How to qualify a part manufactured by LPBF (and why projects often fail there)

Qualifying a part isn't about getting it right. It's about demonstrating that it will always work well, and being able to show someone the evidence. That distinction is what separates a project that reaches production from one that remains in the demonstration phase with two pretty parts sitting on a table.

And that's where most of them fail, because the effort to qualify is usually discovered late, when the technical part is already solved and everyone took the project for granted.

The five levels

Qualification is not a single thing. It consists of five layers that support one another, and skipping one invalidates the layers above it.

Level	What is demonstrated	What is frozen
1. Material	The powder meets chemistry, particle size, and cleanliness requirements, batch by batch	Supplier, specification, and reuse criteria
2. Process	The parameters yield the expected microstructure and density	Power, speed, thickness, strategy, atmosphere
3. Machine	That specific, calibrated equipment reproduces the result	Calibration, maintenance, and verification plan
4. Part	That geometry, in that orientation, meets the specifications	Orientation, supports, complete post-processing
5. Production	Repeated batch after batch, with evidence for each	Process record, controls, and traceability

The keyword in the right-hand column is freeze. Qualified means that from that point forward, any change to what is frozen requires re-proofing. Changing powder suppliers, moving the part on the platform, or updating the software can invalidate the qualification.

It's uncomfortable, and that's exactly what gives the seal its value.

Why does LPBF cost more than machining?

In a machined part, the material arrives already qualified in bar form, with its certificate. The process removes material from something that already had the properties.

In LPBF, the material and the part are created simultaneously. The microstructure, and therefore the mechanical properties, depend on how it was manufactured: the orientation, the scanning strategy, the layer thickness, and the heat treatment. The same alloy, on the same machine, yields different results depending on how the part was positioned.

From this, three consequences follow that are surprising to someone coming from a machining background:

The properties are anisotropic. The part does not respond the same way in the growth direction as it does in the plane. It must be characterized in both.
Orientation is part of the specification. Changing it changes the product.
Test specimens must be manufactured with the part, in the same assembly, to be representative.

What evidence needs to be kept

This is the list you should have ready from day one, not just when the client asks for it:

From the powder: batch number, certificate of analysis, number of reuse cycles.
From the manufacturing process: laminated file, applied parameters, layer-by-layer process log, oxygen and chamber pressure.
From the machine: last calibration and maintenance.
From the post-processing: thermal cycle with its actual graph, not the nominal one.
From the part: dimensions, density, and tests on specimens of that same manufacturing process.

Assembling this later, for a part that's already been manufactured, is impossible. Data that wasn't recorded doesn't exist.

Why do projects die here?

Four reasons, in order of frequency:

It's discovered too late. The budget and timeframe were set for developing the part, not for qualifying it. When the actual cost of qualification comes to light, the project no longer works.

The wrong level is being requested. A critical aeronautical component protocol is being applied to a support that doesn't need it, because no one dared to decide what level was sufficient.

The machine doesn't allow it. If the equipment doesn't record the process, or doesn't allow exporting the history, or doesn't allow setting the parameters for a custom material, there's no way to freeze anything.

The acceptance criteria are missing. The client and manufacturer never agreed on what "good" means. Without that written agreement before starting, the discussion drags on and never ends.

How would we approach it

Start with the level that's truly necessary. Most industrial parts—tooling, spares, short runs, non-critical components—can be addressed with levels 1, 2, and 5, and can be completed in weeks. Full levels 3 and 4 are for parts where a failure would cause a complete breakdown, and those take months, quite rightly.

Write down the acceptance criteria before manufacturing the first part. It's a half-hour meeting and it's what decides whether the project succeeds or gets bogged down.

And register from day one, even if no one has requested it yet. It's free while it's being manufactured, but impossible to register afterward.

Regarding the machine, there are three questions worth asking before buying it: does it record the process layer by layer, can that record be exported, and can the parameters of a material not listed in the manufacturer's catalog be set and saved? Without these three questions, the machine's performance depends on the supplier.

—

What can be measured and corrected while the part is being manufactured, in closed-loop control. How to measure the density that needs to be demonstrated, in porosity and density. Why orientation is part of what needs to be frozen, in anisotropy and orientation. And why the entire sector is working on common criteria, in distributed manufacturing and on-demand spares.

And if what you need is the qualified part and not the equipment, that's exactly what we do at parts manufacturing.

Why does process gas also enter into what freezes, in argon or nitrogen.