A HYBRID APPROACH TO STRUCTURAL HEALTH
MONITORING USING NONDESTRUCTIVE EVALUATION (NDE)
AND ACTIVE DAMAGE INTERROGATION
R. W. Engelbart, D. D. Palmer, Jr., and 1D. M. Pitt
Nondestructive Evaluation
1
Smart Structures and Systems
Phantom Works
The Boeing Company
St. Louis, MO 63166
ABSTRACT. With the useful lives of many aircraft being extended, in-service inspection, repair,
and related support activities represent a considerable commitment of resources. Periodic
maintenance operations involve comprehensive nondestructive inspections (NDI) and significant
aircraft downtime for compilation of data, disposition of results, and any needed repairs. Confining
the inspections to known or suspected problem areas still requires time to cover the area adequately,
and to detect and assess flaws. Reducing this inspection time requires advance knowledge of the
condition of the structure with respect to the existence and probable location of defects. Recent work
has demonstrated the viability of active damage interrogation (ADI) for continuous monitoring of
structural health. Piezoelectric transducers are used to actively excite and sense the vibration
characteristics of the structure and use this information to make estimates regarding the health of the
structure. By detecting changes in the structure's vibration signature, damage can be detected,
localized, and assessed; however, additional information may be needed to accurately quantify flaws
or damage. This paper will present a concept for a hybridization of ADI and NDI technologies.
INTRODUCTION
A current focus area of research in smart structures is the development and
implementation of an automated structural health (SHM) monitoring system to provide
diagnostic and prognostic capability for an aircraft's health management system using
smart sensors and actuators integrated into the structure. A reliable SHM system will
permit the use of a condition-based maintenance approach to aircraft in service. Such an
approach can significantly reduce life cycle costs by eliminating unnecessary inspections,
by minimizing the time and effort expended on those that are necessary, and extending the
useful life of new and aging aircraft structural components. The most promising techniques
under development involve the use of piezoelectric transducers to actively excite and sense
the vibration characteristics of the structure and use this information to make estimates
regarding the health of the structure. These techniques are encompassed in an SHM method
developed at The Boeing Company called Active Damage Interrogation (ADI), (U.S. Patent
#US6006163). By detecting changes in the structure's vibration signature, damage can be
detected, localized, and assessed. Full utilization of these techniques in a reliable and
practical SHM will require a sophisticated on-board data storage and management system,
CP657, Review of Quantitative Nondestructive Evaluation Vol. 22, ed. by D. O. Thompson and D. E. Chimenti
© 2003 American Institute of Physics 0-7354-0117-9/03/S20.00
1577
one that can acquire flaw and damage data from all sensor locations during flight, catalogue
and store the data, and download it in a format easily understood by maintenance
technicians. This paper suggests a concept for a more immediate approach using a
combination of ADI and the conventional nondestructive inspection techniques already in
use in maintenance operations.
OVERVIEW OF ADI TECHNOLOGY
The ADI system is a sensor-based approach for health monitoring which provides a
method for detecting and localizing structural damage according to changes in the
characteristic vibration signature of the structure. The ADI system uses piezoelectric (PZT)
transducers attached to the structure which act as both actuators for exciting the structure,
and as sensors for measuring the vibration response of the structure. When the ADI system
is operating, a single transducer at a time is used as an actuator, while the other transducers
act as sensors, receiving the actuator input. As each transducer in turn becomes an actuator,
the others "listen." The structure's vibration signature is then characterized by computing
the transfer function between each actuator and sensor. This transfer function contains the
magnitude and phase information versus frequency for the sensor response as a function of
the actuator input. These transfer functions are compared to a reference, or baseline, which
represents the normal "healthy" state of the structure. The baseline is generated by
collecting several sets of actuator/sensor data when the structure is in a healthy state. The
mean and standard deviation is then computed to produce the baseline.
