advanced eddy current imaging - トップページ | tia institute for ceramic technologies and...

F R A U N H O F E R I N S T I T U T E F O R C E R A M I C T E C H N O L O G I E S A N D S Y S T E M S I K T S

Advanced Eddy Current Imaging

High Frequency | Industrial Solutions

CFRPs | Ceramics | Coatings | Polymers

Customized developments | validation of systems | delivery of components and algorithms | OEM supplier

Fraunhofer Institute for Ceramic

Technologies and Systems, Branch

Materials Diagnostics IKTS-MD

Maria-Reiche-Strasse 2

01109 Dresden, Germany

Contact

Jun.-Prof. Dr.-Ing. Henning Heuer

Phone +49 351 88815-630

[email protected]

www.ikts.fraunhofer.de

F R A U N H O F E R - I N S T I T U T F Ü R K E R A M I S C H E T E C H N O L O G I E N U N D S Y S T E M E I K T S

3D-FREIFORMSCANNER FÜR

BILDGEBENDES ABRASTERN

Für Untersuchungen an Bauteilen stehen

Ultraschall-, Wirbelstrom-, OCT- und weitere

Sensoren zu Verfügung. Für die Materialdi-

agnose soll eine abbildende Messtechnik

verwendet werden. Um gerasterte Bilder von

oft unbekannten Formen herzustellen, ist

der Einsatz eines robotergeführten Sensors

naheliegend. Notwendige Voraussetzung ist,

den Aufwand der Programmiertätigkeiten

für dieses mechanische Prüfsystem gering zu

halten. Aktuell wird das Drapierverhalten

von CF-Gelegen bei der Herstellung von

Leichtbauformteilen untersucht.

Methode

Ein unbekanntes Bauteil wird mit einer

Streifenlichtkamera digitalisiert. Auf der

rückgeführten Fläche wird eine parametri-

sche Bahnplanung durchgeführt. Diese

Parameter (z. B. Bahnabstand, Überfahrbe-

reich, Sensororientierung und Offsettie-

rung) werden in ein Programm überführt

und in einer virtuellen Kollisionsprüfung auf

Erreichbarkeit kontrolliert bzw. justiert.

Anschließend wird der Sensor orthogonal

über die Bauteiloberfläche gesteuert. Die

Sensormessergebnisse und die Bahndaten

werden zu einem Rasterbild (C-Scan) zu-

sammengesetzt.

Vorteile

Eine schnelle Adaption von Prüfaufgaben

an topografischen Oberflächen wird durch

eine virtuell justierte Oberflächendigitalisie-

rung mit Flächenrückführung möglich. Das

Scanraster wird parametrisch auf der ge-

wonnenen Fläche generiert, womit das

zeitaufwändige Programmieren des Robo-

ters entfällt. Durch vier verschiedene Siche-

rungen wird gewährleistet, dass Messgerät,

Prüfling und Messergebnisse nicht beschä-

digt werden. Das Konzept ist variabel für

berührende und berührungslose Sensoren

einsetzbar. Durch die Reproduzierbarkeit

der Oberflächenabtastung können ver-

schiedene Methoden evaluiert werden.

Fraunhofer-Institut für Keramische

Technologien und Systeme IKTS

Maria-Reiche-Straße 2

01109 Dresden

Ansprechpartner

Jun.-Prof. Dr. Henning Heuer

Telefon 0351 88815-630

[email protected]


1 Digitalisierung mit Streifenlichtkamera

2 Prüfling nach Flächenrückführung

3 Parametrisch generiertes Scanraster (rot)

4 Virtuelle Kollisionsprüfung

5 Mechanisch geführtes Abrastern mit

Roboter.

6 Wirbelstromscan-Ergebnis der CF-Gelege-

struktur.

1 2 3


3D FREE FORM IMAGING

SCANNER

Materials diagnostics for 3D unacquainted

Materials diagnostics for 3D unacquainted

forms can be facilitated by guiding a sensor

with a robot across the specimen and

mapping the sensor information to an

image. The mechanical scanning system

developed by Fraunhofer IKTS-MD reduces

programming efforts in order to adapt it

easily to very different testing tasks. The

scanning system can be used e. g. with

ultrasonic, eddy current and optical sen-

sors, also as parts of tomographic systems.

Currently, we are applying the 3D free

form scanner to investigate the manufac-

turing process of lightweight form parts

made of CFRP.

Method

The specimen with the 3D unacquainted

form is digitalized with a stripe light cam-

era. Based on the calculated surface, a

parametric path is planned, specifying path

distance, scan area, sensor orientation and

offset. These parameters are input to the

robot program and are tested and adjusted

by performing a virtual collision check.

Subsequently, the sensor is guided orthog-

onally across the surface. The measuring

results and path information are composed

to the mapping image (C scan).

Advantages

The robot based scanning system of

Fraunhofer IKTS-MD supports a fast adap-

tation to testing tasks on varying topogra-

phies. This is realized by a virtually adjusted

surface digitalization and surface calcula-

tion. The scan path is parametrically gener-

ated for the calculated surface, saving a

time-consuming programming of the ro-

bot. A fourfold safeguarding mechanism

guarantees that measuring device, speci-

men and acquired data are not damaged.

This concept is flexibly applicable for con-

tact and non-contact sensors. Due to its

reproducibility of surface scanning, differ-

ent techniques can be evaluated.

1 Digitalization with stripe light camera.

2 Calculated specimen area.

3 Parametrically generable scan pattern

(red).

4 Virtual collision testing.

5 Mechanically guided scan with robot.

6 Eddy current c scan result of cf structures.


Technologies and Systems IKTS



Contact

Jun.-Prof. Dr. Henning Heuer

Phone +49 351 88815-630

[email protected]


4 5 6


EDDYCUS® POLARLAB

Labormessgerät zur Winkellagenver-

messung von CFK-Bauteilen

Werkstoffe aus Kohlefaserkompositen

haben sich in Bezug auf ihre mechanischen

Eigenschaften bewährt. Aufgrund der

oftmals sehr komplexen Fertigungsverfah-

ren ist es notwendig, Fabrikationsfehler im

unverharzten als auch konsolidiertem Zu-

stand zerstörungsfrei zu prüfen. Hierfür

eignen sich die bewährten Prüfsysteme der

EddyCus®-Serie zur bildgebenden Wirbel-

stromprüfung. Diese Systeme benötigen

durch die abrasternde Messdatenaufnahme

eine gewisse Zeit bis auswertbare Daten

verfügbar sind. Oftmals bedarf es jedoch

nur einer genauen Kenntnis der Lagenori-

entierung an spezifischen Punkten auf dem

Bauteil, um Aussagen zur mechanischen

Festigkeit ableiten zu können.

Wirtschaftlich sinnvoll dafür erscheint die

Substituierung der klassisch angewandten

computertomografischen Messung durch

eine selektive winkelaufgelöste Hochfre-

quenzprüfung. Erste, mit einem neuartigen

Geräteprototyp EddyCus® PolarLab aufge-

nommene Polardiagramme zeigen eine

eindeutige Korrelation zu den CT-Auf-

nahmen.

Technische Spezifikationen

- Prüffrequenzen: 100 KHz–100 MHz

- Bis zu 4 Frequenzen parallel im Zeit-

multiplexbetrieb

- Prüfzeit: 6 Sekunden

- Empfindlichkeit:10 Lagen ∑ 900g/m²

- Automatische Nullpunktkalibrierung

- Messung in Reflektion und Transmis-

sionssetup möglich

- Automatische Vermessung der gefun-

denen Winkellagen

- Bestimmung von winkel- und amplitu-

denabhängigem Anisotropiewert

- Datenexport zur externen Datenaus-

wertung

- CE-Konformität nachgewiesen




01109 Dresden

Ansprechpartner

Martin Schulze

Telefon 0351 88815-628

[email protected]


1 EddyCus® PolarLab-Rotationsprüfstand.

2 Frequenz-Setup.

3 Reflektions- und Transmissionanordnung.

4 Datenauswertung und Exportfunktion.

2 1


EDDYCUS® POLARLAB

Laboratory device for angle resolved

layer evaluation of CFRPs

Carbon fiber composites have proven their

mechanical properties. Owing to the often

very complex production processes, it is

necessary to be able to test fabrication

defects in the dry and in the consolidated

state non-destructively. The EddyCus®

testing systems for eddy current imaging

are suitable for this purpose. However, due

to the scanning data acquisition, these

systems require a certain amount of time

until evaluable data are available. Very

often it is only necessary to have a precise

knowledge of the layer orientation of the

CFRP at specific positions on the compo-

nent in order to be able to derive state-

ments on the mechanical strength.

For economic reasons and radiation protec-

tion restrictions, the standard applied com-

puter tomography measurement was sub-

stituted by a selective angle-resolved high-

frequency testing system. The so called

polar diagrams from the light-weight com-

ponent acquired with the novel device

prototype EddyCus® PolarLab show a clear

correlation with the CT images.

Technical specifications

- Measuring frequencies:

100 KHz–100 MHz

- Up to 4 frequencies in time-multiplex

operation

- Testing duration: 6 seconds

- Sensitivity: up to 10 layers ∑ 900g/m²

- Automatic zero point calibration

- Reflection and transmission operation

modes

- Automatic measurement of detected

layers

- Determination of angle and amplitude

dependent anisotrophy magnitude

- Data export option for external evalua-

tion

- CE conformity approved

1 EddyCus® PolarLab test stand.

2 Frequency setup.

3 Sensor setup (reflection/transmission).

4 Measurement view and data export.




01109 Dresden

Germany

Contact

Martin Schulze

Phone +49 351 88815-628

[email protected]


3 4

343-W-17-3-6


WIRBELSTROMSCANNER

EDDYCUS® MPECS

Der EddyCus® MPECS wurde als Offline-

Lösung zur prozessnahen Stichprobenprü-

fung von Kohlefaserverbundwerkstoffen

(CFK) entwickelt. Aufbauend auf Erkennt-

nissen zur Detektion von Kohlenstofffaser-

lagen an planaren Objekten wurde ein

Freiformscanner entwickelt, dessen Senso-

ren leicht und nahezu druckfrei über die

Oberfläche gleiten.

Methode

Wirbelstrom basierte Prüfmethoden nutzen

die elektrischen Eigenschaften der Kohlen-

stofffasern zur Qualitätsbeurteilung und

eignen sich durch ihre einfache Anwend-

barkeit (koppelmittelfrei, ohne Strahlen-

schutz) für die schnelle prozessnahe Prü-

fung. Der universell parametrierbare Scan-

ner ist durch die Erzeugung von verzer-

rungsfreien Leitfähigkeitsbildern auch an

realen (3D)-Strukturen einsetzbar. Uneben-

heiten werden fast vollständig ausgegli-

chen, um Abhebeeffekte zu minimieren.

Kenndaten

Frequenzbereich 100 kHz–100 MHz

Anzahl Frequenzen 1–4

AC Verstärkung 0– 43,5 dB

Abtastrate 2000 Samples/s

Max. Scanfläche 300 x 300 mm

Min. Schrittweite 0,255 mm

Fahrgeschwindigkeit 500 mm/s

Topographienach-

führung

100 mm (aktiv),

30 mm (passiv)

Wirbelstromsenso-

ren

an die Messaufgabe

angepasst

Videokamera 1 Megapixel

Vorteile

- Hoher Scangeschwindigkeit bei gleich-

zeitig hoher Auflösung

- Nachführung des Sensors auf schrägen

planaren Flächen

- Flexible Parametrierung der verschiede-

nen leicht auswechselbaren Sensoren




01109 Dresden

Ansprechpartner

Jun.-Prof. Henning Heuer

Telefon 0351 88815-630

[email protected]


1 EddyCus® MPECS-Auswerteeinheit und

Aufsatzscanner als transportables System.

2 Wirbelstromsensor und Videokamera auf

gefederter Nachführung.

3 Wirbelstromscan-Bild (C-Scan) mit ver-

schiedenen Gelegelagen.

4 FFT-Lagenseparation (Rohbild, C-Scan).

5 FFT-Lagenseparation (0°-Lage).

6 FFT-Lagenseparation (+ 45°-Lage).

7 FFT-Lagenseparation (- 45°-Lage).

1 2 3


EDDY CURRENT TESTING

DEVICE EDDYCUS® MPECS

The EddyCus® MPECS was developed as an

offline solution for process-oriented sam-

pling inspection of carbon fiber composites

(CFRP). Based on the technical expertise for

the detection of carbon fiber layers of

planar objects, a free-form scanner was

developed, whose sensors glide slightly and

nearly pressure free over the surface.

Method

Eddy current based test methods, using the

electrical properties of carbon fabric fibers

for quality assessment, are well suited to

perform fast process-oriented testing due

to their simple applicability (couplant free

without radiation protection). The all-

purpose configurable eddy current scanner,

producing distortion-free conductivity

images, is deployable on real (3D) struc-

tures. Surface roughness is almost com-

pletely balanced to minimize lift-off effects.

Characteristics

Frequency range 100 kHz–100 MHz

Scanning frequencies 1–4

AC gain 0– 43.5 dB

Sample rate 2000 Samples/s

Max. scan area 300 x 300 mm

Min. pitch 0.255 mm

Speed 500 mm/s

Topography

arrangement

100 mm (active),

30 mm (passive)

Eddy current sensors Diff. types for

special applica-

tions

Video camera 1 Megapixel

Advantages

- High scanning speeds at high resolution

- Tracing of the sensor positions on slop-

ing, planar surfaces

- Flexible configurations of various easily

replaceable sensors

1 EddyCus® MPECS portable system with

analysis unit and attachment scanner.

2 Eddy Current sensor and video camera

with spring-loaded tracking.

3 Eddy Current image (C scan) of various

fabric layers.

4 FFT layer separation (raw image, C scan).

5 FFT layer separation (0° layer).

6 FFT layer separation (+ 45° layer).

7 FFT layer separation (- 45° layer).





Contact

Jun.-Prof. Henning Heuer

Phone +49 351 88815-630

[email protected]


4 6 5 7 50 mm 50 mm 50 mm 50 mm

EddyCus® MPECS HIGH RESOLUTION MULTIFREQUENCY EDDY CURRENT IMAGING DEVICE

The Multi Parameter Eddy Current Scanner

utilizes Eddy Current (EC) techniques for

evaluation of the local conductivity and

the generation of conductivity images for

quality assurance in carbon fi ber materials.

The testing system is designed for reliable

inspection of Raw Carbon Fiber (RCF) or

Non-Crimp-Fabrics (NCF) material of various

production types or Carbon Fiber Reinforced

Plastics (CFRP). Carbon fi ber based materials

show a marginal electrical conductivity that

is suffi cient to induce Eddy Currents in the

fi bers. The applied sensors have a very high

spatial resolution especially eligible for carbon

materials. The device scans with 4 discrete fre-

quencies whereas the variation of sensor pa-

rameters and frequencies allow the detection

and separation of various defects in multiple

layers depending on the sample. Imaging and

an optimized image pre-processing enables

to detect defects and inhomogeneities such

as, missing fi ber bundles, lanes, suspensions,

fringes, missing threads and angle errors.

Applications

Quality assurance of CFRP fabric –

and plates

Inline Inspection–

Defect detection in multiple layers–

Detection of missing bundles, lanes, –

suspensions and angle errors

Characteristics

Contact-free –

Measurement duration for 120 x 120 –

mm2 @ scan resolution: 0,5 x 0,5 mm2

Sensor confi guration: 4 minutes –

Frequency sweeping for up to 256 –

frequencies for evaluation process

Realtime imaging of results–

Automated evaluation with colored –

failure coding and comprehensive

fi ltering options

Data archiving–

1

1 EddyCus® MPECS.





01109 Dresden | Germany

Contact:

Jun.-Prof. Dr.-Ing. Henning Heuer

Phone +49 351 88815-630

[email protected]


Characteristics

EddyCus MPECS 15 (MPECS 100)

Frequency range (freely selectable) 10 kHz - 15 Mhz (MPECS 100

handles up to 100 Mhz)

Scanning frequencies 1 - 4

Channels 1

AC Gain / DC Gain 43,5 dB / 50 dB

Max. scanning area 300 x 400 mm

Min. scanning resolution 20 μm

Max. sample height 50 mm

Speed Max. 100 mm/s

Sensor arrangement 1 x single sensor

Measurement spot size 1 x 1 mm

Camera Included

Detectable effects

Misplaced lanes –

Angle Errors –

Missing fi bre bundles –

Suspensions –

Fringes –

Missing threads –

Thread undulation –

Inhomogeneities –

1-3 Defects in CFRP fabric

(missing bundles, lanes and fringes).

