Nonconformance and MRB
A Part You Cannot Rework

Why Additive Breaks Conventional MRB Assumptions

A material review board grew up around subtractive and moulded manufacturing, and its default dispositions assume something that is usually false for a printed part: that the part can be brought back into conformance by removing material or by adding it. A machined part that runs oversize can be cut again. A moulded part with a short shot can sometimes be salvaged at the press. A printed part that has already been built is a solid, near-net object created layer by layer, and the options for changing it after the fact are far narrower than the vocabulary of standard MRB practice implies. Approving a disposition framework that was written for steel and applying it to resin or filament is how a print operation ends up "reworking" parts that were never reworkable and shipping them out of conformance.

The consequence is not academic. A disposition decision is a statement about what the customer will receive and what the organisation accepts as a quality cost. Getting the categories wrong inflates the cost of poor quality in one direction, by scrapping parts that could have been used, and deflates quality in the other, by passing out-of-specification parts that will generate a warranty claim. Both failures trace back to a disposition vocabulary that does not fit the process. The regulatory and documentation context around this sits alongside our guide to ISO 13485 regulatory requirements, and the upstream risk analysis that should have predicted the failure is covered by FMEA and control plan for additive.

The NCR Document: Fields That Actually Matter

A nonconformance report for a printed part needs to capture more than a generic form allows, because the root cause runs back through the build file and the machine, not just through a batch number. A form that records only "part number, quantity, defect" cannot support a capability investigation, and the person filling it in will start guessing to fill the blanks. The fields that carry the diagnostic weight are the ones that connect the physical part to its digital origin and its process history.

Two of these fields do more work than the rest. The build file revision is what turns a single nonconformance into a searchable one: if three NCRs in a month point at the same file revision, the problem is the design and not the machine. The detection stage is what sets urgency. A defect caught in-process costs a build cycle. The same defect caught at final inspection costs the whole lot. Caught in the field it costs a customer relationship, and the organisation needs to know which of those it is looking at inside the first hour.

Disposition Categories for Additive: Four, Not Three

Conventional MRB practice uses three primary dispositions: use-as-is, rework or repair, and scrap. Additive needs a fourth, because the cheapest way to satisfy a specification on a printed part is often to build another one, and that option has a different cost structure from reworking the part in hand. Treating "reprint" as a subtype of rework hides the fact that it consumes machine capacity and material rather than labour, which changes who should approve it.

• Use-as-is. The part exceeds a nominal dimension or fails a cosmetic requirement but is functionally acceptable for the specific application. Requires a documented engineering justification tied to the application, not to the general specification. This is the disposition that most often lacks evidence and most often causes a downstream field failure.
• Rework. The part can be brought into conformance by removing material. On a printed part this means drilling, tapping, chasing threads, facing, or trimming support remnants. It cannot mean adding material back, because the layer bonding and dimensional control that would require are not achievable after the build.
• Repair. The part is salvaged by an operation that does not restore nominal dimensions but restores function: installing a threaded insert, bonding a reinforcement, or filling a void with a compatible resin and accepting the resulting surface. Repair almost always requires a concession from the customer, because the part delivered is not the part specified.
• Reprint. The nonconforming quantity is rebuilt from a corrected file. This is the additive analogue of remachining from bar stock and it should be dispositioned with the same seriousness as scrap, because it consumes capacity that the schedule had allocated elsewhere. It is usually the correct disposition when the nonconformance is a structural or build-process failure rather than a dimensional near-miss.
• Scrap. The part is unrecoverable. The only decision left is whether the failure mode demands a CAPA, and whether the affected population extends beyond the parts already in hand.

The Disposition Decision Table

Additive disposition is a function of two variables, and separating them is what makes the decision repeatable. The first is the class of deviation. The second is how critical that class is to the application the part will actually serve. A cosmetic deviation is scrap in a visible consumer product and use-as-is on a drone frame. The same measurement produces opposite dispositions, and that is correct.

A disposition is the documented decision on what happens to a specific quantity of nonconforming material, recorded against an NCR and approved by the material review board before the parts move. It is not the same as a concession, which is the customer's separate agreement to accept a delivered part that does not meet the stated specification.

