Private beta

Show what your inspection can find, with numbers that hold up

Ask for an invitation to the beta Try the point demonstration calculator Create a free account

What a study produces: a POD curve, its lower confidence curve, and the size found nine times in ten with the bound above it. An illustration, not measured data.

What makes it different

Correctness you can check

Read the validation reports

Guardrails for inspectors, not statisticians

More than one variable

In the browser, saved and reviewable

Plan, report, prove

The analyses

Hit/miss POD with logit, probit, log-log and complementary log-log links
â vs. a POD, with recording threshold and saturation handled by censored maximum likelihood
Model comparison by AIC and BIC on a comparable basis, next to the diagnostics
Binned POD with exact binomial bounds
Point demonstrations of the 29 out of 29 kind, and the trials a demonstration needs
Noise distributions and the probability of false call, and false-call rates
Decision threshold studies: a90/95 and false calls against the threshold
Sizing accuracy: bias, scatter and limits of agreement, with Deming regression when the actual sizes carry error too
Sizing calibration: corrections from measured to actual size, compared out of sample
Trial planning: the a90/95 a planned trial would give, simulated from an assumed curve before the specimens are made
POD against a second size dimension and against flaw type, side, zone, operator or run, each term tested against the model without it
Operator variability: a curve for each operator or run, and one for the study whose bound is clustered on flaw
POD with a floor and a ceiling, for inspections that miss some flaws however large they are
Blind trials: what the operators reported, matched to the flaws by position under your rule

What it does not do

NDT Infer reports statistics against the criteria you enter. It does not certify, qualify or approve a technique, a procedure or an operator. That decision stays with you and your qualification body.