1 2 3

Fail

ure

De

tect

ion

Laye

r D

eco

mp

osi

tio

n

Fraunhofer Institute for Ceramic Technologies and Systems IKTS Maria-Reiche-Strasse 2, 01109 Dresden, [email protected]

Dipl.-Ing. (BA) Martin SchulzeJun. Prof. Henning HeuerDr. Dieter JoneitProf. Norbert Meyendorf

Modular high frequency eddy current sensor system and image preprocessing algorithms for CFRP testing

The project was funded by the German VDI / VDE under the funding program no.: 16SV5292

Project partner:

G. Zeidler, A. Wegner, S. PetrenzKARL MAYER MALIMO Textilmaschinenfabrik GmbH [email protected]

R. SeligmannDELTEC GmbH, [email protected]

R. Schallert, R. Sydow, S. Franciscoarxes GmbH, [email protected]

Typical defects in raw carbon fiber material and CFRP – scanned with high resolution Eddy Current single sensors. First non modular arrays for inspection of metallic surfaces were tested in 2010. This type was the basis for the project.

Motivation

Implementation of multiparametrizable Eddy Current array at the end of production proccess of KMM Malitronic MULTIAXIAL knitting machine.

Solution Results and further stepsWire diameter: 60 m CuProblem of reproducibilityInductivity in potcore: 8 HPlanar transformer conceptunder development

Modular Eddy Current arrayswith line resolution of 875 m96 sensors / array with 288different sensor combinations(orientations)One multiplexer and EddyCurrent channel per array

Left: multiplexer (MPX)Middle: 3 MPX per arrayRight: high frequencyEddy Current DeviceFrequency range:100kHz – 100MHz

Array combined on rawcarbon fiber materialTriaxial type (+45° / 0° / -45°)Scanning speed up to 6m/minPlanned sensor size: 2,5m

60mm

Huge variation of complex impedanceOptimization of sensor board is in progress (signalpropagation, parasitic effects, HF-matching)Calibration algorithms helps to normalize influenceFirst results on carbon fiber material with low SNR

Eddy Current C-Scans were analyzed with 2D-FFT algorithms. In frequency domain simple layer decomposition is possible by selection of interesting orientations. After segmentation a classification is needed. Gap size, length and orientation measurements with resolution of +-0,5° is possible.

0°

-45°

+45°

2D-FFT Plot

Raw carbon fiber material

Layer fusion

C-Scan on raw carbon fiber material

Coil production

Modular Eddy Current arrays with marked sensor focus points

Multiplexer and stackable multiplexer PCB HF-Eddy Current boards

Modular Eddy Current arrays on raw carbon fiber material

Complex impedance plane

Raw image of real part

Raw image of imaginary part

Calibrated C-Scan with line offset recalculation

First Eddy Current array in 2010

Karl-Mayer-MalimoMalitronic Multiaxial High-tech knitting machine

Single sensor configuration on EddyCus® MPECS industrial

Typical defects in CFRP and raw carbon fiber material

IKTS

lable at ScienceDirect

Composites Part B 77 (2015) 494e501

Contents lists avai

Composites Part B

journal homepage: www.elsevier .com/locate/compositesb

Review on quality assurance along the CFRP value chain e Non-destructive testing of fabrics, preforms and CFRP by HF radio wavetechniques

H. Heuer a, c, *, M. Schulze a, M. Pooch a, S. G€abler a, e, A. Nocke b, G. Bardl b, Ch. Cherif b,M. Klein g, R. Kupke g, R. Vetter d, F. Lenz d, M. Kliem d, C. Bülow f, J. Goyvaerts h, T. Mayer i,S. Petrenz j

a Fraunhofer Institute for Ceramic Technology and Systems, Material Diagnostic, Dresden, Germanyb TU Dresden, Textile Machinery and High Performance Material Technology, Dresden, Germanyc TU Dresden, Chair Sensor Systems for Non-Destructive Testing, Dresden, Germanyd TU Dresden, Institute for Lightweight Engineering and Polymer Technology, Dresden, Germanye Leibniz Institute of Polymer Research, Dresden, Germanyf DLR, Center for Lightweight-Production-Technology, Stade, Germanyg SURAGUS GmbH, Dresden, Germanyh SABCA LIMBURG N.V., Belgiumi BMW AG, Landshut, Germanyj Karl Mayer Malimo, Chemnitz, Germany

a r t i c l e i n f o

Article history:Received 17 October 2014Received in revised form2 February 2015Accepted 4 March 2015Available online 17 March 2015

Keywords:CFRPA. Carbon fibreD. Non-destructive testingHigh frequency eddy currentB: Electrical properties

* Corresponding author. Fraunhofer IKTS-MD, Marden, Germany. Tel.: þ49 351 88815 630; fax: þ49 351

E-mail address: [email protected]

http://dx.doi.org/10.1016/j.compositesb.2015.03.0221359-8368/© 2015 The Authors. Published by Elsevier

a b s t r a c t

Eddy current testing is well established for non-destructive testing of electrical conductive materials [1].The development of radio frequency (RF) eddy current technology with frequency ranges up to 100 MHzmade it possible to extend the classical fields of application even towards less conductive materials likeCFRP [2][3](Table 2). It turns out that RF eddy current technology on CFRP generates a growing number ofvaluable information for comprehensive material diagnostic. Both permittivity and conductivity of CFRPinfluence the complex impedance measured with RF eddy current devices. The electrical conductivitycontains information about fiber texture like orientations, gaps or undulations in a multilayered material.The permittivity characterization influenced by dielectric properties allows the determination of localcuring defects on CFRP e.g. hot spots, thermal impacts or polymer degradation. An explanation for thateffect is seen in the measurement frequency range and the capacitive structure of the carbon rovings.Using radio wave frequencies for testing, the effect of displacement currents cannot be neglectedanymore. The capacitive structures formed by the carbon rovings is supposed to further strengthen thedielectric influences on eddy current measurement signal [3]. This report gives an overview of severalrealized applications and should be understood as a general introduction of CFRP testing by HF RadioWave techniques.© 2015 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND

license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction

Along the value chain of carbon fibre reinforced plastic (CFRP)different non-destructive testing (NDT) methods such as ultrasonic,thermography, X-rays or opticalmethods are successfully applied [7].

ia-Reiche Str. 2, 01109, Dres-88815 509.(H. Heuer).

Ltd. This is an open access article u

However, looking in more detail at the process chain, there is a gapwhere standard NDT methods cannot be used. For automated massproduction facilities based on infusion processes (e.g. RTM e ResinTransfer Molding), it is important to acquire quality parameters priorto the resin infiltration step. This opens up the possibility of in-timeprocess recalibration, for rework or repair of the fiber preform, ifnecessary.With knowledge of the incomingmaterial characteristic atthe textile- or preform stage, the following process steps can beadjusted in-time to reduce final part rejects resulting from defects inthe pre-stage material. If problems like missing or misaligned fibre

nder the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

http://creativecommons.org/licenses/by-nc-nd/4.�0/

mailto:[email protected]

http://crossmark.crossref.org/dialog/?doi=10.1016/j.compositesb.2015.03.022&domain=pdf

www.sciencedirect.com/science/journal/13598368

www.elsevier.com/locate/compositesb

http://dx.doi.org/10.1016/j.compositesb.2015.03.022

http://creativecommons.org/licenses/by-nc-nd/4.�0/



H. Heuer et al. / Composites Part B 77 (2015) 494e501 495

bundles, waves or insertions are detected in time, the further pro-cessing can be stopped and readjusted resulting in less materialwaste. In addition, subsequent process steps can be controlled byutilizing the information on the incoming product quality e.g. gapsbetween fiber bundles which will influence the infiltration behavior.The current trend in aerospace industries is to evolve to more largeprimary aircraft structures in CFRP with high levels of functionintegration, resulting in more expensive parts and increasing theneed for first-time-right products to avoid complex repairs. A criticaldefect originated in the lay-up phase, is Foreign Object Debris (FOD)introduced in the manual process steps. FOD prevention campaignscannot reduce the occurrence to zero and therefore an inspectionstage before subpart consolidation and -final cure can be economi-cally interesting. Visual FOD inspection is no option for CFRP mate-rials due to the lack of transparency. Also does classic ultrasonicinspection techniques require a couple medium (generally water)which should be avoided in contact with prepregs or dry fiber pre-forms. Thermography results in prepreg out-time reductions andprocessing is required to improve FOD contrast. RF radio wavetechnique is a promising alternative without couple medium andhigh sensitivity. This leads to an increasing demand for NDTmethodsthat can be applied inline to dry multilayered carbon textiles as wellas wet and consolidated materials. The application of ultrasound,thermography or shearographie requires solid state material formechanical or thermalwave propagation or for deformation analysesrespectively. Due to this limitation these standard NDT techniquescannot be applied to dry or wet pre stage components.

An ideal NDT method should be applicable along the wholeCFRP value chain, from the fibre bundle over the fabric/prepreg andpreform stage up to components and its life time. This paper pro-vides an overview about a radio frequency technique that wasemerged by an extension of state of the art eddy current tech-niques. Typical tasks of radio frequency techniques are the deter-mination of fibre bundle orientation, misalignments, gap sizedistribution, local fibre areal weight/thickness and waviness. Bymeasuring dielectric properties curing effects, hotspots, dry spotsand polymer degradation can be analyzed, too. The RF techniqueshows a high potential to be used as an integrated inspectiontechnique along the whole value chain and product life time byimaging the electrical and capacitive properties of CFRP non-destructively (Table 1).

2. Radio frequency inspection

Eddy current (EC) technology is a well-established non-destructive method for the characterization of surfaces or materialsby analyzing conductivity and permeability variations [2]. A pri-mary magnetic field is generated when an alternating current isapplied to an induction coil. Eddy currents are generated in a

Table 1Questions to be answered by NDT along the CFRP value chain.

Process step Question to be answered by NDT

1 Carbon rowing Number of filaments?Type of filaments, cuts or damage?

2 Fabric Number of layers and its orientation?Misalignments/waves/gaps?FOD insertions?

3 Preform Right stack pre assembly?Waves or misalignments after draping

4 Component Dry spots, curing process successful?Delamination?

5 Product life cycle Degradation of fibres?Degradation of polymer?Re-qualification or recycling?

conductive specimen when the coil is placed near that specimen(Fig. 1).

The eddy current within the specimen generates a secondarymagnetic field opposed to the primary field. If the material prop-erties are changed e.g. due to a deviation of current paths resultingfrom cracks or insertion in the sample, the secondary field alsochanges and causes an complex impedance shift in the pick-up coil.The measured values from the pick-up coil are evaluated on thecomplex impedance plane. An important parameter is the fre-quency of the excited alternating current. Due to the skin effect, thedepth of excitation in the specimen decreases with increasing fre-quency. The point where the eddy current density has decreased to1/e, or approximately 37% of the surface density, is called thestandard depth of penetration d [1]. The excitation depth into amaterial is affected by the frequency of the excitation currentu ¼ 2pf, the electrical conductivity s, and magnetic permeability mof the sample as given by:

d ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffi2=usm

p(1)

This skin depth effect is the most important limitation for fre-quency selection during inspection. E.g. due to the good conduc-tivity of Aluminum a frequency in the kHz range or lower has to beused to generate an acceptable probing volume for conventionalcrack detection in Aluminum structures. But beside the penetrationproblem frequency has a second important effect. By the use ofhigher frequencies, the penetration depth will be lost butincreasing signal strength can be obtained. The density of the eddycurrent is influenced by the frequency itself. The Faraday's law ofinduction states that the induced voltage is proportional to the rateof change of the magnetic flux. In other words, since a frequency isan inverse of time, when the frequency of the flux increases thepick-up signal, which can equally be regarded as induced voltage(V), also increases. For CRFP a probate solution for determining thesignal Voltage for CFRP is given in Ref. [5]. Higher frequenciestherefore represent a good option through which the sensitivity ofthe Eddy current method can be increased for low conductivematerials such as Carbon materials due to the extended tradeoffbetween penetration depths and signal amplitude.

Depending on conductivity and permeability of the materialunder test, an optimal frequency within the electromagneticspectrum exists, exhibiting a suitable ration between penetrationdepth (equals to probing volume) and signal strength [6]. Table 2gives an overview about the use of electromagnetic spectrum forNDT purposes. The HF range that is suitable for testing CFRP ma-terials is located between the traditional medium frequency eddycurrent and the microwave and terahertz range.

3. Electrical properties of CFRP and measurement setup

Unidirectional single layered carbon fiber material has a con-ductivity up to s ¼ 5*106 S/m in longitudinal and s ¼ 1*103 S/m inthe lateral direction. Through variation of fibre type, fibre orienta-tion, stacking sequence, fibre/volume content, fibre density andcompaction around the carbon rovings, these values differ. In

Fig. 1. Schematic diagram of probe and specimen configuration for eddy currenttesting.

Table 2The electromagnetic spectrum [9,10] and its use for NDT purpose.

Frequency nomenclature NDT purpose

VLF Very low frequency 3 kHze30 kHz Metal objects/high penetration depth (e.g. Steel Plates/tubes with several cm penetration)LF Low frequency 30 kHze300 kHz Metal objects/surface inspection (e.g. Weld inspection, Surface Cracks)MF Medium frequency 300 kHze3 MHz Thin metal objects with high resolution needs (e.g. Crack detection in thin (mm) Aluminum sheets)HF High Frequency 3 MHze30 MHz Weak Conductive Materials (e.g. Thin Film Characterization, CFRP Testing)VHF Very high frequency 30 MHze300 MHz Semiconductors, dielectric spectroscopy (e.g. Solar Cell, Polymers, Curing)UHF Ultra high frequency 300 MHze3 GHz Insulators, coatings, (e.g. Ground Penetrating Radar GPR, Microwave Testing of GFRP)SHF Super High Frequency 3 GHze30 GHz (e.g. Radar, Microwave Testing of dielectric Material, Paint thickness measurement)EHF Extremely High Frequency 30 GHze300 GHz (e.g. Radar, “Terahertz” Testing, Security Application “naked scanner”)THz “Terahertz” 300 GHze3 THz (e.g. “Terahertz” Testing, Spectroscopy)

H. Heuer et al. / Composites Part B 77 (2015) 494e501496

addition to the conductivity of the CFRP also its permittivity isimportant for RF testing. It generates a displacement current in thematerial, which influences the measurement signal as well as theeddy current [11,12]. In a dry carbon material, the permittivity isdominated by the fiber coating and the surrounding air. In a CFRPthe air is substituted by a polymer resin, so that the permittivity isdepending on type and processing quality of the resin [11]. Fig. 2shows the main electrical effects of applying an alternating mag-netic field to strongly anisotropic CFRP material that are influencedby three main parameters.

a) Fiber/Volume Ratio: The Fiber/Volume ratio determines theamount of conductive carbon fibres in a volume section anddefines the average electrical conductivity of the material,which needs to take into account for general measurementparameterization (frequency, penetration depth).

b) Electrical connection between fiber bundles: Depending on thestructure (woven, crimped, non-crimped etc.) chemical andstructural conditions of the interfaces of filaments and theconsolidation density due to mechanical pressing, the electricalcontact between neighboring bundles can vary. Identical ma-terials can provide different degrees of eddycurrent propagationdue to the quality of internal electrical connections duringconsolidation. The horizontal electrical connection between fi-ber bundles (in plane, parallel to the surface) is direct influ-encing the contrast of the RF image. The vertical connection indepth direction is influencing the interlaminar interfaces [4] andthereby the penetration depth.

c) Capacitive effect and displacement current: In addition to theelectrical connection of fiber bundles, the dielectric properties ofthematrixmaterial also influence the complex signal impedance.

Fig. 2. Electric and dielectric behavior of carbon matrix composites [3].

This three parameters can be analyzed by RF EC instruments.Medium frequency eddy current (MF EC) inspection instruments areavailable in many configurations by different commercial suppliers.Typically, the instruments are designed for manual handling of asingle coil sensor or array probe with wheel tracker. The frequencyrange andparameterization of such standard devices is optimized forconventional NDE task such as crack detection or material identifi-cation for metallic specimens. Also, mechanical manipulators like X-Y-Z axle scanners or robot basedmanipulators are in commercial useto scan an EC probe over a 3 dimensional surface (e.g. rotor blades).For low conductive carbon based materials, a contrast enhancementat frequency's up to 50MHz, and in some special cases up to 80MHzwas observed. One reason is the increasing signal strength due to theFaraday law of induction, second the decrease impedance of thecapacitive structure of CFRP at higher frequency [5]. This frequencyrangewas usually observed only in laboratory based equipment [12].Initiated by the increasing demand for HF eddy current measure-ments, instruments operating in the range of 100 kHz up to 100MHzwere developed. The results shown in the following chapters whereacquired with EddyCus® instruments. Combined with a precision X-Y-Z manipulator with minimum 25 mm step width, high resolutionHF EC images can be acquired. The sensors glide non-contact orlightly contacted over the surface. The used system can capture amaximum surface area of 300 mm by 300 mm with a maximumspeed of 300mm/s at a sampling rate of 3000 samples per second. Inaddition, the EddyCus® software provides a frequency sweep modebetween 100 kHze100 MHz with 256 steps or a sequential multi-frequency data acquisition with up to four frequencies. The instru-ment allows the acquisition of complex HF EC signals in amplitudeandphase shift in the complex impedance plane andas a time plot byC-Scan [6].

To perform eddy current measurements in a frequency rangeabove 1 MHz mechanical vibrations (lift off variations) and elec-tromagnetic disturbance needs to be controlled precisely. Also,electrical conductive dust in carbon contaminated environmentscan deteriorate the measurement results. To solve this practicalproblem a robust setup needs to be used that is shielded againstdust and has an EMS safe architecture (Fig. 3).

