The Constraints That Shape Every Printed Part in Additive Manufacturing

Every printed part in additive manufacturing is a negotiation. Not with your slicer, not with your printer: with physics. Here's how to stop fighting those constraints and start designing with them.

There is a question engineers ask when they first encounter additive manufacturing seriously, not casually: "What are the rules?"

Not rules in the abstract sense. Actual rules. The kind that tell you whether your part will come out right or come off the build plate in pieces.

The frustrating honest answer is: it depends. It depends on the process, the material, the geometry, the machine, and the ambient temperature in the room. But the more useful answer is that there is a small set of physical phenomena that drive almost every constraint you will encounter across every additive manufacturing process. If you understand those phenomena, you can reason about new situations instead of just memorising a checklist.

That is what this post is about. Not rules: the physics behind the rules.

Every additive manufacturing process shares one defining characteristic: it builds material incrementally, in a sequence, over time. That seems obvious. But it has consequences most people do not fully trace.

Additive manufacturing constraints are not arbitrary process limitations. They are the direct physical consequences of building a solid object one layer at a time, in a specific direction, against gravity, while thermally and mechanically unstable.

Because material is added sequentially, each layer depends on what came before it. The part is not created instantaneously: it exists in an intermediate, partially-built state for most of its production. That intermediate state is where most failures happen. Understanding the constraints means understanding what makes that intermediate state unstable, and designing to keep it stable long enough for the part to finish.

Nearly every additive manufacturing constraint traces back to one of three physical sources. Think of them as a hierarchy. The further down the hierarchy you go, the more design freedom you recover.

1. Gravity and structural support

A layer of material being deposited cannot float in mid-air. New material needs something beneath it to rest on: either the previous layer or a dedicated support structure. When a feature projects outward without that support, it either sags, collapses, or prints poorly. This is the overhang constraint, and it is the one most designers encounter first.

The practical threshold for most processes: overhangs at angles beyond 45 degrees from the vertical start to require support. Some processes handle 50 or even 55 degrees well. A few (those using powder beds, where unmelted powder surrounds the part as it builds) are almost immune to this constraint. But for extrusion-based processes and direct energy deposition, the 45-degree rule is a good first approximation.

2. Thermal gradients and residual stress

Most additive manufacturing processes involve depositing material in a liquid or semi-liquid state that then cools and solidifies. Cooling is not uniform. The area just deposited is hot; the area underneath it is cooler. This differential causes the material to contract at different rates in different places, and that differential contraction is the source of residual stress.

Residual stress is why parts warp. It is why a long, flat part with a thin cross-section can curl upward from the build plate like a piece of paper left in the sun. It is also why certain geometries crack mid-build: not because the geometry was impossible, but because the accumulated stress exceeded the material's local strength before the part finished.

The constraint this imposes on design is less obvious than the overhang rule, but it is often more consequential. Large flat areas, abrupt cross-section changes, and sharp internal corners all tend to concentrate stress. Designs that manage thermal gradients (through wall thickness transitions, rounded features, and strategic build orientation) fail far less often than those that ignore them.

3. Layer adhesion and anisotropy

When material is deposited in layers, the bond between layers is almost always weaker than the material within a single layer. This is not a flaw in the process: it is a physical consequence of how interlayer bonds form (partial remelting, mechanical interlocking, diffusion bonding, depending on the process). The result is anisotropy: the material behaves differently depending on which direction you load it.

In a typical FDM (Fused Deposition Modelling) part, the Z-direction (the build direction) can be 20 to 50 percent weaker than the X-Y plane. In metal laser powder bed fusion, the difference is smaller but still present. This means orientation is a structural decision, not just a support-minimisation decision. A bracket that needs to resist loads in a specific direction should be oriented so those loads run across layers, not through them.

Let me make this concrete with the most common friction point in additive manufacturing design: support structures.

Support structures are temporary geometry added to the build to prop up overhanging features. They are removed after printing, either mechanically (by breaking them away) or chemically (by dissolving them in a solvent bath, in processes that use soluble support material).

The problem with supports is not that they exist: it is that they create secondary problems. They leave surface marks where they were removed, they add post-processing time and cost, and for internal channels or cavities, they can be impossible to remove entirely. A support inside a closed internal passage is essentially permanent.

Support structures are not free. They add material, print time, and post-processing labour. They also affect surface finish: the surface touching a support is rougher than a free surface. In a production context, every support is a cost that should be challenged during design review.

This is why the most practical additive manufacturing design skill is not knowing how to add supports: it is knowing how to design them away.

Techniques for reducing support dependency include:

Self-supporting geometry. Arches, chamfers, and teardrop holes are all shapes that distribute the overhang across a broader base, staying within the self-supporting angle even as the overall form gets more complex.

