FSE 3D Printing Lab: How Metal AM Parts Actually Get Made

• 1
  What happens inside an FSE 3D printing lab
• 2
  Geometry that additive handles and machining cannot
• 3
  Why the lab and the machine shop sit together
• 4
  Design rules that decide whether the print is worth it
• 5
  When to print, when to machine, when to do both
• 6
  Additive vs CNC vs hybrid for the same part

An FSE 3D printing lab is not a room full of filament spools. The metal side runs powder bed fusion: a 20–60 μm layer of gas-atomized powder is spread across a plate, a laser or electron beam traces the slice geometry, and the melt pool fuses powder into solid metal. The plate drops by one layer thickness. Then it happens again. A 40 mm tall bracket at 40 μm layers takes roughly 1,000 passes.

The melt pool is the whole story. It is 100–200 μm wide and lives for a few hundred microseconds. Cool-down runs at 10⁴ to 10⁶ K/s. That speed is why the process can build a conformal cooling channel no drill can reach, and also why the part walks off the plate with residual stress locked into it.

Layer thickness sets three things at once: build time, surface finish, and how much post-machining you owe. Thin layers cost time. Thick layers cost stock allowance. A 60 μm layer on a vertical wall leaves a stair-step of about 60 μm, which is roughly Ra 8–12 μm as-built. That is not a sealing face.

Most parts leave the machine welded to a build plate by support structures. Those supports carry heat out of the part and hold overhangs from curling. Removing them is a manual or CNC job, and the witness marks they leave usually define which faces need finishing.

The honest case for metal AM is internal geometry. A hydraulic manifold with curved channels, a mold insert with conformal cooling that follows the cavity, a heat sink with lattice fins at 0.4 mm wall thickness. None of these can be cut from solid without splitting the part and brazing it back together. Additive builds them in one piece.

The second case is low volume in a hard material. Titanium TA2 and TC4 (Ti-6Al-4V) machine slowly and eat tooling. Inconel 718 is worse. If you need 20 parts, printing near-net and then machining only the critical faces often beats cutting the whole shape from bar stock.

Now the limits. A flat 200 × 200 mm face will not come off a powder bed flat. It comes off with residual stress bow, typically 0.2–0.5 mm across that span, and it needs a stress-relief anneal followed by face milling. If your part is mostly flat plates and drilled holes, printing it is an expensive way to make a part you could have milled in two setups.

Porosity is the other boundary. Laser powder bed fusion reaches 99.5%+ density when parameters are dialed in, but keyhole-mode defects and lack-of-fusion voids still show up at overhangs and thick sections. For a static bracket that is fine. For a fatigue-critical part, you need HIP and CT inspection, and both add cost.

An as-built metal part is a blank, not a finished component. Support removal comes first, then stress relief, then the build plate is cut off with wire EDM or a band saw. Only after that can you touch the part with a cutter, because machining a stressed part just moves the distortion around.

Heat treatment is where the mechanical properties land. Annealing relieves the internal stress. Hot isostatic pressing (HIP) closes internal voids and homogenizes the microstructure. For Ti-6Al-4V, HIP at roughly 900–955 °C and 100–150 MPa for two hours is the usual route before you compare it to wrought properties.

Then the interfaces get machined. Threads, bearing bores, seal grooves and dowel holes are cut to tolerance on a 5-axis mill or a mill-turn center. That is where we hold ±0.005 mm (±0.0002 in) and Ra 0.8–1.6 μm on the critical features, while the as-built lattice stays as-built.

Surface finishing closes the loop. Bead blasting knocks down the stair-step on non-critical faces. Polishing and electropolishing reach Ra 0.2–0.8 μm on fluid paths. Anodizing, electroless nickel and laser marking handle the rest. Doing all of this under one roof is what keeps a printed part from stalling between five vendors.

Overhang angle is the first number to check. Below 45° from horizontal, the melt pool has nothing under it and needs support. Supports are removable, but they consume laser time and leave a rough surface. Redesign the face to 45° or steeper and the part gets cheaper.

Minimum feature size is the second. Walls thinner than 0.4 mm may not survive the recoater. Holes under Ø1 mm tend to close up or come out oval, so print them at Ø0.3 mm oversize or drill them afterward. Internal channels below Ø2 mm are risky for the same reason.

Trapped powder is the third. Any enclosed void with no escape path holds powder you cannot remove, and that is a contamination problem in a clean assembly. Design a Ø2–3 mm drain path to an accessible face.

Finally, plan the machining stock before you print. Add 0.3–0.5 mm on faces that need a finish cut and 0.5 mm on datum faces. Printing to net shape and hoping for a clean surface is how a part ends up undersized, scrapped, and reprinted at your cost.

Start with quantity. One to twenty parts in a difficult alloy with internal geometry: print, then machine the interfaces. A few hundred parts: look at die casting or vacuum casting for the body and CNC for the critical features. Thousands of simple parts: CNC from bar or plate will beat additive on unit cost every time.

Then look at the shape. If the geometry can be reached by a Ø6 mm end mill from three or four directions, milling is faster and the surface is better. If it cannot, additive earns its place.

Then look at the material. Aluminum 6061 and 7075 are cheap and fast to cut. The same alloys in powder form cost more per kilogram and print at a slower rate. Titanium and nickel alloys flip that math, which is why aerospace and medical work dominates metal AM.