The used probes for CFRP inspection are based on copper wireswinded on ferrite cores. The coil diameter differs between 2 and7 mm in accordance to the requested spatial resolution and sensi-tivity. The following experimental results were obtained by halftransmission probe configuration with two 5 mm coils. Due to thegeometric distance between the sending and receiving coil (Fig. 1)the sensitive field of view of the probe is elliptical and the observedspatial resolution is higher than the core diameter. The parameter-ization of the frequency is done bya frequency sweeping at differentpoints on the CFRP surface. The frequency's showing the bestcontrast for the specify inspection task are than used for scanning.The resulting frequency parameterization is a function of a) coilproperties, b) material properties to be tested and c) the type ofneeded contrast mechanism inside CFRP. The determination of fiber

Fig. 4. HF EC Images of CFRP plates with in-plane and out-off-plane waves. (Size150 � 150 mm x 5 mm). a) indication of wave position, b) raw data image.

Fig. 3. EddyCus® systems a) Linear Axis system for flat and slightly curved structuresand b) Robot based system 3d structures.1


texture requires contrast between fiber bundles (gaps) whereas theinspection of hot spots or polymer defects need contrast bydielectric properties of the specimen. The used optimal frequency isalways a setup individual parameter and differs between differentsamples and inspection task. The frequency adjustment is typicallydone by 0.1 MHz steps for maximizing the image contrast.

4. Experimental results

4.1. Inspection of texture and waviness

As shown in Refs. [12,13] the HF eddy current technique can beapplied to dry, wet and consolidated carbon based materials. Onemajor task of HF EC is the characterization of waves inside carbonfibre preforms or consolidated materials. The image of Fig. 4 showsin-plane and out-of plane waves perpendicular to each other. TheIn-plane wave can be seen directly due to the typical not straightforward orientation of the fibre bundle. A out-of- planewave showsa modulation of the signal amplitude as gray value contrast.

Due to local changing of fiber volume content the signalamplitude changes depending on the out-of-plane wave position.In Fig. 4 the out-of-plane waves are indicated by black contrastchange compared to the typical undulation of in-plane waves. Dueto this different contrast mechanism it follows that in-plane wavesand out of plane waves require a different algorithm for evaluation.For industrial application of HF EC imaging techniques, the texturalinformation like roving orientation, gap sizes or waviness needs tobe extracted by image processing. Especially for multi axial mate-rials, a two dimensional fast Fourier transformation (2D-FFT) showshigh potential for automated fiber texture analyses [13]. In texturedmaterials the image frequency is correlated with the periodicstructure. In relation to the texture of CFRP fabrics, the image fre-quency correlates with gap size and fiber bundle size whereas therotation of the image frequency maxima represents a layer orien-tation. The Fig. 5a) shows a HF EC Image of a 3 axial non-crimpfabric with two horizontal layers on the back side. The corre-sponding 2D-FFT of Fig. 5b shows the 3 characteristic lines indi-cating the 45�, 0� and �45� layer.

Inside the 2D FFT data in Fig. 5 b each point bar contains in-formation for one specific CFRP orientation. By setting a filter that issuppressing all points except the point bar from the layer andperforming an inverse 2D-FFT the textural image of one individuallayer can be reconstructed. The procedure is descript in Ref. [13].The Fig. 6 shows such a reconstruction of one layer of Fig. 4 withsignificant in-plane undulations [13].

Fig. 5. Analyzing the image frequencies of a 5 layer CFRP (30 � 30 cm) with twohorizontal layers that differ by 2� . a) shows the raw HF EC, b) corresponding 2D FFT

4.2. Characterization of defects according to their depth

The modulation of the signal amplitude as a function of thedepth of an out-of-plane wave can be explained by formula (1) and

1 http://www.youtube.com/watch?v¼wNng6ClM1CM.

(2). The eddy current density decays exponentially with increasingdepth. For an isotropy material with good conductivity, the currentdensity can be calculated by:

JðdÞ ¼ Jse�dd (2)

where Js is the Current density at the surface, d is the depth ofobject and d the penetration depth shown in (2), [8]. But due to thestrong anisotropy of CFRPs, this relation can be used only as roughestimation. To validate the in-depth resolution for a specific CFRPmaterial, a reference sample with insertions in different depth hasto be used. Fig. 7 shows a HF EC Image of CFRP plate with 14thlayers of GV 300 U TFX (12 k) non crimped fabrics with a grammageof 317 g/m2 for each layer and fiber volume content of 67%. Duringthe laydown process, pieces of copper foil where inserted betweendifferent layers. The shape of the copper foil was used to identifythe insertion depth by reference after curing.

With increasing depth the signal amplitude and the phase anglewhere the maximum amplitude occurs is changing. In thefollowing diagram Fig. 8, the signal amplitude over the number ofCFRP layers covering the copper foil is shown. An exponential decayof amplitude such as expected by formula (2) was found. In thisspecial case the copper insertion was seen until the 12th layer. Thephase angle where the amplitude reaches a maximum shows alinear dependency until the 7th layer from the surface. For thedeeper regions between 7th and 14th layer the phased angle seemsto be constant or is slightly decreasing. This behavior was notinvestigated yet but was be reproduced in additional experiments.

It turns out that not only insertions of higher conductive ma-terials like copper, but also lower or non-conductive materialscould be found. A rectangular sandwich panel measuring300 � 210 mm with FOD inclusions was made consisting of an Ustiffener of Rohacell 110 WF-HT covered with 8 CFRP fabric layers

(rotated by 90�), The green line indicated the misaligned horizontal layer. (For inter-pretation of the references to colour in this figure caption, the reader is referred to theweb version of this article.)

http://www.youtube.com/watch?v=wNng6ClM1CM

http://www.youtube.com/watch?v=wNng6ClM1CM

Fig. 6. Analysis In-Plane waviness of Sample in Fig. 4 by Filtering in the Fourier Spaceand following inverse FFT.

Fig. 7. HF EC Image at 8.12 MHz (phase 50,5�) of CFRP with inserted Copper foils indifferent depths.


(977-2A-37%-3kHTA-5H-280-1200) and 4 additional UD layers(977-2-34-12KHTS-196-T1-150) at the U top. Four types of FODinclusions are inserted at different depths, all measuring10 � 10 mm: 1. copper mesh (M21/67%/ECF73 þ V12/900 mm), 2.Flashbreaker® tape 0.18 mm thick, 3. UD-backing paper (0.1 mmthick, ±130 g/m2) and a 0.13mm thick Polyethylene prepreg releasefoil. In Fig. 9 the position of the insertion an the U is illustrated.

The copper mesh was included as a reference and is stilldetected under 8 layers of CFRP fabric and 4 UD layers (in total 12layers). The detection depth of the tape, polyethylene foil and paperare 6 fabric layers and 3 UD layers respectively. On top of the Ustructure, this is increased to 7 CFRP fabric and 3 UD layersrespectively. The increased detection depth is probably due a

Fig. 8. Decay of signal amplitude with increasing depth of the inserted copper foil.

reduced amount of air inclusions on the U-top. In another experi-ment a CFRP plate with cut out wedges was produced. In Fig. 10a)the position of the cut outs are shown, in Fig.10b the correspondingHF EC image is shown respectively.

In the area of the cut out, the fiber orientation is disturbed butshaped CRFP parts often require a termination of hidden layersinside a multilayer stack.

Shaped CRFP parts often require a termination of layers inside amultilayer stack (so called “ply drops”). For quality assurance thenumber of the terminating layers must be determined. As shown inFig. 11, the area around ply drops has a characteristic electricalbehavior that allows an imaging of the terminating zone by a signalstrength variation.

4.3. Determination of local areal weight and carbon fiber volumecontent

Besides analyzing the texture, another quantitative parameter,the local areal weight, can be determined by HF EC. Particularly, fornon-woven fabrics such as fleece or recycled short fibers, thisparameter is more of interest for quality assurance. Basis weight isdefined by weight per area square typically given in g/m2 or gsm.The realized prototype system is able to non-destructively measurethe basis weight within a 25 mm diameter spot size. In comparisonto sensors used for textural analysis, the sensors for basis weightdetermination are using a comparably larger measurement spot,which allows quantitative measurement at a reasonably sized area.The advantage of using eddy current measurement is its ability forinline integration (no contact, no radiation) and the potential forcomprehensive data analyses. In order to determine properties ofan unknown specimen the system will be calibrated with knownreference samples first [16,17].

The base weight data can be displayed as single value or, whencaptured as an EC-scan, as a homogeneity image. Such an EC-scan isdepicted in Fig. 12, showing the uniformity of the sample. Whenusing a narrow tolerance or a filtering of the scale, minor in-homogeneities are revealed (Fig. 13).

As shown in Fig. 12 five A4-sized dry carbon fiber fleece sampleswere measured in the center. The darker the image appears thehigher the carbon fiber content within the measurement spot is.The fleece uniformity can be visualized by enhancing the contrastof a narrow EC value range (Fig. 13).

For fast inline applications a prototype of a multiple sensorsystems for base weight monitoring was realized (Fig. 14).

The system has been evaluated for grammages of 30 gsm up to2500 gsm. The achievable accuracy is better than þ/5%. In additionthe method can be used to determine the carbon fiber volumecontent of composites. Particularly, for chopped fibers in SMC orthermoset composites, the uniformity mapping shows areas oflarge accumulation of fibers or areas of low fiber density. In addi-tion, the general fibre alignment can be investigated too. Sensorwith oval shaped measurement spots are very sensitive to fibresaligned in the same direction. This focus of the sensor can be uti-lized to obtain more information on the dominant fiber orientationwithin the sensors.

4.4. Characterization of infiltration and curing

As explained in chapter 2, not only the conductive carbon fibresbut also the permittivity of the matrix material influences the eddycurrent measurement signal. This effect can be used for qualitycontrol of processes, where the complex permittivity of the com-posite changes [11]. So, potential applications could range fromcure monitoring (e.g. resin flow front detection, determining thedegree of polymerization) to the testing of consolidated CFRP with

Fig. 9. Schematic image FOD sample shows the type and position of different insertions with the corresponding RF EC image.

Fig. 10. HF EC Image of dry NCF plate with 5 UD Layers with cut out wedges in severaldepth. Sample Size 300 � 300 mm.

Fig. 11. HF EC Image from backside of NCF sample with 5 layers of which 3 are ter-minating. Sample Size 300 � 300 mm).


focus on infiltration or curing defects. A typical infiltration defect isthe occurrence of unimpregnated dry spots [14,15]. That means thatthere is not enough resin reaching a certain area of the CFRP part.Consequently, matrix material is lacking locally. Which result in adeviation of local permittivity and local conductivity, especially atthe contact points between roving's. Consequently, this defect canbe identified using eddy current, even when it occurs within thenon-visible layers. The dry spot sample in Fig. 15 was measured at13, 22 MHz and shows the transition zone between infiltrated andno filtrated areas as black contrast.

In addition to the ‘dry spots’ also ‘hot spots’ within a CFRPcomponent can be identified using eddy current technology(Fig. 16). Those defects can occur when component structuredoesn't allow an appropriate heat distribution, while dissipatingthe thermal energy that is generated during the curing reaction. Ifthe temperature of the spot stays bellow the glass transition tem-perature of the resin (at that specific state of cure) the curingprocess gets faster. If temperature climbs above that temperature,thermal damage of the matrix material is the consequence [11].

Even in complex CFRP structures the ‘hot spot’within thematrixcan be well separated from fiber related effects. The change of thecomplex eddy current signal due to conductivity variations (e.g.varying number of layers) has a different direction compared to thechange resulting from permittivity variations (Fig. 17). The differ-entiation towards edge and lift-off effects is more challenging andrequires further research work to be conducted.

Fig. 12. EC contrast function for Carbon fleece with different weight per area(100 � 100 mm).

Fig. 13. Visualization of Uniformity of fleece (100 � 100 mm) by image contrast enhancement (adaption of histogram).


5. Conclusion

In the decision matrix of NDT methods for CFRPs, the HF ECmethod can provide unique information when compared to otherNDE methods. CFRPs exhibit an electrical conductivity enabling theuse of electromagnetic testing techniques. In recent years, highresolution eddy current imaging technology has been substantiallydeveloped further. HF EC Imaging is a potential technology for in-spection of raw carbon fiber fabrics, infiltrated wet prepregs and

Fig. 14. Prototype EddyCus® Inline for Inline gramature determination by RF EC.

Fig. 15. Eddy current Image of CFRP sample with a ‘dry spot’ (Image Size150 � 150 mm).

consolidated CFRPs. HFEC based methods are interesting due to thesimplicity of machinery integration, allowing HF EC sensors to bedirectly integrated into the process chain without affecting thematerial properties and quality. Textural analyses and fault testingcan be performed with an image quality up to 500 mm resolution.Analyzing the quality of semi-finished products such as

Fig. 16. Photography and Eddy Current Image of 10 � 10 cm CFRP sample with a ‘hotspot’.

Fig. 17. eData of 10 � 10 cm CFRP sample with a ‘hot spot’ displayed in the complexplane.


reinforcement fabrics very early in the value chain can help to in-crease the yield of the production process and the safety of the finalproduct. The initial trials of FOD detection of a U stiffener in theprepreg stage proved to be successful up to 4 CFRP fabric layersdeep in this configuration. This opens up the potential of two in-spection intervals per stiffener in serial production. Small scaletrials with manually operated probes are required to validate shopfloor use and quantify real life inspection speeds. The potential toinspect terminating layers offers the possibility to integrate HF ECtechnique into a fully automated system for research on the pro-duction of airplane parts e.g. frames in the fuselage, produced in aRTM process at the DLR in Stade, Germany. The characterization ofpermittivity changes allows at the characterization of the curingprocess. In a long term vision the possibility of a non-destructivecharacterization of CFRPs textural properties may open the wayfor more accurate structural design and more consequent lightweight design. Experiments currently conducted have to show if HFEC can be used to monitor CFRP ageing by observing deviation ofdielectrical properties of the polymer matrix due to matrixdegradation.

Appendix A. Supplementary data

Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.compositesb.2015.03.022.

References

[1] Non-destructive testing e eddy current testing e general principles. 2008. ISO15549.

[2] Heuer H, Schulze MH, Meyendorf N. Non-destructive evaluation (NDE) ofcomposites: eddy current techniques. In: Karbhari VM, editor. Non-destruc-tive evaluation (NDE) of polymer matrix composites: techniques and appli-cations. Cambridge, UK, Philadelphia, PA: Woodhead Publishing; 2013.p. 33e55.

[3] Lange R, Mook G. Structural analysis of CFRP using eddy current methods. NDTE Int. 1994;27(5):241e8.

[4] Cheng J, Ji H, Qiu J, Takagi T, Uchimoto T, Hu N. Role of interlaminar interfaceon bulk conductivity and electrical anisotropy of CFRP laminates measured byeddy current method. NDT&E Int 2014;68:1e12.

[5] Cheng J, Ji h, Qiu J, Takagi T, Uchimoto T, Hu N, et al. Novel electromagneticmodeling approach of carbon fiber-reinforced polymer laminate for calcula-tion of eddy currents and eddy current testing signals. J Compos Mater March2015;49(5):617e31.

[6] Schulze M, Heuer H. High-resolution eddy current sensor system for qualityassessment of carbon fiber materials. Microsystems Technol 2010;16(5):791e7.

[7] Beine C, Boller C, Netzelmann U, Porsch F, Venkat R, Schulze MH, et al. NDT forCFRP aeronautical components a comparative study. In: Presented at the 2ndInternational Symposium on NDT in Aerospace, Hamburg, Germany; Nov.2010.

[8] .Hagemaier DJ. Eddy-current standard depth of penetration. Mater Eval1985;43(11):1438e41.

[9] IEEE standard letter designations for radar-frequency bands. 1984. IEEE Std521.

[10] International Telecommunication Union, Nomenclature of the frequency andwavelength Bands used in telecommunications, itu-t recommendation B.15.

[11] Gaebler S, Heuer H, Heinrich G. Measuring and imaging permittivity ofInsulators using high-frequency eddy current devices, Instrumentation andmeasurement. IEEE Transactions 2015;PP(99):1.

[12] Mook Gerhard, Lange Rolf, Koeser Ole. Non-destructive characterisation ofcarbon-fibre-reinforced plastics by means of eddy-currents. Compos SciTechnol May 2001;61(6):865e73.

[13] M.H. Schulze, H Heuer, Textural analyses of carbon fiber materials by 2D-FFTof complex images obtained by high frequency eddy current imaging (HF-ECI), SPIE Smart Structures and Materialsþ Nondestructive Evaluation andHealth Monitoring, 83470S-83470S-7.

[14] Gubernatis S, Balvers J-M, Weimer C. Concept development for inline Proc-essControl of the preform-LCM Production Chain. In: 4th International Sym-posium on NDT in Aerospace; 2012.

[15] Ehrenstein GW, Bittmann G, Hoffmann L. Duroplaste: aush€artung, prüfung,eigenschaften. Hanser; 1997. p. 14e6.

