A Clear CT Image Is Not Proof of Zero Defects in a 3D-Printed Part

An X-ray computed tomography scan can reveal internal features of an additively manufactured part without cutting it open. A clean-looking image, however, is not a universal proof that no flaw exists. Buyers and engineers should ask what the inspection was intended to detect, how the method was evaluated and what limitations apply to the particular part and scan.

Define the inspection question before the scan

Start with the part, material, geometry and the type of feature relevant to its intended use. A request to “scan the part” does not define whether the goal is dimensional measurement, investigation of a suspected anomaly or a specified nondestructive inspection. Agree on the applicable acceptance requirements and the qualifications of the inspection provider before interpreting a result.

Keep the part identifier, design revision, manufacturing and post-processing state associated with the report. An image may show a region of interest, while the decision concerns the entire part or only specified features. The report should say which areas were inspected and whether any were outside the agreed scope. If the information is missing, ask for clarification rather than assuming complete coverage.

Detection is a performance question

NIST researchers describe a method for assessing probability of detection using additive manufacturing defects and a reference measurement. Their work illustrates that detection capability can be studied with defined samples, an analysis method and uncertainty. It does not give a universal guarantee for all scanners, materials, flaw types or production parts.

A claim such as “the scan detects every defect” needs a defined defect population, inspection conditions and supporting evidence. The size, shape, orientation and location of a feature can matter. A detection study's result should be quoted with its scope and assumptions, not converted into a promise about an untested part. The absence of a reported indication is a result of a particular inspection, not proof of an absolute absence of imperfections.

Image settings and analysis affect the result

Acquisition settings, reconstruction, image processing and the rule used to classify a feature can all influence the reported outcome. NIST's study on acquisition parameters, image quality and detection probability examines these relationships for its defined research conditions. Ask an inspection provider which settings and analysis procedure were used and how changes are controlled.

A voxel is an element of a reconstructed volume, but a nominal voxel size alone is not a guarantee that every feature of that dimension can be reliably detected or measured. Request evidence appropriate to the feature and inspection setup rather than treating a single resolution number as the complete performance specification. The inspector should explain the method's limitations in terms the part owner can use.

Separate a reference sample from a production decision

A reference artifact or test sample can help evaluate an inspection method, but its geometry and material may differ from the production part. Ask how representative the reference is, which feature characteristics were evaluated and whether the same inspection and analysis conditions were used. If the production geometry or material changes substantially, the organization responsible for the inspection should review whether prior evidence still applies.

Consider a hypothetical trial using a sample with known internal features. The provider demonstrates that its process finds some features under a documented setup. That evidence supports discussion of the tested conditions; it does not automatically establish the detection performance for a different part with another wall thickness or flaw orientation. A qualified team needs to decide what additional evidence is required.

When comparing inspection proposals, ask whether the reported performance concerns detecting a feature, measuring its dimensions or classifying its type. These are related but different claims. One method may identify an indication without measuring its volume accurately enough for a particular decision. The buyer should provide the actual decision requirement and let the qualified provider explain the method's demonstrated capability, assumptions and any need for complementary inspection. Keep that explanation with the acceptance plan rather than relying on a marketing image.

Keep the report usable for the next decision

Request the part identity, inspected area, procedure version, relevant acquisition and analysis information, acceptance criteria, findings and stated limitations. Record who approved any deviation and how the data can be retrieved. If an image is reprocessed, distinguish the original data from the later interpretation and retain the version used for the decision. This allows a later reviewer to understand what was actually inspected.