Build orientation selection. Rotating the part so that critical overhangs face upward (and are therefore not overhanging relative to the build direction) often eliminates the need for supports entirely. The trade-off is that orientation affects surface quality, build time, and mechanical anisotropy, so orientation is always a compromise between multiple competing factors.

Feature redesign. Sometimes the cleanest solution is to eliminate the overhang by reshaping the feature. A horizontal through-hole that needs support can often become a diamond or teardrop shape with no support required, and equivalent functional performance.

Thermal constraints require a different kind of mitigation. The main tools here are gradual wall thickness transitions (avoiding sudden cross-section changes), generous fillet radii at corners (reducing stress concentration), and controlled print speed in thick sections (giving material more time to equilibrate thermally). Some processes also use heated build chambers or controlled cooling rates to manage residual stress at the process level, but those are process parameters, not design parameters.

Anisotropy management comes down to one question: where are the critical load paths in service, and does the orientation you chose for printability align those paths with the strong direction of the material? A well-designed part answers yes to both. A poorly designed one picks orientation for support minimisation and ignores structural implications entirely.

Here is the thing most introductions to additive manufacturing constraints do not tell you: the constraints are not symmetric across processes.

Powder bed fusion processes (including both metal and polymer variants) use unmelted powder as a support medium throughout the build. This changes the overhang constraint dramatically. Since the powder holds up partially-fused regions, you can print features that would be impossible in extrusion-based systems. Internal channels, complex lattice structures, and organic geometries with no flat horizontal surfaces are all achievable without dedicated support structures.

This is not a minor technical distinction. It is a capability discontinuity. Parts designed for laser powder bed fusion can look fundamentally different from parts designed for FDM: not because the designers were more creative, but because the physics of the process removed a whole category of constraint.

SLA (Stereolithography) sits in an interesting middle position. Like FDM, it requires support structures, because the part builds up through a resin vat and overhanging regions have nothing to rest against. But unlike FDM, the solidification mechanism is photopolymerisation rather than thermal deposition: a UV laser or light source cures liquid resin layer by layer. This means thermal gradients and residual stress are far less severe than in FDM or laser metal processes. The overhang and anisotropy constraints still apply, but the warping and cracking risks are significantly reduced. SLA parts can be remarkably accurate precisely because one of the three constraint sources has been almost entirely removed.

This is why the mental model matters more than the specific rules. The 45-degree overhang rule is real for FDM. It barely applies to selective laser sintering (SLS). The thermal warping risk is real for FDM and metal processes; it is nearly absent in SLA. If you memorise the rules without understanding the physics behind them, you either over-constrain your design in contexts where the constraint does not apply, or you violate it in contexts where it does.

The same logic applies to anisotropy. FDM parts have very pronounced directional weakness. Selective laser melting metal parts are significantly less anisotropic, especially after post-process heat treatment. SLA parts fall somewhere in between: layer adhesion exists, but the photopolymer bonds more uniformly than a fused thermoplastic. Continuous fibre composite additive processes, on the other hand, can be engineered to be strongly anisotropic in specific directions, which is a feature, not a defect.

Think about carpentry vs masonry.

A carpenter working in wood can make cuts in any direction, plane surfaces smooth, and join pieces at arbitrary angles. The grain of the wood matters for strength but the tool does not know or care which direction it is moving.

A stonemason working in dry-stone walling is constrained by gravity in every single course. Each stone must rest stably on what came before it. You cannot cantilever a stone three feet out and expect it to stay there without support. The entire design vocabulary of a dry-stone wall (the battered face, the through-stones, the hearting) exists because every element is negotiating with gravity, one layer at a time.

Additive manufacturing is closer to masonry than to carpentry. The process has a direction, gravity exists during construction, and stability at every intermediate stage is part of the design problem: not an afterthought you solve by adding more supports.

• Nearly every additive manufacturing constraint traces to one of three sources: gravity and structural support, thermal gradients and residual stress, or layer adhesion and anisotropy.
• The 45-degree overhang rule is real, but process-dependent. Powder bed processes largely eliminate it. Knowing why the rule exists helps you know when to apply it.
• Support structures are a cost, not a default. The best designers in additive manufacturing design them out through geometry, orientation selection, and feature redesign.
• SLA removes thermal gradient risk almost entirely, which shows how eliminating even one constraint source changes what is possible and what is safe to attempt.
• Build orientation is simultaneously a structural, surface quality, and process decision. Optimising for only one of those dimensions usually creates problems in the others.
• Anisotropy is not uniformly bad. Processes that allow engineered directionality can use it as a design variable.

Next post picks up where this one leaves off: once you understand the constraints, you are ready to talk about the explicit trade-offs between strength, speed, and cost, and what happens when you try to optimise all three at once.

Curious to exchange some ideas? Reach out via the contact form or connect on Linkedin!