[16] García-Martín J, G�omez-Gil J, V�azquez-S�anchez E. Non-destructive techniquesbased on eddy current testing. Sensors 2011;11(12):2525e65.

[17] Hellier C. Handbook of nondestructive evaluation. New York: McGraw-Hill;2003. p. 8e15.



http://refhub.elsevier.com/S1359-8368(15)00141-9/sref1



























































lable at ScienceDirect

Composites Part B 96 (2016) 312e324

Contents lists avai

Composites Part B

journal homepage: www.elsevier .com/locate/compositesb

Automated detection of yarn orientation in 3D-draped carbon fiberfabrics and preforms from eddy current data

Georg Bardl a, *, Andreas Nocke a, Chokri Cherif a, Matthias Pooch b, Martin Schulze b,Henning Heuer b, c, Marko Schiller d, Richard Kupke e, Marcus Klein e

a Institute of Textile Machinery and High Performance Material Technology (ITM), TU Dresden, Dresden, Germanyb Fraunhofer Institute for Ceramic Technology and Systems, Material Diagnostic (IKTS-MD), Dresden, Germanyc Electronic Packaging Laboratory, Chair of Sensor Systems for Non-Destructive Testing, TU Dresden, Dresden, Germanyd HTS GmbH, Coswig, Germanye SURAGUS GmbH, Dresden, Germany

a r t i c l e i n f o

Article history:Received 28 January 2016Received in revised form24 March 2016Accepted 14 April 2016Available online 23 April 2016

Keywords:A. Carbon fibreA. PreformD. Non-destructive testingE. FormingHigh frequency eddy current testing

* Corresponding author. ITM TU Dresden, Hohe Str.Tel.: þ49 351 463 33766; fax: þ49 351 463 34026.

E-mail address: [email protected] (G. Ba

http://dx.doi.org/10.1016/j.compositesb.2016.04.0401359-8368/© 2016 The Author(s). Published by Elsev

a b s t r a c t

Ensuring the correct fiber orientation in draped textiles and 3D preforms is one of the current challengesin the production of carbon-fiber reinforced plastics (CFRP), especially in resin transfer molding (RTM).Small deviations in fiber angle during preforming have a considerable effect on the mechanical prop-erties of the final composite. Therefore, this paper presents an automated method for determining localyarn orientation in three-dimensionally draped, multi-layered fabrics. The draped fabric is scanned witha robot-guided high-frequency eddy current sensor to obtain an image of the sample's local conductivityand permittivity. From this image, the fiber orientation not only of the upper, but also of the lower,optically non-visible layers can be analyzed. A 2D Fast Fourier Transform is applied to local segments ofthe eddy current image to determine the local yarn orientation. Guidelines for processing the eddycurrent data, including phase rotation, filtering and evaluation segment size, are derived. For an intuitivevisualization and analysis of the determined yarn orientation, reference yarn paths are reconstructedfrom the determined yarn angles. The developed process can be applied to quality inspection, processdevelopment and the validation of forming simulation results.© 2016 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license

(http://creativecommons.org/licenses/by/4.0/).

1. Introduction

In the production of carbon fiber reinforced plastics (CFRP), oneof the current challenges is the quality assessment along the pro-cess chain. This is especially true for the resin transfer molding(RTM) process, which involves the steps of textile production,preforming and the injection of the resin in a closed mold. Thepreforming step, in which dry textiles are cut, stacked, draped to a3D shape and bonded, is critical for the quality of the final com-posite part. Any deviations in the fiber orientation can lead to adrastic loss of strength in the composite part. E.g., a fibermisalignment of 10� results in a loss of 30% or more in compressionstrength for unidirectional-reinforced composites [1e4]. It istherefore of great importance to ensure the correct fiber

6, 01069, Dresden, Germany.

rdl).

ier Ltd. This is an open access artic

orientation. Furthermore, forming defects like gaps and wrinklesneed to be eliminated.

Different methods have been applied for the detection of thefiber orientation in 3D-draped textiles and preforms. Most re-searchers and industrial users rely on optical methods [5e8], whichare fast and low-cost. However, optical methods only allow mea-surement of the uppermost fabric or preform layer, which isinsufficient for typical preforms consisting of 10 or more fabriclayers. Ultrasonic inspection, which is state of the art for the qualityassessment of the finished composite part, is in general not appli-cable to stacks of dry textiles, which lack a solid medium for wavepropagation [9,10]. Although air-coupled ultrasound techniquescan be operated on dry materials, the attainable resolution andcontrast are not sufficient for fiber orientation analysis [11,12].High-resolution thermography has been shown to be able to detectthe fiber orientation in draped textiles [13,14]. Challenges, in thiscase, are the limited penetration depth and the required techno-logical adaptions for high-resolution measurement.

le under the CC BY license (http://creativecommons.org/licenses/by/4.0/).

http://creativecommons.org/licenses/by/4.0/

mailto:[email protected]

http://crossmark.crossref.org/dialog/?doi=10.1016/j.compositesb.2016.04.040&domain=pdf

www.sciencedirect.com/science/journal/13598368

www.elsevier.com/locate/compositesb


http://creativecommons.org/licenses/by/4.0/



G. Bardl et al. / Composites Part B 96 (2016) 312e324 313

One technology with the potential to fill this current gap, andthereby provide an inspection technique for multi-layered textilestacks and preforms, is the high-frequency eddy current inspectiontechnique.Due to the considerably lower conductivityof carbonfibermaterials compared to metals, higher frequencies can be used toincrease sensitivity on carbon fibers (1e50 MHz, compared to0.1e1 MHz for the eddy current testing of aluminum). Although thefeasibility of eddy current testing for carbon fiber composites hadalready been shown in the 1970s [15,16], early research focused onthe detection of cracks in composites [17e23]. Only with the recentdevelopmentof high-resolution testingequipment, itwaspossible touse local changes in conductivity to trace individual yarns andthereby detect fabric and preform defects such as gaps, local wavi-ness and fiber orientation [24e27]. In recent years, high-resolution,high-frequency eddy current measurement has been successfullyapplied for theonline-inspectionofnon-crimp fabricproduction [28]and automated fiber placement [29], as well as in the online arealweight (grammature) testing of carbon fiber nonwovens [28,30].

Fig. 1 (left) shows an example eddy current scan for a quadraxial(þ45/90/�45/0) carbon fiber non-crimp fabric (four layers of ca.300g/m2areaweighteach).Ascanbeseen, yarns fromboth theupperand the three lower layers can be distinguished as stripe patterns.

By integrating a high-frequency eddy current measurementsystem with an industrial 6-axis robot, it was shown that the yarnorientation in three-dimensionally draped textiles and preformscan be visualized [28]. The result is a 3D map of the local electricalproperties of the sample, in which individual yarns from thedifferent layers can be distinguished (Fig. 1 right).

A dedicated inspection software, EddyEva by Suragus GmbH[31], is available for the automated analysis of eddy current scansfrom multi-layer carbon fabrics and preforms. Different layers can

Fig. 1. Eddy current scan results for a planar quadraxial non-crimp fabr

Fig. 2. EddyEva detection of local yarn positions and gaps for a biaxial fabric (le

be separated and quality parameters like yarn orientation, fiberdistribution, gaps and waviness can be automatically evaluated,with the option to check user-defined critical thresholds and togenerate automatic reports for each sample. The inspection can beapplied for the local analysis of both 2D and 3D scan results (Fig. 2).

However, no research has been reported so far on the globalanalysis of 3D eddy current scan results. From the perspective ofprocess development and quality assessment in RTM preforming,as well as for the validation of draping simulations, it would bedesirable to not only determine local yarn orientations for thewhole 3D surface automatically, but to use this local fiber orien-tation to reconstruct the actual yarn paths over the 3D surface. Thiswould finally allow for an easy comparison between differentsamples, e.g. with different draping parameters, and could serve asinput e.g. for infusion or structural simulation [32].

Therefore, this paper aims at developing such an automated,robust method for the extraction of local yarn angles from 3D eddycurrent data and for the reconstruction of local yarn paths. Thedetection of the yarn orientation will be done by a local Fouriertransform of projected segments of the measurement data; in ordertoprovideageneralbackground, theconceptofusingFourieranalysisfor the determination of fiber orientationwill be explained briefly.

Supplementary data related to this article can be found online athttp://dx.doi.org/10.1016/j.compositesb.2016.04.040.

2. 2D-Fast Fourier analysis for the evaluation of fiberorientation

As was shown above, high-frequency eddy current scanningcan be used to generate a conductivity map in which the indi-vidual yarn systems are visible as overlaid stripe patterns. For the

ic (left) and for a hemispherically draped biaxial non-crimp fabric.

ft and center), and of a critical gap width (right) in a 3D eddy current scan.


Table 1Material specification.

Fiber orientation [þ45/�45]

Areal weight 580 g/m2

Fiber material Toho Tenax HTS45 E23 800 texFiber spacing 2.76 mmStitch length 4 mm

G. Bardl et al. / Composites Part B 96 (2016) 312e324314

analysis of such periodic patterns, Fourier analysis been used in avariety of textile and related applications. It has been successfullyapplied for the detection of local fabric defects in weaves[33e44] and knitted fabrics [45]; the determination of fiberorientation in paper sheets [46], nonwovens [47,48] and biolog-ical cell networks [49]; and the measurement of fiber orientationin planar non-crimp fabrics, based on optical inspection [50]. Ithas also been used for the evaluation of 2D eddy current mea-surements [51,52].

In Fourier analysis, the 1D or 2D signal is decomposed into sinefunctions, which, when summed up with their respective fre-quency and amplitude, generate the original signal [33]. TheFourier transform, therefore, can be used to analyze the periodiccomponents of a signal. For an example, see the striped black-whitepattern in Fig. 3 (left). It can be represented as a single sine wavewith the respective angle (45�), spatial frequency (ca. 0.03 mm�1)and amplitude, which are shown in the Fourier transform (Fig. 3,right). Since the Fourier transform is symmetrical to the origin, allfrequencies are present twice, resulting in two red peaks for themain frequency. The angle between the vertical axis and the lineformed by these peaks represents the angle of the periodicity of thepattern. The stripe or “fiber” orientation is perpendicular to the axisof the periodicity and can thus be easily derived from the Fouriertransform.

When two perpendicular sine waves of the same frequency areoverlaid, the result will be a pattern as in Fig. 4, with the Fouriertransform showing both components at their respective angles(þ45� and �45� in this case).

Fig. 4. Two overlaid sine waves at þ45� and �45� (le

Fig. 3. Stripe pattern (left) and i

3. Experimental procedures

3.1. Tested material

All experiments were carried out with bidiagonal [þ45/�45]carbon fiber non-crimp fabrics. The textiles were manufactured atthe ITM, TU Dresden on a Karl Mayer Malimo 14024. Table 1 con-tains the material specifications.

3.2. Draping test stand

For the purpose of this research, a draping test system wasdesigned and manufactured by HTS Coswig GmbH in a jointresearch project (Fig. 5). The test system features a hemisphericalpunch which can be extended at different speeds and to setheights. Force and displacement are measured and logged to acontrol PC. For tensioning, the circular blank-holder is divided into8 blank-holder segments, whose forces can be adjustedindividually.

ft) and corresponding Fourier transform (right).

ts Fourier transform (right).

Binding Tricot

Fig. 5. Draping test system (schematic) (left), picture (right).


Table 2 shows the available options for the draping test stand,their limits and the selected values for the first experiment.

3.3. Robot-guided eddy current measurement system

After draping, the surface of the draped textile is scannedwith a high-frequency eddy current sensor mounted on an in-dustrial 6-axis robot (Fig. 6). Due to symmetry, a representativepart of the hemisphere was used for scanning. Within this area,parabolic paths are projected onto the surface, along which thesensor scans the draped multi-layered textile. For the sensor andthe measurement hard- and software, the commercially avail-able EddyCus® Integration Kit by Fraunhofer IKTS, Dresden, wasused (see Ref. [28] for further details). At defined time intervals,the complex impedance signal (conductivity and permittivity ofthe sample) is measured and recorded. The testing is donesimultaneously with four eddy current excitation frequencies

Table 2Experimental values for draping experiments.

DisplacementSpeedBlank holder force (individually for each of the 8 blank holders)

Fig. 6. Draped textile at punch extension height 100 mm (left), area used for scannin

(test frequencies). Best results were obtained for a test frequencyof 6 MHz, so only data for this frequency was used for evaluation.

Table 3 lists the scanning parameters.

4. Development of image analysis algorithm

4.1. Automatic phase rotation of eddy current data

The measurement provides a data set of ca. 60,000 complexelectrical impedance values (real and imaginary components) ofthe eddy current signal, with each value associated with a point in3D space. In order to visualize the data and to apply image analysisalgorithms to it, this complex (2D) signal value at each point has betransformed to a real (1D) signal which can be represented as a grayvalue.

Fig. 7 (left) plots the real and imaginary components of all datapoints. The real components spreadout between ca. 7000 and10,000digits, the imaginary components between 1000 and 5000 digits. By

Min Max Selected value

0 mm 100 mm 100 mm0.3 m/s 13.3 m/s 0.3 m/s0 96 N 0 N

g (center), scanning process by EddyCus® robot at Fraunhofer IKTS-MD (right).

Table 3Eddy current scanning parameters.

Robot Fanuc ARC Mate 120iC

Sensor- Coil configuration- Coil diameter- Shape of coil

S14148- “Half-transmission” (separated excitation and pick-up coil, both on same side of material)- 3.3 mm- Helical

Eddy current test frequency 6 MHzDistance between data points (average) 0.6 mmDistance between parallel paths (average) 0.5 mm

Fig. 7. Complex plot of eddy current scan data (left), visualization as gray values after projection of complex data (right).

Fig. 8. Phase rotation of complex impedance data.


projecting this point cloud onto the real axis (i.e., only taking the realcomponent), the complex 2Dsignal at eachdatapoint canbe reducedto 1D data, which can be interpreted as a gray value (with 7000digits ¼ white, 10,000 ¼ black). Fig. 7 (right) visualizes the result,where at each sample point in 3D space the gray value derived fromthemeasurement is plotted. Notice that the paths of individual yarnscan be distinguished, althoughwith low contrast. Also, yarns of bothlayers e both the upper þ45� and the lower �45� layer, whichwouldn't be accessible with optical methods e are visible.

The contrast between individual yarns in Fig. 7 can be consid-erably improved by applying an additional phase rotation to thecomplex data. What appears as different brightness between the(darker) left side and the (brighter) lower-right part of the surface isactually the result of a lift-off effect (a small air gap between thesensor and the fabric), which occurred in the lower-right partduring the scanning process. The lift-off effect and the electricalconductivity of the sample are the two dominating contrastmechanisms during eddy current scanning, with both contrast ef-fects occurring at different phase angles. The difference betweenthese phase angles is ca. 90�. Thus, by applying an appropriatephase rotation to the data set, the effect of the lift-off can bemapped mostly to the imaginary axis and the effect of the con-ductivity on the real axis. The absolute value of the required phaserotation angle will be influenced by parasitic effects from themeasurement instruments.

Fig. 8 shows the point cloud rotated by different phase angles.For the marked sample point, the phase angle is changed from a0 toa0 þ30� etc.

The rotated point clouds are again projected onto the real axis(only the real component is taken) and visualized on the 3D surface(Fig. 9). Rotations above 180� are not considered, since a rotation of180�þa only inverts the real and imaginary components comparedto a rotation of a.

Comparing the results of the different rotations, it can be foundthat the best contrast between individual yarns is achieved for aphase rotation of 60�. At this rotation, the lift-off influence is

considerably decreased, with the different brightness between theleft part and the right tip of the surface is removed.

In order to automatically determine the appropriate phaserotation for removing liftoff and compensating parasitic effects,it is assumed that the standard deviation caused by the lift-offis equal or higher compared to the standard deviation causedby the sample's electrical properties, which is a valid assump-tion for Fig. 7. Thus, when a regression line is calculated for thecomplex data set, the direction of the regression line e whichby definition is the direction of the highest standard deviatione will indicate the direction of the lift-off. By rotating the pointcloud so that this regression line stands vertically (Fig. 10), thelift-off effect will be mapped mostly to the imaginary axis andis therefore removed when the data is projected to the realaxis.

For the current sample, this procedure results in a calculatedlift-off angle of 62�, which is in good agreement with the analysis ofFig. 9 as discussed above. Furthermore, it was found that a phaserotation calculated by this procedure gives best detection results for

Fig. 9. Visualization after phase rotation.

Fig. 10. Procedure for automatic phase rotation.

Fig. 11. Evaluation points.


the other samples discussed in this paper, too. The automaticallycalculated phase rotation is therefore applied to the complexmeasurement data set prior to image processing.

4.2. Image segmentation, orthogonal projection and local Fouriertransform

In order to extract information about local yarn orientation, theimage is evaluated at individual points (centers marked with circles

in Fig. 11). In this chapter, section (2,2) will be used for the illus-tration of the algorithm. For the discussion of the algorithm pa-rameters, results from sections (2,2) and (5,2) will be compared(both marked red).

