Metal 3D Printing for Production Parts: Mechanical Properties You Can Actually Rely On

Every few months for the past several years, someone at whatever manufacturer I’m stationed at asks whether we can just “print” a production part instead of machining or casting it. My answer has evolved a lot since my first exposure to metal additive manufacturing over a decade ago, when the honest answer was almost always no. Today, after specifying laser powder bed fusion parts for several real production applications, the answer is a careful, qualified yes — but only if you understand which mechanical properties you can trust and which ones will bite you if you assume they behave like wrought or cast metal. This article is about the practical property differences I’ve learned to respect when a metal-printed part is actually going into a machine, not a display case.

Anisotropy Is the Property Nobody Warns You About Enough

The single biggest surprise for engineers new to metal AM is that mechanical properties are directionally dependent. A part built with its long axis vertical (in the Z direction, parallel to the build platform’s travel) will often show different yield strength, elongation, and especially fatigue life than the same part built lying on its side. This comes from the layer-by-layer melting process itself — each layer is a discrete weld pool, and the boundaries between layers can act as preferential crack paths, particularly for fatigue loading perpendicular to the layers.

On a bracket I designed for a test fixture that later got scaled toward low-volume production, I initially oriented the part to minimize support structures, which put the highest-stress fillet parallel to the layer lines rather than across them. Static testing looked fine. It wasn’t until we ran cyclic loading that we saw premature cracking initiating right along a layer boundary at that fillet. Reorienting the build by 45 degrees, at the cost of more support material and post-processing, fixed it. I now treat build orientation as a structural design decision, not a manufacturing convenience decision, and I specify it on the drawing rather than leaving it to the print shop’s discretion.

Porosity and Defects: What Process Parameters Actually Control

Like powder metallurgy parts, metal AM parts are never perfectly dense unless something has gone very right in the process control. Lack-of-fusion porosity, caused by insufficient energy input between adjacent scan tracks or layers, creates irregular, sharp-edged voids that are far more damaging to fatigue life than the small, rounded gas porosity that comes from trapped shielding gas in the powder itself. The difference matters enormously — a part with 99.5% density full of rounded gas pores can outperform a part with 99.8% density that has a handful of lack-of-fusion defects in the wrong location.

This is why I stopped accepting a single density percentage as adequate quality evidence. I ask suppliers for their process parameter validation data, and for anything fatigue-critical, I request CT scanning on a sample population rather than relying on density-by-Archimedes-method alone, since that method can’t tell you where the pores are, only how many there are on average. For one hydraulic manifold-style part, we found through CT scanning that porosity clustered near a specific internal channel intersection where the laser scan strategy had to change direction abruptly. The supplier adjusted their scan pattern for that geometry specifically, and the defect cluster disappeared on the next build.

Surface Finish Drives Fatigue Life More Than the Bulk Material Does

As-built metal AM surfaces are rough, typically in the 6-15 micrometer Ra range depending on the process and build angle, and downward-facing (support-touching) surfaces are usually worse than upward-facing ones. For static strength this barely matters. For fatigue, it can dominate the result entirely — I’ve seen fatigue life differences of 3-5x between an as-built surface and the same geometry with the surface machined or polished smooth, well beyond what the base material’s S-N curve alone would predict.

I learned to treat any highly loaded, cyclically stressed feature on a metal-printed part as needing post-machining or at minimum shot peening, almost as a default assumption rather than an exception. On a rotating component I worked on, we left a fillet in the as-built condition to save a machining setup, reasoning that the static stress there was low. It wasn’t the static stress that mattered; the surface roughness created enough local stress concentration under vibration that we saw early crack initiation in accelerated life testing. Adding a simple post-machining pass on that one fillet solved it completely, and it’s now a standing rule on my drawings: any surface subject to meaningful cyclic load gets called out for machining or peening, full stop.

Comparing AM Properties to Wrought and Cast Baselines

This table reflects a rough pattern I’ve observed across several projects with titanium and nickel alloys rather than a universal rule — actual numbers vary a lot by alloy, machine, and post-processing. What surprised me most when I first built something like this table was that yield strength in metal AM parts is often comparable to or even higher than wrought material, because the rapid solidification creates a very fine grain structure. Fatigue strength is where the picture flips, and it’s exactly the property that’s easiest to overlook if you’re only checking a tensile certificate.

Hot Isostatic Pressing: The Step That Changes the Conversation

Hot isostatic pressing, or HIP, applies high temperature and high pressure simultaneously to close internal porosity, and for anything I’d consider putting into real service with meaningful fatigue loading, I now treat HIP as close to mandatory rather than optional. It doesn’t fix lack-of-fusion defects that are open to the surface, and it doesn’t fix a bad surface finish, but it dramatically improves the internal porosity picture and brings fatigue scatter down to something much more predictable.

I ran a small internal comparison once, testing coupons from the same build plate with and without HIP. The as-built coupons showed fatigue life scatter across almost two orders of magnitude, which is unusable for any serious design allowable. The HIP’d coupons tightened that scatter substantially and brought the average life up meaningfully. That data point alone changed how I write specifications now — for any part where I’m claiming a specific fatigue allowable, I specify HIP and a defined post-HIP surface treatment together, because neither one alone gave me confidence.

Design Rules I Apply Differently Than for Machining or Casting

Metal AM lets you build internal channels, lattice structures, and consolidated assemblies that would be impossible any other way, and that freedom is real. But I’ve learned to apply some conservative rules regardless of that freedom. Minimum wall thickness around 0.4-0.8 mm depending on the alloy and machine, self-supporting overhang angles typically above 45 degrees from horizontal to avoid heavy support structures (and the rough surfaces those supports leave behind when removed), and generous fillets everywhere a stress-critical feature exists, since sharp internal corners compound both the anisotropy issue and the surface finish issue at once.

I also design in sacrificial machining stock on any surface that needs to meet a real tolerance or surface finish requirement, typically 0.5-1 mm depending on part size, because as-built dimensional accuracy on features spanning more than roughly 50 mm can drift by several tenths of a millimeter due to thermal distortion during the build. Trying to hold a tight tolerance directly out of the printer, without planning for finish machining, is a mistake I made once on a mating flange and won’t make again.

When Metal AM Is Actually the Right Call

After all these caveats, I don’t want to leave the impression that metal AM isn’t worth using — it absolutely is, in the right circumstances. It earns its place when part consolidation eliminates several separate components and their joints (which are often the actual weak points in an assembly anyway), when internal geometry like conformal cooling channels delivers a real functional benefit that no other process can achieve, or when production volumes are low enough that tooling for casting or forging can’t be justified.

Where I steer people away from it is when someone is chasing metal AM purely because it sounds advanced, for a part that a simpler, better-understood process would handle with less risk and lower cost. My rule of thumb after years of this: if the part’s function depends heavily on fatigue life in an as-built or lightly post-processed condition, budget for HIP, surface finishing, and real test data before you commit the design, not after the first field failure shows up.