The probability that an inspection detects a specified defect under stated conditions.
Also called: detection probability; detection rate.
Defining the defect and detection conditions
Name the target and the conditions in which it is to be detected. For bottle inspection that might be a specified count defect; for another detector it may be a different feature. A detection probability without the reference condition leaves the event undefined.
Relevant conditions depend on the inspection principle. Metal-detection results, for example, depend on target shape, position, material and product effects; they cannot supply a detection probability for counting tablets or identifying a short bottle.
The detected fraction is an observation from a defined sample group. It can estimate conditional detection performance; it is not a guarantee for every future presentation. Keep the reference-positive group as the denominator when estimating sensitivity.
Compare detected fractions only when the target definition, reference labels and test conditions match. The miss rate is the complementary fraction for the same known-defective group.
Does detecting a defect mean the bottle was rejected?
Detection is the inspection response. Physical removal is a later function. A POD statement about the detector therefore needs separate removal evidence before it becomes a claim about defective bottles leaving the line.
An overall escape-risk estimate needs both detection and removal performance under stated conditions. A model developed for another inspection task cannot supply those inputs for bottle counting.