transient (pulsed) eddy current inspection of cf188 inner ... · • simulated wing spar –...
TRANSCRIPT
Transient (Pulsed) Eddy Current Inspection of
CF188 Inner Wing Spar Through Carbon/Epoxy
Skin with Ferrous Fasteners Present
Capt. Peter F. Horan
T.W. Krause, P.R. Underhill, V. Babbar
Royal Military College of Canada
Kingston, ON
Outline
• Introduction
• Experimental Setup
– Principal Components Analysis
• Results
• Conclusion
Introduction• RCAF CF188 Fleet has a need to detect cracks in Inner Wing spar over
carbon-epoxy skin without ferrous fastener or skin removal
• Transient Eddy Current (TEC) inspection seen as possible solution
Experimental Setup
• Central Driver Differential Pickup probe design
Source: P. Whalen, TEC Inspection in the presence of
ferrous fasteners in multi-layered aluminum structures,
RMC Thesis, 2010
Experimental Setup
• Simulated Wing Spar
– 7075-T6 aluminum spar
– Nylon 6 polymer “skin”
– Ferrous Hi-Lok fasteners
– Notches used to simulate cracks
• Only full-through notches between
fasteners examined, i.e.
12 June 2012
Fastener with
no notch
Fastener
with notch
Through notch
in aluminum
Experimental Technique
• Measurement technique used meant to simulate on-aircraft inspection method for
hundreds of fasteners
– Removing and replacing probe on spar, aligning by eye
– Aligning 180-degree coils in-line with fasteners (determined to be optimum detection
position)
Experimental Methodology
• Signal response difference due to technique:
TEC responses of 21 measurements of fastener #16 (blank) (right) with controlled measurement technique, and (right) with realistic technique
#16 (Blank)
Experimental Methodology
• Analysis of raw TEC, time-domain responses incapable of
accurate crack differentiation:
TEC responses of 21 measurements of fastener #16 (blank) (right) and fastener #17 (left) (270°) using modified measurement technique
#16 (Blank) #17 (Notched)
Principal Components Analysis (PCA)
• Principal Components Analysis (PCA) used on numerous measurements of like
fasteners
• Statistical technique used to reduce the # of relevant dimensions of large data
arrays to a few vectors that best describes the signals in a least squares sense
• PCA can then reduce each signal measurement to 3 or 4 values (scores) that
describe almost all of the variance between them
• PCA (Z) scores – uncorrelated combinations of original data and eigenvectors,
expressing each signal as a single dimension with maximum amount of variation in
the data accounted for
z� =�X��u��
�
Principal Components Analysis
� = � � � ����+���� �⋯
Results• Eigenvectors
Eigenvector ui
u1
u2
u3
u4
Eigenvalue λiλ1 = 14334.02 λ2 = 102.83 λ3 = 66.02 λ4 = 6.340
Results• PCA Scores – 1st and 2nd components (Z1 and Z2)
Scores of first two principal components for TEC responses of numbered fasteners.Left: fasteners with notches at bore. Right: fasteners without notches at bore
Z1 vs. Z2 – fasteners w/notched holes Z1 vs. Z2 – fasteners w/out notched holes
Results• PCA Scores – 2nd and 3rd components (Z2 and Z3)
Scores of the second and third principal components for TEC responses of numbered, cracked ferrous fastenersLeft: fasteners with cracks at bore. Right: fasteners without cracks at bore
No Crack
Z2 vs. Z3 – fasteners w/notched
holes
Z2 vs. Z3 – fasteners w/out
notched holes
Results
• Eigenvectors
– 3rd eigenvector is what is describing the presence of a defect
Results
• PCA score outliers indicate measurement anomalies
Results• Anomalies – likely due to probe not being centered over fastener, and/or
pickup coils not aligned properly with crack/spar direction
#15
Conclusions
• Reliable method using PCA to detect notches beneath a
non-conducting layer without ferrous fastener removal
has been developed
• A realistic measurement technique, and statistical analysis
method allow for signal variation caused by user errors to
be identified and differentiated from variation caused by
the presence of cracks
• Ongoing development of probe design to detect
• Low-cost system, potentially fast and efficient with real-
time analysis capability
Questions?