A 100% inspection-rate tradeoff question lands on the NDT (nondestructive testing) desk

Quality Magazine has published a comparison study weighing the time savings of high-throughput digital radiography (DR) against the detection rates of automated defect recognition (ADR) software in aerospace and automotive parts. The piece, titled "High-Throughput DR Aerospace and Automotive Inspection with ADR | And a Time and Detection Rate Comparison Study," is positioned as a practitioner-facing benchmark rather than a vendor announcement.

DR replaces conventional film radiography, the legacy X-ray inspection method that produces images on photographic film, with a flat-panel detector that delivers a digital image in seconds. ADR layers software on top of that image, flagging indications (localized density or geometry anomalies on a radiograph that may signal a defect) for human review rather than relying on a technician to find every flaw by eye.

What does the study actually compare?

The headline frames two metrics that plant quality teams track separately: throughput, measured as the number of parts inspected per shift or hour, and detection rate, the share of relevant defects the system catches. The two metrics typically pull in opposite directions. Faster scanning means fewer dwell angles and shorter integration times, which can shrink the contrast-to-noise ratio (the signal clarity between a flaw and the surrounding material) and let smaller defects slip past the algorithm. A slower scan, with more projections and longer exposure, raises detection probability at the cost of cycle time.

The study addresses both industries in a single article. Aerospace inspection generally tolerates longer cycle times because flight-critical components carry tighter acceptance thresholds and the cost of a missed flaw dwarfs the labor cost of extra dwell time. Automotive inspection operates the opposite way, with high-volume castings and welds priced per second and a different defect-size threshold.

Why does the same DR-plus-ADR chain behave differently in each plant?

Because the part population is different. Aerospace workpieces tend to be titanium or aluminum alloys with thinner walls and tighter geometric tolerances, requiring lower X-ray energies (lower kV, or kilovoltage, the peak voltage driving the X-ray tube) and higher resolution detectors. Automotive components are often steel or iron castings with thicker sections, needing higher energies and accepting coarser resolution because the relevant defect size is larger.

ADR settings follow the part, not the industry. A 0.2-millimeter indication that matters on a turbine blade is noise on a brake caliper. Vendors often sell the same algorithm package for both markets, leaving the plant quality engineer to tune thresholds, region-of-interest masks (areas of the image where the algorithm focuses its search), and reject criteria for each part family.

What does this change for the plant quality manager?

If the study confirms a measurable detection-rate penalty at maximum throughput, the practical question for a manufacturing plant becomes: which is slower across the portfolio, the inspection cell or the downstream machining and assembly stations that feed and follow it. Most plants already run the X-ray cell as a bottleneck, so any cycle-time gain goes straight to output.

What to watch: whether the comparison delivers a defensible operating point — a throughput number paired with a measured probability of detection (POD, the statistical likelihood that the system will flag a defect of a given size) — that a quality engineer can hand to an OEM (original equipment manufacturer) customer auditor without a follow-up conversation. That POD-plus-throughput pair is the deliverable the automotive tier-1 (direct-to-OEM) suppliers will look for first.