Weld-Joint Radiographic Interpretability in Student-Built Engineering Education Inspection Cells: A Seeded-Defect Benchmark Linking Exposure Geometry to Discontinuity Misclassification Rates

Student-built radiographic inspection cells are increasingly used in engineering education to train weld-joint interpretation, yet the exposure geometry that students select is rarely treated as a controlled variable in the assessment of their interpretive accuracy. The benchmark was evaluated across 2000 replications per design cell, with cluster counts of 8, 16, 24 and 32 and effect sizes of 3, 5 and 8 percentage points. The intraclass correlation was 0.64, the between-cluster standard deviation was 6.0 and the residual standard deviation was 4.5. Power was estimated for the before-after-control-impact interaction, and the minimum detectable effect at 80% power was derived for each cluster count. The results indicate that 52 clusters are required to detect a 5 percentage point effect at 80% power under the declared variance structure. The benchmark provides a reproducible basis for distinguishing interpretive error that arises from exposure geometry from error that arises from trainee judgement, and it supports the design of inspection-cell exercises that treat geometry as a documented parameter rather than an uncontrolled source of variation.