A continuing project at the Center for NDE in Ames involves the development of models which predict the probability of detecting flaws using a given inspection system.[1] Our general approach as it applies to through-transmission immersion inspections is as follows. With the two transducers to be used in the inspection, a reference experiment is performed to determine relevant information concerning equipment characteristics and transducer efficiencies. This may be done by placing a calibration specimen into the ultrasonic beam and measuring the time-domain electrical signal in the output cable of the receiver. Using models, we then predict how this received electrical signal would be changed if the calibration specimen were removed and unflawed and flawed components were placed into the beam in turn. The two components are assumed to have identical geometries except for a hypothetical flaw of given type, size, and orientation. If the difference between the predicted output signals for the two components is sufficiently large, compared to system noise, then the hypothetical flaw is said to be detectable.