How 3D Printing to Industrial Mass Production Actually Works

• 1
  What Changed When Laser Deposition Reached the Shop Floor
• 2
  Where Each Metal AM Process Stops Being Economic
• 3
  The Cost Crossover: When Printing Beats Milling
• 4
  Post-Processing Decides the Final Tolerance
• 5
  How to Read a Quote Before You Commit
• 6
  Process Selection by Part Characteristic

Metal additive manufacturing stopped being a prototyping tool the moment deposition rates climbed and powder handling became closed-loop. The two production-grade families are laser metal deposition (LMD, also sold as DED) and laser powder bed fusion (LPBF, also called SLM). Both melt metal with a fiber laser, but they build in opposite directions.

LMD feeds powder or wire through a nozzle into a moving melt pool. The head sits on a gantry or robot arm, so the part grows in open space with no powder bed around it. That freedom lets you add material to an existing forging or casting instead of starting from a blank. Deposition rates run roughly 10 to 100 times higher than powder bed, which is why the process is quoted for near-net shapes rather than fine features.

LPBF spreads a 20-60 μm layer of metal powder, melts a cross-section with a galvo-scanned laser, then recoats. Layer thickness typically lands between 30 and 60 μm, and the machine builds the whole part inside a sealed chamber under argon. This is the process behind thin walls, internal channels, and lattice structures that no cutter can reach.

The engineering consequence is simple. LMD gives you throughput and the ability to grow on existing metal. LPBF gives you geometry fidelity. Neither one is a drop-in replacement for a machining center, and treating it that way is where most cost estimates go wrong.

LMD earns its place when the part is large, the shape is near-net, and machining the whole thing from bar would remove 60-80% of the stock as chips. A 400 mm titanium bracket that starts as a 30 kg forging and finishes at 4 kg is a classic candidate. You deposit the ribs and bosses, then finish the interface surfaces on a 5-axis mill to ±0.005 mm.

The process loses when the part needs thin internal cooling channels under 1 mm, sharp internal corners, or a surface finish better than roughly Ra 6 μm as-deposited. The melt pool is 1-3 mm wide, so feature resolution follows the nozzle, not the laser spot. Anything smaller than that has to be machined afterwards or moved to powder bed.

LPBF wins on complexity, not on size. A part with conformal cooling channels, an impeller, or a topology-optimized bracket that would need five setups on a mill is exactly the target. Build envelopes in the common 250-400 mm class cover most of these parts, and the support structures come off in the same post-processing cell.

LPBF loses on cost per kilogram and on surface quality. Powder is expensive and partly non-reusable, the build runs for hours to days, and every downward-facing surface needs support that must be cut and ground away. As-built surfaces sit around Ra 10-20 μm, so any sealing face, bearing bore, or mating flange still goes to a CNC for finishing.

The crossover is not a fixed part count. It moves with three variables: buy-to-fly ratio, material price, and feature count. Buy-to-fly is the weight of stock purchased divided by the weight of the finished part. A milled part from plate usually sits between 3:1 and 10:1. Additive near-net shapes bring that down to roughly 1.5:1 to 2:1.

Take titanium at high cost per kilogram. If milling removes 85% of a 20 kg block, most of what you paid for goes into the chip bin. Printing a near-net preform and finishing it on a 5-axis machine can cut material spend sharply even before you count machining hours. That is the real driver, not the printer itself.

On the other side, aluminium brackets with simple geometry and generous tolerances almost never justify additive. A 3-axis machine cuts them in minutes from cheap plate, and the finish is already ready to use. Additive only pulls ahead when the geometry would otherwise force multiple setups, custom fixturing, or a part split into welded subassemblies.

Volume matters less than people expect. Because there is no tooling cost, the first printed part and the thousandth cost roughly the same per unit. That is the opposite of die casting and injection molding, where the crossover against machining sits in the low thousands of parts. Additive's advantage is in low-to-mid volume with high complexity.

No additive process holds ±0.005 mm as-built. Thermal shrinkage, residual stress, and layer stair-stepping all push as-built deviation into the 0.1-0.5 mm range depending on part size and material. Every production print therefore ships with a machining plan attached to it.

The standard sequence is stress relief, support removal, datum establishment, then CNC finishing. Stress relief matters more than most shops admit. A titanium or Inconel part that goes straight to the mill after cutting off the build plate can move several tenths of a millimeter as internal stress relaxes. Anneal first, then cut datums.

Datums are the hard part. An as-printed surface is not flat enough to locate on, so you need either a fixture that grips the near-net shape or a sacrificial pad machined first. Once you have three reliable datums, a 5-axis machine can bring bores, sealing faces, and threads to ±0.005 mm and Ra 0.8-1.6 μm.

For critical interfaces, plan a machining allowance of 0.3-0.8 mm on any surface that will be cut. Less than that and you risk hitting porosity or an unmelted region. More than that and you are paying to remove metal you just paid to deposit. Hot isostatic pressing closes internal porosity when fatigue life matters, but it adds a process step and a lead-time cost.

Ask what the as-built tolerance is on the specific feature you care about, not on the part as a whole. Suppliers often quote a general number that hides the fact that one bore needs a reaming operation on top of the print. Get the machining allowance stated in writing per surface.

Ask how supports are removed and whether the removal is included. On LPBF parts, support cutting and grinding can be a meaningful share of the total. On LMD parts, the question is whether the near-net shape is close enough that your finishing pass has stable stock.

Ask about powder provenance and reuse. In LPBF, powder that has been through many build cycles shifts its particle size distribution, and that changes melt behavior. For aerospace or medical work, traceability from powder lot to finished part is the normal expectation, not an extra.

Finally, check whether your supplier can finish the part in-house. A print shop without machining capacity hands your part to a third party, and the datum you established on the build plate gets lost in the handoff. Keeping printing and 5-axis finishing under one roof removes that risk entirely.