In a first step, the tangential plane is determined at theconsidered evaluation point and a normal projection of the whole


surface is carried out (Fig. 12, left). From this normal projection, asquare section of defined edge length around the center point istaken for evaluation (Fig. 12, right). Notice that the pattern in theprojected section generally resembles the pattern generated by twooverlaid sine waves in Fig. 4 above.

On this projected section, a 2D Fourier transform is applied inorder to carry out a frequency analysis (Fig. 13, left). In the Fouriertransform image, yellow and red colors correspond to higher en-ergy, which indicate a periodicity in the corresponding directionand at the corresponding frequency. The angle of these peaks, asmeasured form the vertical, thus indicates the direction of theperiodicity. Thus, the analyzed section shows periodic componentsat angles of 57�, 100� and 142� (Fig. 13, right).

4.3. Detection of yarn orientation

Since the yarn direction is perpendicular to the direction of thecorresponding frequency peak, the actual yarn angles ayarn can becalculated from the frequency angles afreq:

ayarn ¼ afreq � 90�

(1)

Fig. 12. Normal projection at point (2,2) (le

Fig. 13. Fourier transform of section (2,2) (lef

Table 4Determined angles from Fourier transform.

Frequency angle (in degree) Corresponding yarn angle (in degree)

57 �33100 10142 52

Table 4 shows the frequency and corresponding yarn anglesalong with the frequencies and wave lengths of the individualpeaks found at these angles. Note that several high-energy peakswith different frequency are present at the angles of 57� and 142�,indicating periodic information of different frequencies at theseangles.

In order to automatically detect the yarn angles from the Fouriertransform, the energy for all pixels at a certain angle is summed upand the frequency angles are converted to the actual yarn angles(Fig. 14, left). This energy distribution clearly shows two peakswhich correspond to the two yarn directions, as can be confirmedfrom Fig.14 (right). The third frequency peak at 10� is most likely anartifact from the measurement. The corresponding wave length ofca. 4 mm indicates that it might be related to the distance betweenthe excitation and pick-up coil in the sensor, which is roughly thesame length [28].

From Fig. 14 (left), an automatic detection of the fiber angles ispossible if the intervals in which the yarn directions are presumedare known. In this case, it is known that there are two yarn systemsand that the angles of these yarn systems are between 20 and 85and �20 and �85�, respectively. These intervals have been marked

ft), derived evaluation section (right).

t), angles with maximum energy (right).

Frequency (in 1/mm) Wave length (in mm)

0.11, 0.19, 0.33 8.9, 5.3, 3.00.25 4.10.18, 0.28 5.6, 3.5

Fig. 14. Energy distribution (left), visualization of fiber orientation with maximum energy (right).


with a dotted gray vertical line in the energy distribution. Themaximum within these intervals is taken to be the actual yarn di-rection. In future research, this procedure can be extended to pre-forms with multiple fabric layers, since each additional yarn systemwith a distinct angle will be represented as an additional peak inthe energy distribution.

It should be noted that in the 2D-Fourier transform (Fig. 13above), the main red and yellow energy peaks are not, as wouldbe expected, at the frequency of 0.36 mm�1 (which would corre-spond to the original yarn distance of ca. 2.8 mm), but at ca. half thisfrequency (corresponding to twice the yarn distance). The reasonfor this can be seen in Fig. 15. For both yarn directions, a grid hasbeen fitted to the image, with the yarn distances of 3.1 and 3.3 mmchosen so as to best match the white stripes (yarns) in both di-rections.When the underlying image is compared to this grid, it canbe seen that two neighboring yarns (white stripes) usually appearas “merged”, which might be caused by the distance betweensensor's excitation and pick-up coil, which is of the same magni-tude as the yarn distance. In the Fourier transform, the main peri-odicity thus appears at twice the yarn distance.

Fig. 15. Analysis of yarn distances.

5. Influence of algorithm parameters and filtering on theresults

5.1. Evaluation sections

The detection of the yarn orientation can be influenced by twofactors: the size of the evaluation segment, and an optional fre-quency filtering, which can improve the results. These influencesare discussed for sections (2,2), which had been used for the il-lustrations above, and section (5,2), which is one of the mostdifficult sections of the surface: the fabric is highly draped, while atthe same time the section features an edge and has reduced in-formation (yarn only in 50% of the area) (Fig. 16).

5.2. Filtering

High-pass filtering is a standard tool in image analysis for takingout low-frequency components, thus sharpening the image andhighlighting any edges. Therefore, different high-pass filters whichcorrespond to multiples of the yarn distance are tested on theimage (see Table 5).

Fig. 16. Normal projection for section (5,2).

Fig. 17. High-pass filter radii in Fourier transfor

Fig. 18. Energy distribution for different high-pas

Fig. 19. Calculated fiber orientation for different high

Table 5Filter frequencies and corresponding wave lengths.

Frequency Wave length

0.08 mm�1 12 mm0.11 mm�1 9 mm0.17 mm�1 6 mm0.33 mm�1 3 mm0 (no filtering) Infinite


Fig. 17 shows the Fourier transform for both sections. Notice thatthe peaks for section (5,2) are much wider, indicating that slightlydifferent fiber angles are present in the evaluation segment. Theminimum frequency for each filter is shown as circle. Whencalculating the energy distribution, only energy content outside ofthe filter radius is considered.

Fig. 18 shows the energy distribution of the two samples for thedifferent filter widths. The detected fiber directions are drawn intoFig. 19.

ms for section (2,2) (left) and (5,2) (right).

s filters, sample (2,2) (left) and (5,2) (right).

-pass filters, sample (2,2) (left) and (5,2) (right).


The results show that the detection of the fiber orientation forsection (2,2) is rather independent of the chosen filter. Small de-viations in the �31 … �35� angle are to be expected, since yarnswith slightly different orientations and slightly different yarn dis-tances are present in the evaluation section due to draping. Basedon the analysis of this section, all filter widths, including 0 mm�1

(no filtering) would be suited for the automated detection.For section (5,2), on the other hand, the different filters lead to

noticeable different results. Without filtering (fmin ¼ 0 mm�1, redcurve), the orientation of the yarn at ca. þ56� was not detected (themaximumof the redcurvewas foundat82.5). Thus, theapplicationofa filter is necessary in this case. Also, note that the filtering atfmin ¼ 0.33 mm�1 (magenta curve) accentuates the peaks at ca. ± 5�

and leads to ratherbig secondarypeakswithin thedetection intervals(at ca. 70� and �65�). The filter sizes of 0.08, 0.11 and 0.17 mm�1 alllead to reasonable results, but it can be seen that the blue and greenlines best match the orientation of the yarns in the evaluation seg-ment's center. Thus, a filter width of 0.11 mm�1 (blue line) is chosen,which corresponds to awave length of three times the yarn distance.

5.3. Influence of segment size

The size of the evaluation segment influences the detection re-sults. The bigger the size, themore yarns and thus periodical patternsare present, resulting in higher peaks in the 2D Fourier transform.However, since the surface is draped, the bigger the evaluation sec-tion, the higher the variety of yarn angles in the evaluation segmentwill be, leading to wider peaks. The determined angle will represent

Fig. 20. Visualization of different evaluation segmen

Fig. 21. Energy distribution for different evaluation se

only the average yarn orientation for a bigger section. It is, therefore,desirable to have a section as small as possible.

Fig. 20 shows the included area for different sizes of the eval-uation section; Fig. 21 shows the calculated energy distribution. Ahigh-pass filtering of 0.11 mm�1 was applied.

The comparison of the calculated yarn orientations with thescanning results can be seen in Fig. 22. For section (2,2), thedetected angles for widths 15, 20 and 30 mm are almost equal.However, a width of 40 mm leads to a slight change for both yarnangles, which is most likely due to a variation in yarn angle in theouter parts of the increased evaluation section.

For Section (5,2), it can be seen that a minimumwidth of 30 mmis necessary for proper detection of the yarn angles. Thus, a filter of0.11 mm�1 and a segment size of 30 mm (corresponding to roughlyten times the yarn distance) is chosen for further evaluation. Noticethat the distance between the evaluation points was chosen as20mm, so an evaluation segment width of 30mmwill lead to someoverlap, which however doesn't affect the calculation.

6. Visualization of the results

The described algorithm is applied to determine the local yarnorientation at all evaluation points (Fig. 23, left). From the pictures,it can be seen that the calculated yarn directions correctly show thedirections of the yarns in the fabric at all evaluation points.

At each evaluation point, the two determined yarn anglesrepresent the slope of the two yarn systems in the fabric. This in-formation can now be used to draw reference yarn paths onto thesurface, which provide an intuitive visualization:

t sizes for section (2,2) (left) and (5,2) (right).

gment sizes, sample (2,2) (left) and (5,2) (right).

Fig. 22. Calculated fiber orientation for different evaluation segment sizes, sample (2,2) (left) and (5,2) (right).

Fig. 23. Calculated yarn directions at evaluation points (left) and calculated reference yarn paths (right) for sample 1.


- From the known slopes at the evaluation points, the slope of theyarn can be calculated at any given point on the surface bymeans of 2D interpolation from the neighboring evaluationpoints.

- Starting from any given point on the surface, the path of a yarncan be reconstructed by numerical integration of the slope field:the yarn direction is determined at the starting point, and a smallstep is taken in this direction, arriving at a newpoint. The processis repeated until the path of this yarn is fully reconstructed.

Fig. 23 (right) shows the paths for a set of reference yarns for theevaluated sample. In order to limit the number of lines in the pic-ture, the distance between the starting points of the reference yarnswas set to 10mm. Each reference yarn thus represents several yarnsof the fabric. It can be seen that the reference yarns follow the pathsof the actual yarns very closely. This serves as the final validation ofthe developed algorithm.

The calculation and comparison of reference yarns can be usedto easily visualize the effect of different forming parameters on theyarn paths. Fig. 24 shows the detected yarn paths for two additionalsamples. In sample 2, the blank holder force of all eight blankholder segments was set to 48 N each, resulting in a total blankholder force of 348 N. In sample 3, a blank holder force of 96 N persegment (768 N total) was used.

Comparing the yarn paths on the surface shows the influence ofthe blank-holder force (Fig. 25). As expected, the displacement ofthe yarns by draping is highest for sample 1 (red) and restricted forsamples 2 and 3. Interestingly, the difference in yarn paths betweensamples 2 and 3 is much smaller than between samples 1 and 2.

7. Conclusions

In order to process 3D eddy current measurement data, an al-gorithm was developed that automatically determines and visual-izes the yarn orientation on the surface. It was shown that the yarnorientation can be derived from a 2D Fourier transform. Withproper choices for a high-pass filter and for the size of the evalu-ation section, it is possible to determine the yarn paths even inhighly draped border regions of the surface. By drawing referenceyarn paths onto the surface, the influence of different experimentalsettings on fiber orientation can be easily evaluated. The presentedevaluation method is not limited to eddy current measurementdata, but may be used for optical inspection of draped fabrics, too,although in this case only the yarn paths for the upper layer may bereconstructed.

The developed algorithm is not only of high usability for theevaluation of draping results, but can also be used for the validation

Fig. 24. Reference yarn paths for samples 2 (left) and 3 (right).

Fig. 25. Comparison of yarn paths for samples 1, 2 and 3.


of textile forming (draping) simulations. The simulation results,usually shear angles, can be processed to compute simulated yarnpaths, too, which can be compared to those derived from experi-mental results. Current research at the ITM is aimed at using thedetermined fiber orientation as input for infusion and structuralFEM simulation.

Acknowledgment

This research was funded by the European UnionRegional Development Fund (EFRE) and the Free State ofSaxony (grant “3D-Fast”, number 100224749). The authors wouldlike to thank the mentioned institutions for providing the funding.

References

[1] Hancox NL. The compression strength of unidirectional carbon fibre rein-forced plastic. J Mater Sci 1975;10:234e42. http://dx.doi.org/10.1007/BF00540347.

[2] Kawai M, Suda H. Effects of non-negative mean stress on the off-axis fatiguebehavior of unidirectional carbon/epoxy composites at room temperature. JComposMater2004;38:833e54.http://dx.doi.org/10.1177/0021998304042477.

[3] Erb T. Methodik zur Bewertung von Fehlern in Strukturbauteilen aus Faser-Kunststoff-Verbunden im Automobilbau. Aachen: Shaker; 2004.

[4] Yokozeki T, Ogasawara T, Ishikawa T. Nonlinear behavior and compressivestrength of unidirectional and multidirectional carbon fiber composite lami-nates. Compos Part Appl Sci Manuf 2006;37:2069e79. http://dx.doi.org/10.1016/j.compositesa.2005.12.004.

[5] Zhong T, Hu H. Formability of weft-knitted fabrics on a hemisphere. ComposPart Appl Sci Manuf 2007;8:245e51.

[6] Lomov SV, Boisse P, Deluycker E, Morestin F, Vanclooster K, Vandepitte D,et al. Full-field strain measurements in textile deformability studies. ComposPart Appl Sci Manuf 2008;39:1232e44. http://dx.doi.org/10.1016/j.compositesa.2007.09.014.

[7] Vanclooster K, Lomov SV, Verpoest I. Experimental validation of formingsimulations of fabric reinforced polymers using an unsymmetrical mouldconfiguration. Compos Part Appl Sci Manuf 2009;40:530e9. http://dx.doi.org/10.1016/j.compositesa.2009.02.005.

[8] Zambal S, Palfinger W, St€oger M, Eitzinger C. Accurate fibre orientationmeasurement for carbon fibre surfaces. Pattern Recognit 2015;48:3324e32.http://dx.doi.org/10.1016/j.patcog.2014.11.009.

http://dx.doi.org/10.1007/BF00540347

http://dx.doi.org/10.1007/BF00540347

http://dx.doi.org/10.1177/0021998304042477



http://dx.doi.org/10.1016/j.compositesa.2005.12.004









http://dx.doi.org/10.1016/j.patcog.2014.11.009


[9] Mahall K. Quality assessment of textiles: damage detection by microscopy.2nd ed. Berlin: Springer; 2003.

[10] Karbhari VM, editor. Non-destructive evaluation (NDE) of composites. Cam-bridge, UK; Philadelphia, PA: Woodhead Pub.; 2013.

[11] Blomme E, Bulcaen D, Declercq F, Lust P. Air-coupled ultrasonic evalua-tion of coated textiles 2002;1:757e60. http://dx.doi.org/10.1109/ULTSYM.2002.1193509.

[12] Imieli�nska K, Castaings M, Wojtyra R, Haras J, Clezio EL, Hosten B. Air-coupledultrasonic C-scan technique in impact response testing of carbon fibre andhybrid: glass, carbon and Kevlar/epoxy composites. J Mater Process Technol2004;157e158:513e22. http://dx.doi.org/10.1016/j.jmatprotec.2004.07.143.

[13] Schmidt T, Dutta S. In: Automation in production integrated NDT usingthermography. Proceedings; 2012.

[14] Schmidt T, Dutta S. In: CFRP manufacturing process chain observation bymeans of automated thermography. Proceedings, Singapore; 2013.

[15] Owston CN. Eddy current methods for the examination of carbon fibre rein-forced epoxy resins. Mater Eval 1976;34:237e44.

[16] Prakash R, Owston CN. Eddy-current method for the determination of lay-uporder in cross-plied CRFP laminates. Composites 1976;7:88e92. http://dx.doi.org/10.1016/0010-4361(76)90018-5.

[17] De Goeje MP, Wapenaar KED. Non-destructive insection of carbon fibre-reinforced plastics using eddy current methods. Composites 1992;23:147e57.

[18] Lange R, Mook G. Structural analysis of CFRP using eddy current methods.NDTE Int 1994;27:241e8. http://dx.doi.org/10.1115/1.4023253.

[19] Gros XE, Ogi K, Takahashi K. Eddy current, ultrasonic C-scan and scanningacoustic microscopy testing of delaminated quasi-isotropic CFRP materials: acase study. J Reinf Plast Compos 1998;17:389e405. http://dx.doi.org/10.1177/073168449801700502.

[20] Yin W, Withers PJ, Sharma U, Peyton AJ. Non-contact characterisation ofcarbon-fibre-reinforced plastics (CFRP) using multi-frequency Eddy currentsensors. In: IEEE Instrum. Meas. Technol. Conf. Proc. 2007 IMTC; 2007. p. 1e4.http://dx.doi.org/10.1109/IMTC.2007.379098.

[21] Yin W, Withers PJ, Sharma U, Peyton AJ. Noncontact characterization ofcarbon-fiber-reinforced plastics using multifrequency Eddy current sensors.IEEE Trans Instrum Meas 2009;58:738e43. http://dx.doi.org/10.1109/TIM.2008.2005072.

[22] Koyama K, Hoshikawa H, Kojima G. Eddy current nondestructive testing forcarbon fiber-reinforced composites. J Press Vessel Technol 2013;135:041501.http://dx.doi.org/10.1115/1.4023253.

[23] Salski B, Gwarek W, Korpas P, Reszewicz S, Chong AYB, Theodorakeas P, et al.Non-destructive testing of carbon-fibre-reinforced polymer materials with aradio-frequency inductive sensor. Compos Struct 2015;122:104e12. http://dx.doi.org/10.1016/j.compstruct.2014.11.056.