The difference between the currently measured transfer function and the baseline
transfer function is normalized by the baseline's standard deviation. This provides a
measure of the vibration signature change in terms of the number of standard deviations
from the baseline. This normalization also compensates for varying signal-to-noise ratios
(SNRs) among the sensors. The normalized difference is input into a windowed averaging
process and then integrated (summed) across the frequency range to produce a cumulative
average delta (CAD), which provides a single metric for indicating damage. This
information processing flow is outlined graphically in the block diagram of Figure 1.
The transfer function magnitude and phase CAD values computed for all sensors in
relation to a single actuator are then averaged to produce a single "damage index" assigned
to each piezoelectric transducer. These damage indices are then sent to downstream
processing algorithms for both detection and localization of the damage. As in all statistical
detection schemes, the detection threshold value must be established according to a
specified cost function that optimally maximizes detection probability while minimizing the
false alarm rate. This threshold should account for all unmodeled variations that normally
occur in the undamaged state, and for changes due to damage smaller than the required
minimum flaw size. Thus the detection threshold for the damage index must be determined
based on variation due to structural loading, environmental conditions, and the minimum
flaw size to be detected.
APPLICATION OF ADI TO AIRCRAFT STRUCTURE
Figure 2 shows a stiffened composite aircraft panel used to test the function of the
ADI system. Piezoelectric transducers were attached to the side of the panel shown; the
panel was then subjected to an array of varying impact loads designed to establish the
damage indices of the transducers as well as to demonstrate flaw detection capability.
Figure 3 shows the locations of the transducers and damage sites.
1578
damage indices of the transducers as well as to demonstrate flaw detection capability.
Transfer Function Amplitude for
Baseline Mean and Standard Deviation
Figure 3 shows the locations of the transducers
and #1damage
sites.
Actuator #4/Sensor
(12 Data Sets)
T
Compute Deltas (CADs) for all Sensors
and Compare to Threshold
Compute
ComputeDeltas
Deltas(CADs)
(CADs)for
forallallSensors
Sensors
and
andCompare
ComparetotoThreshold
Threshold
FIGURE 1. ADI System Block Diagram.
10000
Frequency (Hz)
15000
-40
Amplitude Information
-50
Transfer Function Magnitude (dB)
10
0
15000
20000
Sigma
-50
-60
0
5000
10000
Frequency (Hz)
15000
-20
-30
Phase Information
Mean
-40
200
-50
0 Amplitude
0.5
1
1.5
2
Information
4
Frequency (Hz)
x 10
0
-20
-40
100
-60
0
50
0.5
0
100 0
1
Frequency (Hz)
0.5
1.5
2
x 10
1
Frequency (Hz)
4
1.5
2
x 10
0
0.5
1
Frequency (Hz)
10
1.5
4
2
x 10
0.5
1
Frequency (Hz)
1.5
2
x 10
5
0
FIGURE
FIGURE 1.1. ADI
ADISystem
SystemBlock
BlockDiagram.
Diagram.
FIGURE 2. Composite Aircraft Panel.
FIGURE 2. Composite Aircraft Panel.
FIGURE 2. Composite Aircraft Panel.
