Characterization of impact damage in fibre reinforced composite plates using

embedded FBG sensors

J. Frieden*, J. Cugnoni, J. Botsis, Th. Gmür

CompTest2011 5th International Conference on

Composites Testing and Model Identification14 Feb 2011 - 17 Feb 2011,

Ecole Polytechnique Fédérale de Lausanne, Switzerland

Swiss National Science Foundation, grant N° 116715

Objectives

Primary objective of this work:

• Impact localization and damage identification in CFRP plates with FBG sensors

Methods:

• Interpolation-based impact localization method using high rate FBG signals

• Inverse numerical-experimental damage identification method based on eigenfrequency changes and homogenized damage model

Objectives

Primary objective of this work:

• Impact localization and damage identification in CFRP plates with FBG sensors

Today’s focus:

• Influence of impact damage on the plate’s eigenfrequencies measured with FBG sensors

• Experimental characterisation of impact damage

• Finite element model of the plate with impact damage that reproduces the change of eigenfrequencies

Application

Introduction: Materials and specimen

CFRP cross-ply plate with 28 UD plies

[0°2, 90°2, 0°2, 90°2, 0°2, 90°2, 0°2]s

Embedded FBG sensors

Reference: Frieden J. et al, Composite Structures, 2010

Cross-section of plate

Sensitivity of eigenfrequencies to damage

Intact plate:Experimental modal analysis

kintact

kintact

kdamagedk

f

fff

k

intactf

Damaged plate:Experimental modal analysis

kdamagedf

Impact1.7J – 6.7J

Experiment carried out on 8 plates using different impact energies.

Sensitivity of eigenfrequencies to damage

Relative frequency changes as a function of impact energy

kintact

kintact

kdamagedk

f

fff

High resolution X-ray computed tomographySkyScan, model 1076 Aluminium filter : 1 mm thickness X-ray source voltage : 100 kV X-ray source power : 10 W Exposure time : 1750 ms

Experimental damage characterization

Damaged CFRP plate

Experimental damage characterization

Impact Location

Intralaminar cracks are rare and their occurrence is limited to a region located just beneath the impact point

Cross-section (Cut through plate thickness)Impact energy : 5.1 J

CT Resolution: 9 μm/pixelDistance between cross-section images: 9 μmTotal of 10 000 imagesConvert to black & white images

2 mm

Experimental damage characterization

Absorbed energy per unit of delamination area of 280 J/m2.

Detailed 3D delamination model

FE model in Abaqus 6.8-2:• Numerical modal analysis• 20-nodes brick elements with reduced

stiffness matrix integration• Mesh interfaces without node

connection between plies• Element size : 2 mm x 2 mm

Discrete delamination model

FE model in Abaqus 6.8-2:• Numerical modal analysis• 20-nodes brick elements with reduced

stiffness matrix integration• Mesh interfaces without node

connection between plies• Element size : 2 mm x 2 mm

Eigenfrequency changes are mainly due to delamination type damage

Homogenized damage model

Projected damage shape:•Rhombic area

Diagonal damage tensor D:Affected:•Transverse shear moduli

Not affected:•Longitudinal, transverse and through-the-thickness Young’s moduli•In-plane shear modulus•Poisson’s ratio

Incident energy [J] 3.37 5.06 6.75 6.75

Projected area [cm2] 10.7 16.0 20.7 20.6

Length [mm] 56.7 72.0 87.4 81.6

Width [mm] 37.8 47.2 47.3 47.3

232323

131313

1ˆ

1ˆ

GDG

GDG

Material properties:•Through-the-thickness homogenized material properties

Homogenized damage model

Diagonal damage tensor D:Affected:•Transverse shear moduli

Not affected:•Longitudinal, transverse and through-the-thickness Young’s moduli•In-plane shear modulus•Poisson’s ratio

Values of D13 and D23 identified through least square optimization:Minimize error between experimentally measured frequency change and numerically calculated frequency change

Projected damage shape:•Rhombic area

Material properties:•Through-the-thickness homogenized material properties

232323

131313

1ˆ

1ˆ

GDG

GDG

Homogenized damage model

Diagonal damage tensor D:Affected:•Transverse shear moduli

Not affected:•Longitudinal, transverse and through-the-thickness Young’s moduli•In-plane shear modulus•Poisson’s ratio

Values of D13 and D23 identified through least square optimization:Minimize error between experimentally measured frequency change and numerically calculated frequency change

Incident energy [J] 3.37 5.06 6.75 6.75

D13 [%] 84.4 86.1 88.4 91.9

D23 [%] 85.5 90.2 92.5 93.9

Projected damage shape:•Rhombic area

Material properties:•Through-the-thickness homogenized material properties

232323

131313

1ˆ

1ˆ

GDG

GDG

Prediction of eigenfrequency changeExperimentally measured damage size

Using the previously determined values of D13 and D23

Damage identification procedure

Values of D13 and D23 are fixed to 94 %

Parameters to identify:

• Damage position

• Damage surface

• Damage aspect ratio

Reduce discrepancy between experimentally measured eigenfrequency changes and numerically calculated eigenfrequency changes

Iterative minimization algorithm: Levenberg-Maquardt

Application example

Impact energy: 3.4 J

Predict the impact location

Identify damage size and position

Application

Reference measurements before impact:

• Arrival time delays for interpolation-based localization method

• Eigenfrequencies of intact plate

Application: Reference data

• Non-destructive hammer impacts• Grid of 3 x 3 reference points• Acquisition rate of FBG sensors : 1 GHz• Arrival time delays obtained by threshold method

Application: Reference data

• Non-destructive hammer excitation• Grid of 3 x 3 reference points• Acquisition rate of FBG sensors : 100 kHz• Eigenfrequencies obtained by modal curve fitting

FRF

Application: Impact

• Impact with energy of 3.4 J• Acquisition rate of FBG sensors : 1 GHz

Application: Impact

• Impact with energy of 3.4 J• Acquisition rate of FBG sensors : 1 GHz

Application: Identification of damage

Experimental data

Application: Identification of damage

Parameters to identify:

• Damage position

• Damage surface

• Damage aspect ratio

Initial guess for the damage identification:

• Predicted impact location

• Damage surface = 1 cm2

• Damage aspect ratio = 1

Experimental data

Application: Identification of damage

Convergence graph

Identification resultsPredicted eigenfrequency changes

compared to experimental eigenfrequency changes

Conclusion

• Embedded FBG sensors provide very accurate strain data for modal analysis and acoustic wave sensing.

• The eigenfrequency changes can be mainly attributed to delamination type damage.

• The simple homogenized damage model allows to reproduce the eigenfrequency changes.

• The damage size can be identified by a numerical-experimental optimization method based on eigenfrequency changes.

Thank you

Introduction: Fast FBG interrogation

FBG sensors for modal analysis

1st mode

3rd mode

2nd mode

4th mode