[24] Mook G, Lange R, Koeser O. Non-destructive characterisation of carbon-fibre-reinforced plastics by means of eddy-currents. Compos Sci Technol 2001;61:865e73. http://dx.doi.org/10.1016/S0266-3538(00)00164-0.

[25] Schulze MH, Heuer H, Küttner M, Meyendorf N. High-resolution eddy currentsensor system for quality assessment of carbon fiber materials. MicrosystTechnol 2010;16:791e7. http://dx.doi.org/10.1007/s00542-010-1047-3.

[26] Beine C, Boller C, Netzelmann U, Porsch F, Venkat R, Schulze M, et al. NDT forCFRP aeronautical compontents. A comparative study. In: Proceedings,Hamburg; 2010.

[27] Heuer H. Non-destructive evaluation (NDE) of composites by eddy currenttechniques. In: Karbhari VM, editor. Non-Destr. Eval. Polym. Matrix Compos.Tech. Appl. Cambridge, UK; Philadelphia, PA: Woodhead Pub.; 2013.

[28] Heuer H, Schulze M, Pooch M, G€abler S, Nocke A, Bardl G, et al. Review onquality assurance along the CFRP value chain e non-destructive testing offabrics, preforms and CFRP by HF radio wave techniques. Compos Part B Eng2015;77:494e501. http://dx.doi.org/10.1016/j.compositesb.2015.03.022.

[29] Schmidt C, Schultz C, Weber P, Denkena B. Evaluation of eddy current testingfor quality assurance and process monitoring of automated fiber placement.Compos Part B Eng 2014;56:109e16. http://dx.doi.org/10.1016/j.compositesb.2013.08.061.

[30] H€antzsche E, Kupke R. Inline monitoring of technical nonwovens e potentialsof contactless measurement systems for continuous property monitoring. In:Klein M, Cherif C, editors. CD-Rom Kurzref., Aachen; 2013.

[31] SURAGUS GmbH n.d. http://www.suragus.com [accessed 21.12.15].

[32] D€obrich O, Gereke T, Cherif C. Modelling of textile composite reinforcementson the micro-scale. Autex Res J 2014;14. http://dx.doi.org/10.2478/v10304-012-0047-z.

[33] Wood EJ. Applying fourier and associated transforms to pattern character-ization in textiles. Text Res J 1990;60:212e20. http://dx.doi.org/10.1177/004051759006000404.

[34] Jasper WJ, Potlapalli H. Image analysis of mispicks in woven fabric. Text Res J1995;65:683e92. http://dx.doi.org/10.1177/004051759506501109.

[35] Ravandi SAH, Toriumi K. Fourier transform analysis of plain weave fabricappearance. Text Res J 1995;65:676e83. http://dx.doi.org/10.1177/004051759506501108.

[36] Xu B. Identifying fabric structures with fast fourier transform techniques. TextRes J 1996;66:496e506. http://dx.doi.org/10.1177/004051759606600803.

[37] Tsai D-M, Hsieh C-Y. Automated surface inspection for directional textures.Image Vis Comput 1999;18:49e62. http://dx.doi.org/10.1016/S0262-8856(99)00009-8.

[38] Sakaguchi A, Kim H, Matsumoto Y-I, Toriumi K. Woven fabric quality evalu-ation using image analysis. Text Res J 2000;70:950e6. http://dx.doi.org/10.1177/004051750007001103.

[39] Sakaguchi A, Wen GH, Matsumoto Y, Toriumi K, Kim H. Image analysis ofwoven fabric surface irregularity. Text Res J 2001;71:666e71. http://dx.doi.org/10.1177/004051750107100803.

[40] Cardamone JM, Damert WC, Phillips JG, Marmer WN. Digital image analysisfor fabric assessment. Text Res J 2002;72:906e16. http://dx.doi.org/10.1177/004051750207201009.

[41] Arivazhagan S, Ganesan L, Bama S. Fault segmentation in fabric images usingGabor wavelet transform. Mach Vis Appl 2006;16:356e63. http://dx.doi.org/10.1007/s00138-005-0007-x.

[42] Ngan HYT, Pang GKH, Yung NHC. Automated fabric defect detectionda re-view. Image Vis Comput 2011;29:442e58. http://dx.doi.org/10.1016/j.imavis.2011.02.002.

[43] Wang X, Georganas ND, Petriu EM. Fabric texture analysis using computervision techniques. IEEE Trans InstrumMeas 2011;60:44e56. http://dx.doi.org/10.1109/TIM.2010.2069850.

[44] Swery EE, Allen T, Kelly P. Automated tool to determine geometric mea-surements of woven textiles using digital image analysis techniques. TextRes J 2015. http://dx.doi.org/10.1177/0040517515595031. 0040517515595031.

[45] Shady E, Gowayed Y, Abouiiana M, Youssef S, Pastore C. Detection and clas-sification of defects in knitted fabric structures. Text Res J 2006;76:295e300.http://dx.doi.org/10.1177/0040517506053906.

[46] Enomae T, Han Y-H, Isogai A. Nondestructive determination of fiber orienta-tion distribution of paper surface by image analysis. Nord Pulp Pap Res J2006;21:253e359.

[47] Ghassemieh E, Acar M, Versteeg H. Microstructural analysis of non-wovenfabrics using scanning electron microscopy and image processing. Part 1:development and verification of the methods. Proc Inst Mech Eng Part J MaterDes Appl 2002;216:199e207. http://dx.doi.org/10.1177/146442070221600305.

[48] Tun�ak M, Antoch J, Kula J, Chvojka J. Estimation of fiber system orientation fornonwoven and nanofibrous layers: local approach based on image analysis.Text Res J 2013. http://dx.doi.org/10.1177/0040517513509852. 0040517513509852.

[49] Sander EA, Barocas VH. Comparison of 2D fiber network orientation mea-surement methods. J Biomed Mater Res A 2009;88A:322e31. http://dx.doi.org/10.1002/jbm.a.31847.

[50] Schmitt R, Mersmann C, Sch€onberg A. Machine-Vision zur Messung derFaserorientierung bei der textilbasierten FVK-Herstellung machine-vision tomeasure the fiber orientation in textile-based FRP Production. Tm Tech Mess2010;77. http://dx.doi.org/10.1524/teme.2010.0056.

[51] Schulze MH, Meyendorf N, Heuer H. Analysis techniques for Eddy currentimaging of carbon fiber materials. Mater Test 2010;52:603e9. http://dx.doi.org/10.3139/120.110167.

[52] Schulze MH, Heuer H. Textural analyses of carbon fiber materials by 2D-FFT ofcomplex images obtained by high frequency eddy current imaging. Proc2012;8347. http://dx.doi.org/10.1117/12.914832. 83470S e 83470S.





http://dx.doi.org/10.1109/ULTSYM.2002.1193509

http://dx.doi.org/10.1109/ULTSYM.2002.1193509

http://dx.doi.org/10.1016/j.jmatprotec.2004.07.143








http://dx.doi.org/10.1016/0010-4361(76)90018-5

http://dx.doi.org/10.1016/0010-4361(76)90018-5




http://dx.doi.org/10.1115/1.4023253

http://dx.doi.org/10.1177/073168449801700502

http://dx.doi.org/10.1177/073168449801700502

http://dx.doi.org/10.1109/IMTC.2007.379098

http://dx.doi.org/10.1109/TIM.2008.2005072


http://dx.doi.org/10.1115/1.4023253

http://dx.doi.org/10.1016/j.compstruct.2014.11.056

http://dx.doi.org/10.1016/j.compstruct.2014.11.056

http://dx.doi.org/10.1016/S0266-3538(00)00164-0

http://dx.doi.org/10.1007/s00542-010-1047-3















http://www.suragus.com

http://dx.doi.org/10.2478/v10304-012-0047-z

http://dx.doi.org/10.2478/v10304-012-0047-z

http://dx.doi.org/10.1177/004051759006000404

http://dx.doi.org/10.1177/004051759006000404

http://dx.doi.org/10.1177/004051759506501109

http://dx.doi.org/10.1177/004051759506501108

http://dx.doi.org/10.1177/004051759506501108

http://dx.doi.org/10.1177/004051759606600803

http://dx.doi.org/10.1016/S0262-8856(99)00009-8

http://dx.doi.org/10.1016/S0262-8856(99)00009-8

http://dx.doi.org/10.1177/004051750007001103

http://dx.doi.org/10.1177/004051750007001103

http://dx.doi.org/10.1177/004051750107100803

http://dx.doi.org/10.1177/004051750107100803

http://dx.doi.org/10.1177/004051750207201009

http://dx.doi.org/10.1177/004051750207201009

http://dx.doi.org/10.1007/s00138-005-0007-x

http://dx.doi.org/10.1007/s00138-005-0007-x

http://dx.doi.org/10.1016/j.imavis.2011.02.002

http://dx.doi.org/10.1016/j.imavis.2011.02.002



http://dx.doi.org/10.1177/0040517515595031

http://dx.doi.org/10.1177/0040517506053906





http://dx.doi.org/10.1177/146442070221600305

http://dx.doi.org/10.1177/146442070221600305

http://dx.doi.org/10.1177/0040517513509852

http://dx.doi.org/10.1002/jbm.a.31847

http://dx.doi.org/10.1002/jbm.a.31847

http://dx.doi.org/10.1524/teme.2010.0056

http://dx.doi.org/10.3139/120.110167

http://dx.doi.org/10.3139/120.110167

http://dx.doi.org/10.1117/12.914832

8th International Symposium on NDT in Aerospace, November 3-5, 2016

NDT investigations on C/SiC samples from different manufacturing steps

Mykhailo KYRYCHENKO1, Susanne HILLMANN2, Frank MACHER2, Martin Schulze2, Henning HEUER1,2, Cesar CAMERINI3, Gabriela RIBEIRO PEREIRA3

1 Institute of Electronic Packaging Technology, Technische Universität Dresden; Dresden, Germany Phone: +49 351 462 42924, Fax: +49 351 463 37035; e-mail: [email protected]

2 Fraunhofer Institute for Ceramic Technologies and Systems IKTS; Dresden, Germany; e-mail: [email protected]

3 Laboratory of Nondestructive Testing, Corrosion and Welding of the Federal University of Rio de Janeiro (LNDC / UFRJ)

Abstract

This paper is focused on eddy current measurements of C/SiC samples in different status of the manufacturing process. The measurements are compared with results of X-Ray, Scanning Acoustic Microscope (SAM) and optical images and discussed. Keywords: C/SiC, eddy current, NDT, X-Ray, carbon fiber composite 1. Introduction Using of eddy currents is state of art for nondestructive testing of electrical conductive materials [1]. The developed high-frequency eddy current technology “EddyCus®” (with frequency ranges up to 100 MHz) made it even possible to extend the classical fields of application towards less conductive materials like CFRP [2]. More details to this technology can be found in [3], [4]. C/SiC is ceramic matrix composite (CMC) and belongs to a class of materials developed for aeronautics and space applications, in a domain where superalloys cannot be used anymore [5]. In comparison to metallic structures C/SiC-composites have many advantages as excellent thermal and mechanical properties by lower weight, high and stable friction coefficient, long life, low wear rate, and lower sensibility to surroundings and oxidation [6]. They have potential applications in structures (air intakes, structural panels with stiffners, etc.), in turbines and for brakes [5].

EddyCurrent (EC) testing was already performed on C/SiC samples in [7]. In this paper are provided NDT examinations for three C/SiC manufacturing process steps. The idea is to transfer the collected experience by investigations on CFRP to up-and-coming C/SiC. 2. Experimental Setup and Practice 2.1 Samples

The samples, shown in Figure 1 represent different steps of the manufacturing process of C/SiC and are numbered with 99, 100 and 101. These samples were provided by the working group of Ceramic Matrix Composites at the Universität Bayreuth.

Figure 1: Optical view of C-SiC specimens

Samples dimensions are presented in Table 1: Table 1: Dimensions of C/SiC samples

99 100 101 100,1*81,5*(3,3-4,1) mm3 99,7*81,1*(3-3,7) mm3 99,9*(81-81,7)*(3-3,3) mm3

Sample "99" is a pure CFRP (carbon fiber reinforced plastic). In "100" the resins were burned out, so carbon fibres and carbon particles remain. In “101” the cavities were filled with Si and after that the sample was sintered. Thus all samples have different states with different properties and different defects (if existent). 2.2 Nondestructive Test Methods

2.2.1 Optical evaluation

There was a digital photo performed. Displays in photos and surface optical review were evaluated and documented. The inconstant illumination intensity of the test and the subjectivity of the assessment must be mentioned as disadvantages.

2.2.2 Eddy currents

Eddy current testing is a well-established nondestructive method for the characterization of surfaces or materials by analyzing conductivity and permeability variations [3]. Alternating current passing the excitation coil induces a primary magnetic field. It excites eddy currents in a specimen. They create a secondary magnetic field that opposes the primary magnetic field. Changing the specimen material properties leads to changing the path of eddy currents and correspondingly complex signal changing at the pick-up coil. Measurements were done with EddyCus® MPECS – Multi Parameter Eddy Current Scanner. A semi-transmission coil with middle frequency 4 MHz was used. All eddy current images in this paper are done with 2.3 MHz.

Figure 2: EddyCus® MPECS

Figure 3: EddyCus® MPECS axes

Accordingly to axes on Figure 3, coil angles were defined: 0° along the x-axis, 90° along the y-axis and 45° was the position, as in the Table 2. Table 2: Coil angle definition

0° 45° 90°

Almost all eddy current measurements in this paper were done in contact-mode because of better resolution. To prevent wearout of coil, samples were coated with polyethylene film. By choosing (rotating) a Complex Phase Angle (CPA), noise signals can be filtered out.

2.2.3 X-Ray radiography

The use of X-rays for nondestructive testing is widely spread and offers numerous different methods for the characterization of chemical and structural properties of the samples. The X-ray radiation has wavelengths of about 10-6 m to 10-12 m and therefore belongs to the high-energy, ionizing radiation, which can penetrate a large number of substances. The simple X-ray transmission is most widely used in the NDT because it is simple and fast to perform. In the case of radioscopy, the structures of the sample are imaged as a function of their attenuation, with objects lying one behind the other in the beam direction overlaid. Defects within structure can thus be made visible [8]. For X-Ray radiography was used a phoenix nanomex (Figure 4). 2.2.4 Scanning acoustic microscopy (SAM)

Ultrasound refers to sound waves with frequencies above 20 kHz. It can be applied for material nondestructive investigations. An acoustic microscope uses the ultrasound propagation possibility in solids and liquids. An ultrasonic pulse is generated in a transducer using a piezoelectric material. The generated sound wave, focused via a lens, passes through the coupling medium water into a sample. The interaction at interfaces between different materials (including inclusions or defects) is investigated. Using the focusing lenses allows receive higher resolution as normal ultrasound testing. Scanning method makes it possible to get a two-dimensional image. A scanning acoustic microscope assembled at Fraunhofer IKTS-MD was used for measurements (Figure 5).

Figure 4: C/SiC specimen in X-Ray microscope Figure 5: Scanning acoustic microscope at

Fraunhofer IKTS-MD

3. Results Some of measuring effects are marked with *number* in text and with correspondent number in figure. 3.1 Sample 99

Sample 99 is a pure CFRP (carbon fiber reinforced plastic). The X-ray image on Figure 6 shows material particles that have a higher density *1* (dark spots). They could not be seen with other tested methods. Also regions with smaller densities can be observed *2* e.g. less filled. These areas were also seen visually on Figure 8 – “areas with hollows”. They were not visible in the eddy current images due to an edge effect. In the eddy current image on Figure 9 can be seen fiber structures. Noticeable spots are marked on the picture. Two dark spots *3* resemble two hollow spots in the photo. Other abnormalities couldn’t be captured with other methods. In the optical view can be seen slightly swollen areas. One of them is also to see on Figure 7 (EC-image). Different semi-transmission coil angles or other complex phase angle by eddy current testing can influence a result image. You might get new signals, as *4* on Figure 10. Different phase angles allow some signals to be displayed more clearly.

Figure 6: Sample 99, X-Ray image

Figure 7: Sample 99, EC image: 2.3 MHz; 90°; 180° CPA

Figure 8: Sample 99, optical view, front side

1

2

3



3.2 Sample 100

In Sample 100 the resins were burned out, so carbon fibres and carbon particles remain. With X-Ray radiography on Figure 11 can be seen similar indications as in sample 99. At the bottom right is marked a dark area *5*. It means greater weakening. In the optical view on Figure 13, this area looks as if metal was infiltrated there. The area is also visible on Figure 16 from the back side. It has to be discussed, why the area is not visible in the eddy current image. Some displays in the eddy current image on Figure 12 match optical view, some are new.




Figure 14: Sample 100, EC image:

2.3 MHz; 45°; 180° CPA


Figure 16: Sample 100, optical view, back side

Depending on the angle of the semi-transmission coil, different displays appear more clearly (compare Figure 12, Figure 14 and Figure 15). Optimization of the image contrast allows most favorable image for the eyes.