1579
4
0.5
1
Frequency (Hz)
1.5
2
x 10
4
4
10000
Frequency (Hz)
15000
0.5
1
Information
Frequency (Hz)
20000
1.5
500.5
0
100
1
Frequency (Hz)
0
0.5
1.5
0
0
0 0
4
1.5
1
2
2
x 10
4
50
0
15
10
15
0
0.5
10
1
Frequency (Hz)
1.5
2
x 10
4
5
0
0
0.5
1
Frequency (Hz)
1.5
0
0
0.5
1
Frequency (Hz)
1.5
Frequency (Hz)
2
x 10
3
Sensor ID
2 2
3 3 4
Sensor
ID ID
Sensor
4
4
5
5
5
6
4
x 10<
6
6
2
x 10
5
nnn
1 1
4
2
x 10
1
Frequency (Hz)
42
20
2
x 10
100
0
4
15
0
0
Phase
0
-200
5
0
5000
-200
200
50
15
0
0
0
-60
0
Mean
-40
-60
10000
Frequency (Hz)
-40
10
-20
-10
TF Phase (deg)
5000
-20
TF Phase (deg)
0
-10
bin Delta (Sigma)
-70
TF Amplitude (dB)
0
-60
20
Transfer Function Magnitude (dB)
5000
-30
bl03a
bl03b
bl03c
20000
bl03d
bl03e
bl03f
bl03g
bl03h
bl03i
bl03j
bl03k
bl03l
0
Avg. Cum. Delta (Sigmas)
Integrate
toto
IntegrateOver
OverDesired
DesiredSpectrum
Spectrum
Compute
Delta
ComputeCumulative
CumulativeAverage
Average
Delta
(CAD)
(CAD)for
forAmp
Ampand
andPhase
Phase
-10
-70-20
0
bin Delta (Sigma)
Integrate Over Desired Spectrum to
Compute Cumulative Average Delta
(CAD) for Amp and Phase
0
-60
Sigma
10
-30
Baseline
Mean and Standard Deviation
Avg. Cum. Delta (Sigmas)
Baseline Mean Spectrums. Normalize by Baseline
Standard Deviation Spectrum and Perform
Windowed Averaging of Delta Magnitudes
SubtractTF
TFAmp
Amp&&Phase
PhaseSpectrums
Spectrums
from
Subtract
from
BaselineMean
MeanSpectrums.
Spectrums.Normalize
Normalizebyby
Baseline
Baseline
Baseline
StandardDeviation
DeviationSpectrum
Spectrum
and
Perform
Standard
and
Perform
WindowedAveraging
AveragingofofDelta
DeltaMagnitudes
Magnitudes
Windowed
-50
TF Amplitude (dB)
bin Delta (Sigma)
ComputeTransfer
TransferFunction
Function(TF)
(TF)
Amplitude
Compute
Amplitude
and Phasefor
for CurrentCondition
Condition
Subtract TF and
AmpPhase
& PhaseCurrent
Spectrums from
Transfer Function Amplitude for
Actuator #4/Sensor #1 (12 Data Sets)
-40
bin Delta (Sigma)
Avg. Cum. DeltaAvg.
(Sigmas)
Cum. Delta (Sigmas)
Compute Transfer Function (TF) Amplitude
and Phase for Current Condition
-30
Delta (Sigmas)
Delta (Sigmas)
Generate Reference "Baseline" Mean
Generate Reference “Baseline” Mean
and Standard Deviation Spectra for
and Standard Deviation Spectra for
Transfer Function Amplitude and
Transfer Function Amplitude and
Phase from Several Sets of Data
Phase from Several Sets of Data
Collected
in the Undamaged Condition
Collected in the Undamaged Condition
20
bl03a
bl03b
bl03c
bl03d
bl03e
bl03f
bl03g
bl03h
bl03i
bl03j
bl03k
bl03l
-20
Transfer Function Magnitude (dB)
Generate Reference “Baseline” Mean
and Standard Deviation Spectra for
Transfer Function Amplitude and
Phase from Several Sets of Data
Collected in the Undamaged Condition
Transfer Function Magnitude (dB)
0
-10
4
20000
FIGURE3.3.Composite
CompositePanel
Panelwith
withTransducers
Transducersand
andDamage
Damage Locations.
Locations.
FIGURE
FIGURE 4. Growth of Impact Damage at Location B.
FIGURE 4. Growth of Impact Damage at Location B.
The most significant result was obtained at location B, as illustrated in Figure 4.
most significant
result
obtained grew
at location
B, as illustrated
Figureof4.
UnderThe
successive
impact loads,
thewas
delamination
from approximately
the in
diameter
Under
successive
impact
loads,
the
delamination
grew
from
approximately
the
diameter
of
a quarter to five inches in length. The transducers near the damage location showed
a dramaaitic
quarter to increases
five inches
in
length.
