not stressed in pure tension or compression. Instead,
a lot of stress is applied in the form of bending moments and those distribute
the force unevenly. The force will be greatest in places furthest away from
the center (neutral axis). Thus, it is generally more effective for additional
material to be added at the surface, not the center of a part.

In 3d-printer speak, this means you should increase the number of
perimeters/shells rather than the infill percentage. Stefan from CNC Kitchen
has done a thorough analysis of this (YouTube).

In general, his work is a great resource for in-depth information about the
mechanical behavior of 3d-printed parts (CNC Kitchen Blog/Website
and YouTube).

R1.3
—
Most strength comes from the part’s surface, not the infill.

The Flow of Forces

However, there is more to this. We can actually influence the amount of stress
at the surface by changing the shape of the part. This is often much more
effective than optimizing print settings. A way to conceptualize the stress
inside of a part is to think about the forces “flowing” through it. The visual
representation of this is called a force lines drawing
(Wikipedia).

Stress will concentrate in places where the force lines are close together.
You can see how the sharp corners of the crack in the image above lead to
regions of particularly high stress and thus will probably lead to part failure
in these locations.

This isn’t specific to 3d-printing but as it is such an important topic, I want
to include it here anyway. The rules are generally the same for any
manufacturing method: We want to minimize stress and the best way to do it is
this:

R1.4
—
Guide forces on the most direct path possible.

Or in other words, keep the force lines as short and straight as possible.

One example where this matters a lot are sharp corners. They impact part
strength very negatively and are usually easy to avoid by simply adding a
fillet. You can see how a fillet allows a much more direct path for the force
than a sharp corner:

Cross-sectional Considerations

The inhomogeneous nature of 3d-printed parts with their shell and infill has
yet more implications for part strength that are worth talking about. In
traditional design engineering, you are taught to reduce cross-sectional area
of a part as much as possible. You should only keep material in places where
it has the highest effect on increasing part strength. The reason for this is
that reduced volume directly leads to reduced material usage and thus cost and
weight savings.

With 3d-printing, things work differently. You can usually increase the
cross-sectional area quite a bit, without a noticeable increase in material
usage, as the infill is mostly empty space. You should rather strive to reduce
the surface area of a part, because that is where most material is used.

This means that you can get away with much thicker shapes in 3d-printed parts,
which obviously benefits part strength a lot. Don’t try to work against this
artificially — if a design allows for it, go for the fattest shape
possible.

R1.5
—
Use large cross sections. Prefer thick shapes over thin shapes.

To give an example, think about the traditional I-beam
(Wikipedia) cross-section. The idea is that material is only kept in
the regions far away from the neutral axis of the bending moment, where it has
the highest impact on part strength. This allows for significant weight
savings.

While the I-beam profile works wonders for homogeneous materials, it is not a
good idea for 3d-printing. A square cross-section of the same outer dimensions
will have comparable or better strength but usually does not lead to an
increase in material usage and print-time. It might actually perform better in
those areas, as shown below.

Simulation Struggles

In traditional mechanical engineering, whenever a part needs to be evaluated
for its strength, the tool of choice is simulation. Simulation allows making
accurate prediction of a part’s behavior under load without needing to
manufacture it. Especially for parts made from expensive materials or made using
expensive manufacturing methods, performing simulations beforehand is critical
for the project’s budget.

Unfortunately, simulation quickly breaks down when trying to analyze 3d-printed
parts. The problem is, once again, the inhomogeneous nature of 3d-printing.
While stress analysis may still be useful to find critical areas to give
attention to, extracting realistic values for limiting forces is a lost cause.

Luckily, there is an alternative to simulations that 3d-printing offers. The
exceptionally low cost of manufacturing 3d-printed parts means that printing
prototypes for testing mechanical properties is often the most cost-effective
solution. Need to find yield strength of a design? Just print a few copies
and empirically determine it.

However, there is one things to be careful with: While I see test-prints as a
great way to determine mechanical properties, I advise to not rely on them for
determining dimensional accuracy. I will go into more detail on my reasons for
this in the next chapter.

