Finished pressboard metal inspection must cover fragments introduced by machining and handling. Challenge the detection route with relevant tool materials and part thicknesses to establish its actual coverage. [1]
Calculation basis. Project-specific values and gates in the illustrative screens below are assumptions; cited catalogue values retain their source test conditions.
First identify plausible contamination sources: cutter wear or breakage, drilling debris, fixture contact and fragments introduced during rework. Distinguish these from board fibers and ordinary nonmetallic dust, because the inspection method may respond differently.
Define what the inspection can find
| Challenge variable | Representative condition | Coverage question |
|---|---|---|
| Particle material | Metals associated with actual tools and fixtures | Does the method respond to each relevant material? |
| Shape | Fragment or sliver as well as a convenient reference sphere | Does shape change the response? |
| Location | Surface, hole, recess and internal depth where applicable | Can the relevant location be examined? |
| Orientation | Positions representative of possible trapped fragments | Is a favourable alignment masking weaker response? |
| Part geometry | Actual thickness and overlapping sections | Does the finished configuration remain within validated coverage? |
A visual check covers accessible surfaces under its specified lighting and magnification. It cannot establish the absence of embedded material. A magnetic response likewise does not establish coverage for every metallic composition. An instrument-based route needs its own demonstrated response for the materials and geometry in the matrix.
Challenge the complete finished-part inspection route
Use controlled challenge specimens that cannot be confused with production parts. Their identification, recovery and disposal should prevent the validation exercise from introducing contamination into a kit. Establish the response criterion before comparing challenge results.
Include relevant difficult locations: a narrow slot, a thick laminated section, a recessed hole or the region inspected last in a manual sequence. A reference object held in front of a sensor tests a much simpler condition than a fragment inside the actual inspection envelope.
Record successful and missed challenges alongside part position and inspection settings. A limited series of successful demonstrations defines the conditions examined; it does not prove that every possible particle is detectable. Where coverage is inadequate, change the route or add a complementary check suited to that location.
Detection should also trigger source investigation. Repeated findings near one machined feature may indicate a particular cutting or fixture problem. Removing the discovered fragment alone leaves the generator of contamination active.
State the inspection route and validated challenge coverage in the acceptance record. Revisit affected cells for changed thickness or tool materials, and support any contractual metal-free claim with its defined verification basis.
Separate observed hits from a demonstrated detection bound
Detection capability also needs known unflawed opportunities to assess false calls [1]. Assume twenty independent, identical binary trials on one fixed particle material, size and location, with eighteen detections. The observed fraction is 90%; an exact binomial confidence calculation retains sample uncertainty [2].
The one-sided 95% lower bound pL solves Σk=18…20 C(20,k)pLᵏ(1−pL)²⁰⁻ᵏ = 0.05. It is 0.7174, or 71.74%. Against an assumed project requirement that this bound exceed 80%, the trial fails despite its 90% observed hit fraction.
Improve and re-demonstrate the stated inspection envelope, retaining known misses and false-call controls. Do not pool different particle/location strata without a justified model. NASA’s aerospace signal-response qualification criteria are not adopted as pressboard acceptance limits.
References
- A. Koshti, P. A. Parker, D. S. Forsyth, M. W. Suits, J. L. Walker and W. H. Prosser, Guidebook for Limited Sample Probability of Detection Demonstration for Signal-Response Nondestructive Evaluation Methods, NASA/TM-20210018515, corrected copy, July 2021, section 1.
- NIST, Exact Binomial Confidence Limits, Dataplot Reference Manual, 2010.