The
transducers
near
the
damage
location
showed
in their damage indices between the three and four foot-pound impact
dramaaitic
increases
in their the
damage
indices
the not
three
andtofour
foot-pound
loads. This
demonstrated
capability
of between
the system
only
detect
damage, impact
but to
loads.
This
demonstrated
the
capability
of
the
system
not
only
to
detect
damage, but to
track changes in the extent of the damage.
track changes in the extent of the damage.
HYBRIDIZATION CONCEPT
HYBRIDIZATION CONCEPT
During scheduled maintenance operations, each aircraft may undergo a number of
During
scheduled maintenance
each be
aircraft
may undergo
a number
of
scheduled
nondestructive
inspections.operations,
These may
performed
on parts
that are
scheduled
inspections.
These
may or
be that
performed
on parts
that are
considerednondestructive
critical by design
and structural
analysis,
have a history
of problems
considered
design and
structural Other
analysis,
or that have
a history of problems
discoveredcritical
duringbyprevious
inspections.
inspections
are unscheduled
and are
discovered
previous
inspections.
inspections
are unscheduled
and are
performed during
to provide
additional
informationOther
because
of a condition
found in the field.
In
performed
to provide
a condition
found
the field. an
In
some of these
cases, additional
dependinginformation
on the typebecause
of flawof suspected
(such
as in
corrosion),
some
theseexamination
cases, depending
type of the
flaw
suspected
as corrosion),
an
initialofvisual
will be on
ablethe
to identify
specific
area to(such
be inspected,
reducing
the actual
time.
however,
there
evidence
for
initial
visualinspection
examination
willIn
bemany
able toinstances,
identify the
specific
areaistono
bevisual
inspected,
reducing
guidance
the entire
area In
must
be inspected.
the
actual and
inspection
time.
many
instances, however, there is no visual evidence for
guidance and the entire area must be inspected.
1580
In the case of impact damage to a composite surface such as a wing, visual
evidence may or may not be present. The outer surface sustains the least damage from
impact; the condition becomes much more extensive through the thickness of the laminate.
Visual inspection for evidence of suspected damage depends greatly on the skill and visual
acuity of the inspector. Such inspections can be fatiguing and may yield no useful
information. Without some visual indication of damage location, an ultrasonic inspection
of any large area of composite surface is very time consuming.
The transducers of the ADI system are intended for placement on skins and
structural members of wings and stabilizers. During system operation, damage is
approximately located by identifying the sensor(s) with the strongest responses, very similar
to triangulation in acoustic emission testing. For the transducers attached to the first layer
(below the wingskin), it is possible to reduce the inspection area by using ultrasonic NDI to
establish the sensor locations and the boundary of the suspect area. There are two potential
approaches for this; one is an active approach, using conventional ultrasonic transducers to
locate the ADI transducers on the back side of the laminate. The second is a passive
approach, utilizing a low frequency transducer to "listen" as the ADI transducers are
triggered in "actuator" mode. Either approach requires the establishment of charateristic
signals by which to recognize sensor locations. When those locations have been
pinpointed, conventional NDI is used to accurately locate and the flaw or damage that
was approximated by the ADI system. This hybrid technique can also be used to verify
that ADI transducers are still attached in correct locations.
CONCLUSION
Aircraft in-service inspection time can potentially be reduced and more meaningful
data can be acquired through the use of a hybrid of active damage interrogation and
nondestructive inspection. Work is needed to exploit the best features of both technologies
and to ensure that the two are developed to work effectively together.
REFERENCES
1.
J. Dunne, D. Pitt, K. Killian, and D. Sofge, "Recent Advances in Active Damage
Interrogation," Proceedings, 42nd AIAA/ASME/ASCE/AHS/ASC Structures,
Structural Dynamics, and Materials Conference, 2001.
1581