Finally, an implication of the above is that topology optimization
(Wikipedia) is not really well suited for FFF 3d-printing.
While it is amazing for other additive manufacturing processes, current tools
are usually not able to produce designs that would truly be optimal for FFF
3d-printing. Not to speak of the often less-than-ideal printability of such
parts.

2. Manufacturing Tolerance and Part Finish

Next, I want to talk about optimizing your design to improve manufacturing
tolerances. Certain shapes will generally print cleaner than others.
If you take this into consideration during design, you can achieve parts that
fit on first print and won’t need endless tweaking of the slicer settings and
printer. A welcome side-effect of improved tolerance is a cleaner part finish
as well. Thus, the following chapter covers both: Manufacturing tolerance and
part finish.

Chamfers vs. Fillets

To start, let’s talk about edges. Generally, mechanical parts should avoid
sharp edges because they are unpleasant to touch. Any good design breaks edges
by either adding a round fillet or a 45° chamfer of appropriate size.

While the choice between the two is often made for stylistic reasons, there are
actually some rather important printability aspects to it. There are two
situations that we have to look at, distinguished by the print orientation of
the edge:

Adding a fillet or chamfer to edges parallel to the print surface means the
feature will be built up from multiple layers, potentially with overhang.
Fillets are not well suited for this. They start with an incredibly steep
overhang that will not print well. Often, such fillets will have large
surface deviations. Even if no overhang is involved, when the
fillet sits on top, the changing curvature will make the layer steps very
visible, degrading the part finish. Chamfers are much better suited in this
orientation. They have a constant overhang angle, leading to a very consistent
layer stepping, which looks a lot more pleasing.

Contrary, edges vertical to the print surface are much better suited for
fillets. A fillet means that the printer does not have to make sharp corners
while laying the perimeters. This reduces print-head acceleration and thus
reduces surface artifacts (ringing, overshoot). A chamfer on such edges will
have two sharp corners, which will never look as pretty and will never achieve
the same tolerances as a round fillet.

Thus, unless stylistic choices dictate something else, the cleanest prints will
be achieved by following this rule:

R2.1
—
Use chamfers on edges parallel to the print surface. Use fillets on edges
vertical to the print surface.

Horizontal Holes

With horizontal fillets, a major problem is the steep overhang. There are
other design features where similar problems arise. For example, horizontally
oriented circular holes. The larger they get, the bigger the problem will be.
The solution is to deviate from the perfect circle for a more optimized shape.
For smaller holes, a teardrop shape with a 90° angle works well. Larger holes
can be extended to have a flat “roof”. Keep in mind that a slight amount of
additional clearance will be needed beneath the roof, due to the bridges
drooping slightly.

R2.2
—
Improve horizontal holes by using a teardrop shape or giving the hole a flat roof.

Seemingly Seamless

A topic that is very important for both part finish and dimensional
tolerances is the placement of the perimeter seams. Perimeter seams are the
point where a perimeter line starts and ends. It is very hard to tune a
printer such that there won’t be any seam artifacts. Thus, let’s put thought
into the placement of this seam so it does not interfere with any functionality
or aesthetics.

Usually, the seam placement is chosen automatically by the slicer. For each
layer, it searches for the sharpest corner (concave or convex) and then places
the seam right into it. This has proven to be a very reliable way to
produce good results without requiring any explicit user input.

However, this algorithm breaks down in two situations:

1. On perfectly round perimeters, like those of a circular hole or outline.
2. When all corners have the same angle or when the sharpest corner requires
   tight tolerances.

Especially the first case is very relevant. This effect has severe consequences
for the tolerances you can achieve with a circular hole. While the rest of the
perimeter may deviate by less than 0.1 mm, the seam artifacts can easily
bump that up to 0.4 mm. And worse, it shifts the centerline of the hole
because the seam is only in one place.

The solution is simple: Add a corner where the seam can be placed without
interfering. Instead of a perfect circle, shape the hole like a teardrop where
the corner has a 120° angle. Note the difference to the horizontal holes
discussed previously.

