Engine Titanium Consortium - Center for Nondestructive Evaluation ...

will be used to validate the physical models for flaw response. These, as well as the noise
distribution, will form the basis for the formulation of a physical model for the signal distribution. In
the formulation of this model, attention will be paid to providing “hooks” which will allow a crack
morphology database to be introduced at a later time (the development of such a database is not
priced in this proposal). From the statistical models of noise and signal, POD, PFA and ROC
curves will be determined.
Based on the flat-plate demonstration, a first generation methodology will be specified. This will
include procedures for each of the steps mentioned above. The methodology will be validated by
exercising it on a set of existent samples and comparing the results to those of existent
methodologies such as a-hat versus a. The conditions under which one or the other breaks down
and the degree of agreement between their predictions will be noted and interpreted. Particular
attention will be paid to the sparseness of data that can be accommodated by the new
methodology, and the time that would be required for its implementation in the field. The new
methodology will be used to make predictions of POD, as influenced by a variety of parameters, for
use by the lifting community.
Application to a Complex Geometry: Blade Slots: A demonstration will then be conducted on a
blade slot, using a differential reflection probe. The same steps will be followed as in the flat plate
demonstration, modified by insights gained in the formulation of the first generation methodology,
and considering the edge geometry signals as sources of noise in the analysis of the data. Poorly
managed edge responses can easily mask defect responses. One must assume that the
inspection procedure incorporates appropriate mechanisms (e.g., by correct choices of probes and
scans, and/or by signal processing) so that the edge contamination is suppressed. This is the
reason for the consideration of the differential reflection probe, which is frequently used to suppress
edge responses in the field. The edge signal remaining is effectively a source of noise. Our new
POD/PFA methodology will be designed to quantify the edge-suppression capabilities as well as
the other sources of noise. Once the signal and noise distributions are determined, following
procedures already discussed, predictions will be made of the POD, PFA and ROC.
Based on the experiences gained in the blade slot demonstration, a second generation
methodology will be formulated. The second generation methodology will be validated by
exercising it on an independent set of blade slot samples and comparing the results to those of
existent methodologies such as a-hat versus a. The conditions under which one or the other
breaks down and the degree of agreement between their predictions will be noted and interpreted.
Particular attention will be paid to the sparseness of data that can be accommodated by the new
methodology, and the time that would be required for its implementation in the field. After
validation, the new methodology will be used to make predictions of POD, as influenced by a
variety of parameters, for use by the lifing community.
Improvement/Transfer of the POD/PFA Methodology Software to the OEMs: In order for the new
methodology to have full impact on the damage tolerant design and management of rotating
components of aircraft engines, it is essential that the necessary tools, i.e., software for
implementing the methodology, the associated modules for predicting flaw and noise response as
influenced by geometry, and materials characterization procedures to efficiently provide input to
them, be in a form that can be readily used by the OEMs. These tools will be developed,
Quarterly Report – January 1, 2002 –March 31, 2002
print date/time: 6/6/2002 - 8:39 AM – Page 109