3 4

5

3.3 Sample 101

In “101” the cavities were filled with Si and after that the sample was sintered.




Figure 20: Sample 101, SAM image: 1 MHz; 20mm focal length

Figure 21: Sample 101, EC image, with Lift-off: 2.3 MHz; 45°;

180° CPA


Scratch, that visually can be seen on Figure 19, is also visible in X-Ray image on Figure 17. With eddy current it was not detected. It is supposed that big circular indication *6* in eddy current images (see Figure 18, Figure 21, Figure 22) is silicon infiltration border. This border is partly in SAM image on Figure 20 visible. In X-Ray image can be seen two parts: the darker in the middle *7* and ring-shaped brighter left *8*. The second part is barely visible. On Figure 18 at the bottom in the middle are circular measurements-artefacts *9* visible. Such artefacts were visible only by this sample in contact mode. By measurement with minimal Lift-off (Figure 21) they were not detected. This artefact comes from illustration of coil itself on surface irregularities. On Figure 22 marked displays come clearly. 4. Conclusion and Outlook C/SiC samples from three manufacturing steps were investigated with eddy current testing and compared with X-Ray, SAM und visual images of the samples. Eddy current is promising solution for material controlling during C/SiC manufacturing steps. Supposedly a silicon infiltration border was seen by sample 101.

6

7

8

9

Direction of semi-transmission coil has big influence on defects detectability with eddy current testing. In future are planned systematic defect inductions during production of the same sample and evaluations with nondestructive testing. In this way has to be proven suitability of NDT-methods for detection of specific defects. Acknowledgements We would like to thank the working group of Ceramic Matrix Composites at the Universität Bayreuth, in particular Dr. Langhof, for providing the examined samples. References

1. 'Non-destructive testing – Eddy current testing – General principles', ISO 15549, 2008 2. M. Schulze, H.Heuer, 'Textural analyses of carbon fiber materials by 2D-FFT of

complex images obtained by high frequency eddy current imaging', Nondestructive Characterization for Composite Materials, Aerospace Engineering, Civil Infrastructure, and Homeland Security, SPIE Proceedings Vol. 8347, edited by A.L. Gyenyesi, pp 18-25, 2012

3. H. Heuer, M. Schulze, M. Pooch, S. Gäbler, A. Nocke, G. Bardl, Ch. Cherif, M. Klein, R. kupke, R. Vetter, F. Lenz, M. Kliem, C. Bülow, J. Goyavaerts, T. Mayer, S. Petrenz, 'Review on Quality Assurance along the CFRP Value Chain – Nondestructive Testing of Fabrics, Preforms and CFRP by HF Radio Wave Techniques', Elsevier, Composites Part B Engineering 27, compositesb.2015.03.022, March 2015

4. G. Bardl, A. Nocke, C. Cherif, M. Pooch, M. Schulze, H. Heuer, M. Schiller, R. Kupke, M. Klein, 'Automated detection of yarn orientation in 3D-draped carbon fiber fabrics and preforms from eddy current data', Elsevier, Composites Part B Engineering 96, pp 312-324, 2016

5. G. Boitier, S. Darzens, J.-L. Chermant, J. Vicens, 'Microstructural investigation of interfaces in CMCs', Elsevier, Composites: Part A 33, pp 1467–1470, 2002

6. Shangwu Fan, Litong Zhang, Yongdong Xu, Laifei Cheng, Jianjun Lou, Junzhan Zhang, Lin Yu, 'Microstructure and properties of 3D needle-punched carbon/silicon carbide brake materials', Elsevier, Composites Science and Technology 67, pp 2390–2398, 2007

7. V. K. Srivastava, A. Udoh, H.-P. Maier, P. Knoch, K. Maile, 'Eddy current nondestructive mapping of C/C–SiC composites', Springer-Verlag, Forschung im Ingenieurwesen 68, pp 169 – 172, 2004

8. K.-J. Wolter, M. Bieberle, H. Budzier, G. Gerlach, T. Zerna, 'Zerstörungsfreie Prüfung elektronischer Baugruppen mittels bildgebender Verfahren', Verlag Dr. Markus A. Detert, 2012

Simone Gäbler1, 2, a), Henning Heuer2, b), Gert Heinrich1, Richard Kupke3, c)

1Leibniz Institute of Polymer Research Dresden, Hohe Straße 6, 01069 Dresden, Germany. 2Fraunhofer IKTS-MD, Maria-Reiche-Str. 2, 01109 Dresden, Germany.

3Suragus GmbH, Maria-Reiche-Str. 1, 01109 Dresden, Germany.

a)Corresponding author: [email protected] b)[email protected]

c)[email protected]

Eddy current testing is well-established for non-destructive characterization of electrical conductive materials. The development of high-frequency eddy current technology (with frequency ranges up to 100 MHz) made it even possible to extend the classical fields of application towards less conductive materials like CFRP. Maxwell’s equations and recent research show that the use of high-frequency eddy current technology is also suitable for non-conductive materials. In that case the change of complex impedance of the probing coil contains information on sample permittivity. This paper shows that even a quantitative measurement of complex permittivity with high-frequency eddy current device technology is possible using an appropriate calibration. Measurement accuracy is comparable to commercial capacitive dielectric analyzers. If the sample material is electrically conductive, both, permittivity and conductivity influence the complex impedance measured with high-frequency eddy current devices. Depending on the measurement setup and the sheet resistance of the sample a parallel characterization of both parameters is possible on isotropic multi-layer materials. On CFRP the permittivity measurement is much more complex due to the capacitive effects between the carbon rovings. However, first results show that at least the local permittivity variations (like those caused by thermal damages) are detectable.

Eddy current testing is a well-established electro-magnetic approach for non-destructive characterization of electrical conductivity or magnetic permeability related sample properties [1]. Typical fields of application are crack detection and surface characterization on metals as well as material differentiation of conductive materials where measurements are usually conducted at the low kHz frequency range [2]. To improve the sensitivity of eddy current testing on low-conductive materials, the high-frequency eddy current system “EddyCus®” was developed. This technology is operating between 100 kHz and 100 MHz, so almost up to the low microwave frequencies [3].

By increasing the measurement frequency, the induced voltage Uind in the sample as well as in the receiving coil rises (1), resulting in an improved measurement sensitivity, especially for low conductive materials or very thin conductive films [4].

dtdUind (1)

. On the other hand the standard penetration depth decreases with increasing measurement frequency. However,

when high-frequency eddy current is applied to materials with low electrical conductivity , that effect gets partially compensated (2). Then the probe size might limit the real penetration depth [5], [6].

41st Annual Review of Progress in Quantitative Nondestructive EvaluationAIP Conf. Proc. 1650, 336-344 (2015); doi: 10.1063/1.4914628

© 2015 AIP Publishing LLC 978-0-7354-1292-7/$30.00

336This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:153.96.242.65 On: Mon, 27 Apr 2015 11:13:15

/2 (2) Besides sheet resistivity measurements the EddyCus technology is currently mainly used for carbon fiber testing.

There, applications range from texture analysis (Fig. 1) and defect detection to local areal weight determination [7]. Especially the use on dry carbon textiles as well as a comparable good penetration depth of up the 8mm is seen as very valuable [8], [9].

High-frequency eddy current scan of CFRP and image processing for separation of differently oriented plies.

In addition to the two already mentioned implications of the high measurement frequency in the MHz range, there is a third important effect, especially if the sensor is applied to dielectric materials. The rotating electric field E, which is created by the time-varying magnetic field H of the coil (3), causes significantly higher displacement currents D/ t than those arising in an eddy current measurement setup operating in the kHz range, due to the higher measurement frequency (4).

tBE (3)

ttEEDJH (4)

That effect can be used to measure permittivity related sample properties with high-frequency eddy current

devices, especially on electrically low-conductive or insulating materials (Fig. 2). Previous FEM simulation results support this hypothesis and first experiments have proved that qualitative permittivity measurement on insulating materials is possible (cure monitoring of the epoxy resin L20 and high-spatial resolution permittivity mapping) [10].

(a) (b)

Photography (a) and high-frequency eddy current image (b) of a dielectric Polyoxymethylen (POM) sample with holes, displayed detail 90mm x 25mm.

This paper is intended to show additional possible applications of high-frequency eddy current permittivity

characterization. The focus will be on quantitative permittivity characterization of insulting materials as well as on permittivity measurement of low-conductive materials.

337 This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:153.96.242.65 On: Mon, 27 Apr 2015 11:13:15

The prevalent approach for quantitative permittivity measurement on insulating materials is the dielectric analysis using a capacitive setup with parallel-plate electrodes. If appropriately calibrated, a high measurement accuracy of 0.1 % regarding magnitude and 0.06° regarding phase can be achieved [11]. However, a good electrical contact between electrodes and sample is essential [12], so sample preparation is quite complex. As high-frequency eddy current measurement does not require any special sample preparation [13], there might be cases where it would be preferred compared to the capacitive technology. Hence the potential of quantitative high-frequency eddy current measurement in comparison to a capacitive setup is examined within the following sections.

Experimental Setup and Sample Preparation

One initial measurement and one later measurement of the complex permittivity were conducted on seven different, fully cured resin samples (Table 1) using two commercially available measurement systems.

Overview on resin systems used for sample preparation.

L L 100:40 25° C L20 EPH 161 100:25 25° C, 10 h 60° C Post-Curing L1100 EPH 294 100:30 400 min 25° C, 10 h 70° C MP MP 100:40 8h 100° C RIM 135 RIM 137 100:30 1 h 60° C, 6 h 80° C RIM 935 RIM 936 100:29 25° C, 10 h 95° C Post-Curing EPR 600 (Resin 496) 385-blue 100:80 1 h 90° C, 4 h 160° C

For the capacitive measurement (Fig. 3) the Novocontrol Alpha AN analyzer was used, in combination with the

4-wire impedance interface ZG4 (operated in 2-wire mode), the passive sample cell BDS 1200 and the software WinDETA V5.62. To reach high measurement accuracy, sample capacity is required to be between 50 pF and 200 pF [11]. Under consideration of available electrodes and limits of mechanical processing sample dimensions were set to 31 mm diameter and 0.4 mm thickness. Therefore 2 mm thick resin plates were locally milled down to 0.4 mm and disks of 31 mm diameter were punched. Afterwards the thickness of each sample was measured at several points using a micrometer screw and the average thickness of each sample was calculated. Then a 60 nm gold layer was sputtered on both sides of the samples using the Balzers SCD 050 Sputter Coater. From each resin system one sample was produced, despite resin EPR 600 where two samples were made in order to test the influence of sample preparation on the measurement results. Before the actual measurement was done, an “All calibration” was carried out for the impedance interface ZG4 in order to compensate for long term drift. Additionally, a low impedance “Load Short calibration” was carried out to account for the properties of the connection lines. For the actual measurement the samples were placed between two gold plated electrodes of 30 mm diameter before being mounted in the sample cell [11]. The sample diameter and thickness was given to the WinDETA software for each sample and complex permittivity values at 6 MHz were acquired.

(a) (b) (c) (d) (e)

Sample preparation and measurement setup for capacitive setup: (a) Sputter Coater, (b) gold plated samples, (c) Novocontrol dielectric analyzer, (d) Sample cell, (e) overview measurement setup


For the high-frequency eddy current measurement (Fig. 4) the Suragus EddyCus® CF map 4040 was used, combined with the sensor SURA RD-P020B. This probe was designed for permittivity measurement and has an outer diameter of 24 mm, a coil diameter of 2 mm and the highest sensitivity at 6 MHz. The samples consisted of seven resin plates (100 mm x 80 mm x 4 mm). They were placed, one by one, at the same location of the scan area. The measurement consisted of a 25 mm x 35 mm scan, obtained in slight contact at 6 MHz, 30 dB. With each sample a 4 mm thick PMMA plate was scanned to allow for short term drift correction. The measurements were repeated. The actual high-frequency eddy current measurement value used for permittivity characterization was the drift compensated and averaged data from both scans. Phase rotation was set to 16°, which is a previously determined value for this particular sensor/frequency combination, so that the real part of complex permittivity ’ is influencing the imaginary part of complex impedance and imaginary part of permittivity ’’ is influencing the real part of complex impedance. For calibration purpose complex permittivity values of the resins L1100 and MP, as obtained by the capacitive measurement, were used: the imaginary part of complex impedance was calibrated to a permittivity ’ of 3.69 (L1100) and 4.70 (MP); the real part of complex permittivity to a dielectric loss ’’ of 0.15 (L1100) and 0.26 (MP). A linear change of complex impedance due to sample permittivity variation is assumed, based on previously reported FEM simulation results [10].

(a) (b) (c)

Setup for high-frequency eddy current measurement (a) 4 mm thick polymer samples, (b) EddyCUS CF map 4040, (c) Parallel PMMA scanning for short term drift compensation

Measurement Results and Discussion

The results of the capacitive and the high-frequency eddy current permittivity measurement are shown in Fig. 5. It can be seen that both approaches are delivering very similar values. Looking at the real part of complex permittivity ’ there is only one resin, L20, with a significant difference of the values acquired by capacitive and inductive measurement. For the imaginary part of complex permittivity ’’ this applies for RIM 135. All other values show a very good agreement.

Result of initial and reproduced capacitive and high-frequency eddy current permittivity measurement.

For evaluating the accuracy of the measurements two factors are investigated: The reproducibility of

measurement and the influence of sample preparation on measurement result.


To compare the reproducibility of both measurement approaches, all samples (despite those used for calibration) were measured a second time a couple of weeks later. The changes of the results compared to the initial measurement are indicated by the error bars in Fig. 5. For the capacitive setup the average difference between both measurements was 0.034 regarding ’ and 0.01 regarding ’’. For the high-frequency eddy current measurement the average difference regarding ’ was similar with 0.034 and a little bit higher regarding ’’ with 0.022. So the capacitive setup shows a slightly better reproducibility. One reason might be the better protection of the capacitive samples against environmental influences (e.g. humidity) due to their gold coating.

However, the sample preparation for the capacitive permittivity measurement (see Fig. 3 b) is critical. Here, in addition to a good electrical contact between sample and electrodes, especially a constant sample thickness is essential [11]. The measured sample capacity Cs depends not only on the complex sample permittivity and the sample area A but also on the sample thickness d (5). To reach the maximum measurement accuracy of 1%, only a thickness deviation of less than a 1 m is accepted for a 0.1 mm thick sample.

d

ACS0 (5)

To practically investigate the influence of sample preparation on the capacitive measurement results, we

prepared two samples of the resin EPR 600. Measurement results are reported in table 2. It concluded that the measurement difference in the real part of permittivity is 0.04 (~1%) and in the imaginary part it is 0.02 (~10%).

Influence of sample preparation on measurement result of capacitive permittivity measurement.

ermittivity ’ ermittivity ’’0.4-0.45, Ø 0.425 4.09 0.15 0.42-0.46, Ø 0.44 4.13 0.17

The deviation of the capacitive measurement result caused by sample preparation is almost the difference that

was observed between capacitive and inductive measurement for ’ of the resin L20 and ’’ of RIM 135. Hence sample preparation might possibly explain that observation.

For the permittivity measurement with high-frequency eddy current devices there is no special sample preparation such as gold coating, necessary. This suggests a negligible influence on measurement accuracy. However, it is also important that the thickness is constant, if the sample thickness is smaller than three times the standard penetration depth [13]. Albeit, the tolerated thickness deviations are usually a lot higher compared to the capacitive setup. In general the measurement accuracy of inductive sensing on insulating materials increases with increasing sample thickness, as there is ample material that interacts with the magnetic field. Subsequently, for a sample of a couple of millimeter thickness, a thickness deviation of some m is tolerable.

Eddy current systems allow the measurement of the thickness of a dielectric layer on conductive substrates. Therefore measurements are made in contact with the sample surface, so that a varying thickness of the dielectric layer causes a change in the lift-off between the eddy current probe and the conductive substrate [2]. However, this only allows the measurement of the thickness of a dielectric layer, but not of its permittivity. In addition it has the disadvantage that to require contact to the sample. The following section shows that those drawbacks can be overcome by the use of high-frequency eddy current. Here even a parallel measurement of conductivity and permittivity related properties of multi-layered systems is possible within the examined limits.


For high-frequency eddy current measurement the commercially available EddyCus® CF map 4040 and the sensor SURA RD-P020B were used again. The samples for the multilayer experiments consisted of 180µm PET


foils of 100 mm x 100 mm coated with thin copper films. Following samples were prepared: no copper film as reference; copper films with 4.3 / ; 0.4 / .; 0.144 / and 0.055 / . They were covered by a sheet of paper and were measured solely or in a stack with plates of MP or L1100 resin (100 mm x 80 mm x 4 mm). Measurement was carried out at 6 MHz, 28 dB, with a constant lift-off of 4.5 mm to the copper coated foil (Fig. 6). An area of 40 mm x 40 mm was scanned in the center of the sample and the average value was calculated. The phase was rotated to a previously determined value, so that an increase in the real part of permittivity ’ of a non-conductive sample only influences the imaginary part of complex impedance. The measurement was repeated three times at different days and reported similar results each time.

