Assessment of flaw detectability using transfer function

摘要

In many situations real flaws such as tight cracks with known morphology cannot be manufactured in part configuration specimens. Typically, fatigue cracks are manufactured in simple geometry specimens such as flat plates, dog-bone shaped flat or round specimens. If an NDE technique is required to provide a reliably detectable a_(90/95) flaw size, then the direct method for qualifying the NDE procedure is to use the appropriate real flaw specimens, run inspection procedure, and perform probability of detection analysis. This is can be described as direct POD demonstration testing which follows guidelines of MIL-HDBK-1823. The POD testing is operator specific. When many operators undertake the same test then conclusions can be drawn regarding percentile of operators meeting detection of certain flaw size with minimum 90%/95% POD/Conf. This paper takes a case where, real flaws are not available in part configuration specimens and a direct POD demonstration study cannot be undertaken because of lack of real flaws. In such situation, general practice is to use artificial flaws in part configuration specimens. Just artificial flaws are not sufficient for this analysis. Additional flaws including some real flaws in same and simpler geometry specimen are required. Signal response data is taken on all sets of artificial and real flaws. Noise is measured on each specimen and signal response variations are measured. The paper provides a procedure to calculate the flaw delectability size using the transfer function analysis. The analysis assumes that the chosen transfer function that relates artificial and real flaw signal responses in real parts and/or simple geometry specimens to compute a_(90/ 95) flaw size for real flaws in real parts. Both single-hit and multi-hit flaw detection is addressed.