Model assisted qualification of NDE techniques

The costly and time consuming nature of empirical trials typically performed for NDE technique qualification is a major barrier to the introduction of NDE techniques into service. The use of computational models has been proposed as a method by which the process of qualification can be accelerated. However, given the number of possible parameters present in an inspection, the number of combinations of parameter values scales to a power law and running simulations at all of these points rapidly becomes infeasible. Given that many NDE inspections result in a single valued scalar quantity, such as a phase or amplitude, using suitable sampling and interpolation methods significantly reduces the number of simulations that have to be performed. This paper presents initial results of applying Latin Hypercube Designs and M ultivariate Adaptive Regression Splines to the inspection of a fastener hole using an oblique ultrasonic shear wave inspection. It is demonstrated that an accurate mapping of the response of the inspection for the variations considered can be achieved by sampling only a small percentage of the parameter space of variations and that the required percentage decreases as the number of parameters and the number of possible sample points increases. It is then shown how the outcome of this process can be used to assess the reliability of the inspection through commonly used metrics such as probability of detection, thereby providing an alternative methodology to the current practice of performing empirical probability of detection trials.