The row that matters most is the one for anisotropic weld lines, because it is the failure mode an experienced subtractive quality engineer is least likely to weigh correctly. A printed part is weaker along the interface between layers and along a travel move that crossed a wall, and a deviation that looks purely dimensional on the drawing can run parallel to one of those interfaces. Use-as-is is defensible for a dimensional overshoot in a part loaded in-plane and indefensible for the same overshoot in a part loaded across a layer boundary. That is why the decision table is keyed on application criticality rather than on deviation magnitude alone.

Rework on a Printed Part: What Is Actually Possible

The word rework causes more confusion in additive quality than any other. On a machined part, rework can remove material, add material by welding, or reform the geometry by bending. On a printed part, only the removal half of that list survives, and the geometric and material consequences of the removal have to be considered before it is approved.

• Viable operations. Drilling and reaming holes to size, tapping threads, chasing threads that printed undersized, facing a mating surface down to flat, trimming support or brim remnants, and removing flash from a rafter. These are all subtractive and all leave the part's bulk properties unchanged.
• Borderline operations. Sanding or machining down a cosmetic surface, which removes the outer skin and can expose infill or affect wall thickness. Acceptable only when the remaining wall meets the design minimum, which should be checked rather than assumed.
• Not viable. Adding material to build up an undersized feature. Bonding a printed patch over a defect and calling it rework. Heat-forming a printed part back into shape. Each of these produces a part whose properties are not the properties that were validated, and none of them belongs in a disposition described as rework.

When the only route back to conformance is a reprint, the decision stops being free. A reprint consumes machine hours that the production plan allocated to another order, and on a loaded schedule it displaces revenue. That cost belongs in the disposition conversation, which is why the production-control side of this is worth reading alongside our guide to work order and job traveler control.

MRB Quorum and the Cost of a Late Decision

The composition of the material review board for printed parts is where many organisations under-resource the process. A board made up only of quality staff can assess whether a part meets the specification, but it often cannot assess whether it meets the application, which is the question the one genuinely hard dispositions turn on. The three participants that add the most value are the process engineer who understands the build, the quality representative who owns the specification, and the commercial owner who can price the consequence.

The other half of holding cost is the risk of the part re-entering the flow. Nonconforming parts that sit in an unsegregated area next to good parts on a busy floor eventually get shipped, and the failure surfaces at the customer rather than in the plant. Physically quarantining anything awaiting disposition, with the NCR tag attached and the quantity recorded, is the cheapest control in the whole quality system, and it is also the one most often skipped during a busy week.

Closing the Loop: CAPA, Trend and Supplier Feedback

A disposition closes the problem for the parts in hand. It does not prevent the next occurrence, and that is the job of the corrective and preventive action that follows. The link that makes it work is the trend view: a single nonconformance of a given type is an incident, three nonconformances of the same type in a quarter is a systemic problem that the FMEA and control plan failed to catch. Reviewing the NCR log by failure mode and by build file revision on a fixed cadence is what surfaces the systemic cases before the customer does.

When the analysis points at incoming material rather than at the process, the loop has to close back to the supplier. A recurring dimensional drift in filament, or a batch of resin that behaves differently from the last, shows up in the NCR log long before it shows up in a customer complaint, and the incoming checks that catch it are covered in our guide to filament quality evaluation. Where the analysis points at inspection capability rather than at the process, the measurement system itself needs qualifying, which is the subject of SPC, Cpk and gauge R&R on printed parts.

Bottom Line

Additive manufacturing needs a disposition vocabulary that matches what is physically possible on a printed part. Use four categories rather than three, treat reprint as a capacity decision rather than a clerical one, key the disposition on application criticality rather than on deviation magnitude, and require the evidence each category demands before the parts move. The failure mode that catches out the most organisations is not a missed scrap decision, it is a use-as-is approval granted for a dimensional near-miss on a part loaded across a layer boundary. Getting that one row right is what keeps disposition discipline from becoming a way to ship problems downstream. The cost consequence of getting it wrong is quantified in our guide to the cost of poor quality and 8D root cause analysis.

← Back to Blog