R2.3
—
Use a teardrop shape for vertical holes to avoid inaccuracy due to perimeter
seams.

Similarly, if other corners must not be used for seam placement, add a small
notch somewhere that will serve as the “sharpest corner” and thus take the
seam automatically. While you can of course also manually place the seam using
the slicer software, this trick means no slicer tweaking is needed.

To summarize:

R2.4
—
Consider where the seam will be placed. If tolerances are tight, provide a
sharp concave corner for the seam the hide in.

Expectable Tolerances of FFF/FDM

A fundamental building block of any “design for manufacturing” is knowing the
accuracy limits of the manufacturing process. As stated in the beginning, the
goal is to build portable designs, so we should assume conservative limits
instead of exhausting the capabilities of a particular printer. Note that this
is backwards to the traditional workflow: Instead of specifying tolerances
that someone else then has to achieve during manufacturing, you accept
certain tolerances of the manufacturing process and design accordingly.

It also needs to be noted that a base assumption here is that the 3d-printers in
question are well-calibrated. A machine that is out of calibration will have
an additional constant error on top of its tolerances. For example, badly
configured steps/mm values will lead to a constant factor deviation of
measurements along a certain axis. Part design cannot and should not
accommodate for this — there wouldn’t be much left if we tried.

Before we can talk about tolerance values, it is important to understand what
causes these deviations in the first place. A tolerance value means nothing
without the context that it applies to. In 3d-printing especially, tolerances
will vary widely depending on specific part geometry. This topic is quite
far-reaching, as a very large number of factors play into it. As such, this
chapter will only contain the most important things to consider.

Starting from the bottom, the lower bound on accuracy is due to the step
resolution of the printer’s stepper motors and their drivers. Fundamentally,
there is no way to get more accurate than this, as dimensions will always fall
somewhere between two step locations. However, on state-of-the-art
3d-printers, the theoretical step resolution is around 0.01 mm. Whether
this value is actually achievable is debatable, but the takeaway is that step
resolution is well below the deviations caused by other effects.

Continuing on, there are more effects of the printer’s motion system on
dimensional accuracy. They all have to do with mechanical unrigidness of the
printer. Tuning the slicer settings and advanced motion control (input shaping)
can bring large improvements here. However, an equally big role is played by
the geometry the print-head is printing. Sharp corners are the worst offender
as they force the printer to maximum acceleration. By optimizing a part for
“easy motion”, tolerances of printed parts can often be improved a lot.

R2.5
—
Design part geometry for easy motion paths while printing, to improve dimensional accuracy.

The next anti-accuracy component of a 3d-printer is the extruder and hotend.
Uneven extrusion makes the line width vary, which has an impact on the outer
dimensions of a part. Additionally, the nozzle of the printer always slightly
drags the extruded line behind it. This leads to circles always being slightly
undersized. On inner circles (vertical holes), assume deviation to make the
hole smaller. On outer circles, assume deviation to make the outer diameter
smaller, but to a lesser degree.

As a rule of thumb, for current popular FFF 3d-printers with an 0.4 mm
nozzle at 0.2 mm layer height, these effects lead to deviations in the
range of ±0.1 mm for each surface. The deviations are shifted for circular
paths, as mentioned. While many surfaces will be more accurate, this value is
a safe bet for surface deviation.

Unfortunately, while 3d-printers usually do not have a noteworthy growth in
tolerance for larger dimensions, other effects take over here: Especially
warping and shrinkage of the 3d-printed parts as they cool down.

Warping and shrinkage depend heavily on the choice of material, but there are
also part geometries that experience it more than others. In general, the more
voluminous a part is and the less sharp its edges are, the less warping can be
expected. Think about the forces exerted when the material shrinks and how
the geometry can resist those forces.

R2.6
—
Prevent warping by making parts voluminous and their surfaces smooth and rounded. The ideal shape is a sphere.

Perfect Precision

The previous chapter discussed what tolerances can be expected from the FFF
3d-printing process. However, there are of course