(a) (b) (c)

Experimental setup: (a) stacked layers, (b) Copper coated PET samples, (c) photography during measurement.


The results of the measurement are shown in Fig. 7. The data suggests that the change of complex impedance Z due to a growing sample permittivity (air, L1100 or MP resin) is almost independent from the sheet resistance of the copper coated sample. With growing permittivity imaginary and real parts of the complex impedance are declining. With increasing sheet resistance the change of the imaginary part of complex impedance is slightly bigger, as oppose to the difference in the real part, which decreases marginally. An exception is the measurement on the PET foil without a copper film. There the effect on complex impedance caused by a permittivity change is significantly smaller.

The change of complex impedance due to the differing sheet resistance follows the typical thickness curve [13]: a decreasing sheet resistance first leads to a shrinking real part of complex impedance whereas it rises again when sheet resistance falls below a certain value.

Three implications can be drawn from the described measurement results for parallel permittivity and sheet resistance measurement of multi-layer sample using high-frequency eddy current technology measurement:

(1) The permittivity measurement with high-frequency eddy current seems to be more sensitive if a conductive material is placed below the dielectric material of investigation. A hypothetical explanation might be a kind of focusing effect for the electromagnetic field of the coil. It might be attracted by the conductive layer and therefore has to pass through the dielectric layer whereas it spreads out in all directions when there is no conductive material nearby.

(2) A change in the complex impedance can only be linked towards a changing permittivity or conductivity of the sample within certain limits of sample properties. Due to the characteristic spiral curve caused by thickness or sheet resistance variations, the authors conclude that permittivity and conductivity related effects are not always orthogonal to each other as assumed before [4]. Depending on the resistance of the sample, the curves of the two parameters might well point in the same direction (e.g. for the 0.144 / sample within the used measurement setup). Only if the sample properties are within a certain range of sheet resistance, away from the measurement setup specific turning point of the thickness curve (e.g. k / to 0.4 / here), then a parallel determination of the real part of permittivity and the sheet resistance is possible. However, in this case an increasing dielectric loss still points in the same direction as a decreasing sheet resistivity. If the targeted accuracy for sheet resistance requires a separation of both effects, a multi-frequency measurement could be considered. At a higher frequency an impedance change due to permittivity variation may be much bigger than a conductivity related effect (4), [10].


Result of the high-frequency eddy current permittivity and sheet resistance measurement on multi-layer samples. (3) When the sheet resistance of an isotropic sample is known, real and imaginary part of permittivity can be

determined easily using high-frequency eddy current measurement technology. Depending on the desired accuracy a “global” permittivity calibration (not considering the specific sheet resistance) might be possible.

Measuring permittivity of an anisotropic CFRP is much more complex than permittivity characterization on isotropic materials. The reason lies within the capacitive structure that the carbon rovings are forming within the material [14]. As a result locally strengthened electric fields occur and affect the eddy current signal. Subsequently quantitative permittivity measurement using high-frequency eddy current still remains part of further studies and will be subject to a future publication. Nevertheless, a first example shall be claimed here: the permittivity of the resin used in a CFRP has remarkable influence on the eddy current signal. Therefore the resin was locally damaged by hot air and a high-frequency eddy current image was acquired afterwards.


The CFRP samples were initially designed for another research work [15]: Woven fabrics (Hexcel Prepreg M18-1/G939) which amongst others are used in military helicopters, were stacked to quasi-isotropic plates ((0/+45/90/-45)3)s of 250 mm x 250 mm and were consolidated in the autoclave. The resulting CFRP plates are 6 mm thick at a fiber volume content of 43%. The center was locally overheated at 400°C hot air for 90 minutes. The intention was to specifically damage the resin, but not to the carbon fiber (for more details see [15]). Within an oxidizing environment fiber degradation only starts above 400° C [16]. After the overheating the plate was inspected with several NDT methods (visually, acoustically, by ultrasonic and with IR thermography). In addition samples for microscopic examinations were taken from one half of the sample. The second half was characterized by high-frequency eddy current. Again, the EddyCus® CF map 4040 and the sensor SURA RD-P020B were used. Measurement was done at a 1 x 1 mm pitch in slight contact at 6 MHz, 28dB.


Figure 8 shows the resulting high-frequency eddy current image (100 mm x 220 mm). The phase was rotated in a way, that the real part of complex impedance contains mainly fiber texture information (conductivity related) and a minimum of information about the thermal damage, which was assumed to affect mainly the resin. Thus, the imaginary part of complex impedance contains much more information on the degradation of the resin.


Nevertheless, it cannot be assumed that conductivity and permittivity related effects are behaving exactly orthogonal to each other (see previous section).

Photography and high-frequency eddy current image of thermally damaged CFRP. The real part of the high-frequency eddy current image shows a white spot in the center of the damaged area.

This spot has a lower conductivity than the rest of sample which is caused by a delamination in the material. Besides the eddy current testing this delamination was also visible in the microscopic examination of the cross section and was captured by all other NDE methods used at the sample [15].

The imaginary part of the high-frequency eddy current image shows a darker and a lighter area around the local delamination indicating a change of permittivity of the resin. In fact, it matches the conspicuously, discolored area shown in photography aside. This is a remarkable result, as none of the other NDT approaches was able to detect a damage of the same dimension as the visual inspection indicates [15]. These methods were solely sensitive to the primarily delaminated area, but unlike eddy current technology failed to reveal the optically discolored area and damaged area.

Summing up, it can be said, that high-frequency eddy current is a novel and equally reliable alternative for quantitative permittivity measurement. On insulating materials measurement results and reproducibility are comparable with the prevalent approach of capacitive parallel plate measurement. There, the advantage of inductive sensing is the much easier sample preparation and the higher resulting measurement accuracy as it is less vulnerable to non-optimal sample properties. Experiments on multi-layer materials indicate that the accuracy of the high-frequency eddy current permittivity measurement could be further improved if a conductive material is situated below the dielectric layer. In that case both, permittivity and conductivity influence the complex impedance measured with high-frequency eddy current devices. If the sheet resistance of the sample is known, the complex permittivity can be determined using high-frequency eddy current measurement technology. However, if both parameters are unknown a parallel measurement of permittivity and sheet resistivity of the sample is only suitable when the sample properties are within certain limits. On anisotropic materials like CFRP a quantitative permittivity


measurement with high-frequency eddy current is even more difficult. Nevertheless, it could be qualitatively shown that the approach is much more sensitive regarding thermal damages of the resin than other non-destructive approaches. As consequence this will encourage the authors to further investigate the full potential of high-frequency eddy current technology as means to characterize dielectric properties of CFRP, in addition to the known application for electrically conductive attributes.

The authors want to thank the Dresden University of Technology for the financial support of this research (Programm zur Förderung von Nachwuchswissenschaftlerinnen) and Mr. Andreas Floet for making his thermally damaged CFRP samples temporarily available for high-frequency eddy current measurement.

1. Non-destructive testing – Eddy current testing – General principles, ISO 15549, 2008. 2. D. Stegemann, Der Einsatz von Wirbelstromströmen für die Zerstörungsfreie Werkstoffprüfung (DGZfP,

Berlin, 2010). 3. M. Schulze and H. Heuer, “Textural analyses of carbon fiber materials by 2D-FFT of complex images obtained

by high frequency eddy current imaging”, in Nondestructive Characterization for Composite Materials, Aerospace Engineering, Civil Infrastructure, and Homeland Security 2012, SPIE Proceedings Vol. 8347, edited by A. L. Gyekenyesi (SPIE, 2012), pp. 18-25.

4. H. Heuer, M. H. Schulze, and N. Meyendorf, “Non-destructive evaluation (NDE) of composites: eddy current techniques,” in Non-destructive evaluation (NDE) of polymer matrix composites: Techniques and applications, edited by V. M. Karbhari (Woodhead Publishing, Cambridge, UK, Philadelphia, 2013), pp. 33–55.

5. D.J. Hagemaier, Materials Evaluation (11), 1438-1441 (1985). 6. Z. Mottl, NDT International (1), 11-18 (1990). 7. www.carbon-fiber-testing.com 8. M. Schulze and H. Heuer, Microsystems Technologies (5), 791-797 (2010). 9. C. Beine, C. Boller, U. Netzelmann, F. Porsch, R. Venkat, M. H. Schulze, A. Bulavinov, and H. Heuer, “NDT

for CFRP Aeronautical Components … A Comparative Study,” in 2nd International Symposium on NDT in Aerospace, DGZfP Proceedings (DGZfP, Berlin, 2010).

10. S. Gäbler, H. Heuer and G. Heinrich, „Measuring and Imaging Permittivity of Insulators Using High-frequency Eddy Current Devices”, IEEE Trans. on Instrum. and Meas. (publication pending).

11. Novocontrol Technologies, “Alpha-A high resolution dielectric, conductivity, impedance and gain phase modular measurement system,” User’s manual (2004).

12. A. A. Nassr and W. W. El-Dakhakhni, Smart Mater. Struct. (5), 055014 (2009). 13. C. Hellier, Handbook of nondestructive evaluation (McGraw-Hill, New York, 2003) p.8.15. 14. R. Lange and G. Mook, NDT & E Int. (5), 241–248 (1994). 15. A. Floet, „Zerstörungsfreie Werkstoffprüfung an faserverstärkten Kunststoffen mit definierten thermischen

Schäden,“ Master thesis, Dresden International University, 2013. 16. E. S. Greenhalgh, Failure analysis and fractography of polymer composites (Woodhead Publishing,

Cambridge, UK, 2009).


This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT 1

Measuring and Imaging Permittivity of InsulatorsUsing High-Frequency Eddy-Current Devices

Simone Gäbler, Henning Heuer, Member, IEEE, and Gert Heinrich

Abstract—This paper shows that the high-frequencyeddy-current (HFEC) measurement devices can be used notonly for characterizing conductivity and magnetic permeabilityrelated properties of electrically conductive materials, but also forpermittivity characterization of insulators. Maxwell’s equations,finite-element method simulations, and experimental research areapplied to support this hypothesis. An industrial HFEC deviceis used to measure the change of dielectric properties duringthe curing process of the epoxy resin L20. The measurementresults are in good agreement with the expected behavior of theparameters relative permittivity and tan δ during cure. Usinga capacitive reference device, similar characteristics regardingthe change of the complex permittivity of the resin can beobserved. In addition, HFEC imaging results on polymethylmethacrylate are presented, discussed, and compared withcapacitive imaging. HFEC permittivity mapping benefits from ahigh spatial resolution with a sensitivity and penetration depththat is at least comparable with those of capacitive imagingtechnology.

Index Terms—Dielectric constant, dielectric losses, dielectricmeasurement, EC measurement, eddy currents (ECs), elec-tromagnetic fields, electromagnetic induction, electromagneticmeasurement, epoxy curing, epoxy resins, impedance measure-ment, insulators, nondestructive testing, parasitic capacitance,permittivity, polymers.

I. INTRODUCTION

EDDY current (EC) technology is the standard nonde-structive evaluation of electric conductivity and magnetic

permeability related properties in conductive materials [1], [2].However, this restriction to electrically conductive materialsis also the most cited weakness of EC measurement in com-parison with other electromagnetic technologies, e.g., [3]–[5].As this paper will demonstrate, this disadvantage can beovercome, when using EC devices, which operate in the

Manuscript received April 1, 2014; revised November 12, 2014; acceptedNovember 17, 2014. This work was supported in part by the DresdenUniversity of Technology, Dresden, Germany, through the Program for thePromotion of Young Female Scientists and in part by the Suragus GmbH.The Associate Editor coordinating the review process was Dr. Reza Zoughi.

S. Gäbler is with the Leibniz Institute of Polymer Research, Dresden 01069,Germany and also with the Fraunhofer Institute for Ceramic Technology andSystems, Dresden 01277, Germany (e-mail: [email protected]).

H. Heuer is with the Fraunhofer Institute for Ceramic Technology andSystems, Dresden, Germany, and also with the Dresden University ofTechnology, Dresden 01069, Germany (e-mail: [email protected]).

G. Heinrich is with the Leibniz Institute of Polymer Research,Dresden 01069, Germany, and also with the Dresden University ofTechnology, Dresden 01069, Germany (e-mail: [email protected]).

Color versions of one or more of the figures in this paper are availableonline at http://ieeexplore.ieee.org.

Digital Object Identifier 10.1109/TIM.2015.2390851

high-frequency range (3–30 MHz [6]). In addition to theclassical fields of application, the permittivity characterizationon insulating and low-conductive materials becomes possible.

The technologies, which are currently prevalent forpermittivity measurement on insulating materials, are eitherlumped or distributed circuit methods. The former use acapacitive setup whereas the latter rely on wave propagationprinciples in the microwave or terahertz frequency range [7].Both approaches struggle to penetrate electrically conductivematerials. In addition, the lumped circuit methods are limitedin spatial resolution. Standard lumped circuit methods operatein the low-frequency range and usually require an excellentelectrical contact between the electrodes and the sample [8].When using an appropriate calibration, high measurementaccuracy of 0.1% regarding magnitude and 0.06° regardingphase can be achieved at nonconductive materials [9].Applied to conductive materials, capacitive methods facethe issue of electrode polarization [7], [10]. In addition,two-side access to the sample is required, when using parallelplate electrodes. Interdigital capacitive sensors overcomethis obstacle, resulting in a lower penetration depth [11].For imaging purposes, the preferred methods do not requireelectrical contact to the sample. For the low- to mid-frequencyrange, single-sided stray-field capacitive imaging is available.This method is mainly used for mapping permittivityvariations in nonconductive materials or sandwich structures.However, the use on conductive materials is mainly limitedto surface characterization as charges accumulate there.The penetration depth of capacitive imaging on insulatingmaterials increases in correspondence with the electrode size,but it must be carefully traded off against a decreasing spatialresolution [12]. Another group of permittivity measurementtechnologies are distributed circuit methods, which operate atthe microwave or terahertz frequency band [7]. Microwavemeasurement approaches use either open or closed structuresto characterize the permittivity of the material of interest [13].Closed structures like the cavity perturbation technology orwaveguide and coaxial transmission line methods usuallyrequire a particular sample preparation and geometry [14].Consequently, they are not suitable for measuring complexparts or large area applications. Open structures, such as open-ended probe reflection systems or free-space transmissionsystems, are generally suitable for imaging and operate inthe electromagnetic near or far field. The spatial resolutionin the far field depends on the measurement frequency and isgenerally limited to about half of the wavelength, resulting in∼15 mm at 10 GHz [15], [16]. Consequently, high-resolution

0018-9456 © 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.


2 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT

far-field imaging requires higher measurement frequencies,which is more expensive [15]. In the near field, thespatial resolution is mainly determined by the geometryof the probe. Thus, microwave near-field imaging achievesa high spatial resolution already at lower microwavefrequencies [4], [15]. Typical for near-field microwaveimaging is a reflective setup with open-ended rectangular waveguides or open-ended coaxial probes [5], [17], [18]. A spatialresolution of 1 mm or less and a penetration depth of severalmillimeters are reported for insulating materials [19]. Withevanescent microwaves probes, which are generated at theend of a microstrip resonator, even a lateral spatial resolutionof 0.4 μm could be reached [4]. However, with smaller probes,the lateral resolution improves at the cost of the penetrationdepth [18]. In general, the reflective setups are sensitive tovariations in the standoff distance between the probe and thesample [15]. Using a dual-polarized microwave reflectometer,this influence can be eliminated [5]. Although microwavesare generally not able to penetrate through conductivematerials [11], [15] or materials with a very highdielectric loss [18], some of the near-field methods allowcharacterization of unidirectional carbon-fiber-reinforcedpolymers (CFRPs). In this case, the electrical fieldpolarization needs to be perpendicular to the carbonfibers [5], [15]. Hence, it works for unidirectional stacks,but fails at laminates with changing fiber directions, wherepenetration is not possible [15]. In addition to reflective setups,transmission technologies can be used for imaging purposesas well [20], [21]. Due to diffraction effects at the edges of thesamples, they are most suitable for large flat specimens [14].The last group of dielectric imaging technologies, which shallbe mentioned here, is terahertz imaging. It allows a higherspatial resolution than microwave testing in the far field dueto the higher measurement frequency, but the hardware is stillsignificantly more expensive. Similar to microwave imaging,it is generally restricted to electrically nonconductive materialsor surface characterization of electrically conductive materials.An exception is the terahertz time-domain spectroscopy thatpenetrates several layers of CFRP [22].

Comparing high-frequency EC (HFEC) measurement to theprevalent and previously discussed methods to characterizepermittivity, two potential advantages become obvious:EC allows a high spatial resolution [23], when comparedwith the lumped circuit methods and a good penetrationdepth in conductive materials. The reason can be found in thelower frequency and therefore, higher standard penetrationdepth of EC compared with microwave or terahertz technolo-gies [24], [25]. However, the real penetration depth of ECfurther depends on the diameter of the sensor, as an EC sensordoes not generate a plane electromagnetic wave [25], [26].The penetration depth increases with the sensor size, atbest reaching the standard penetration depth. Nevertheless,previous researchers reported a significant penetration depth inCFRP while maintaining a high spatial resolution [24], [27].Structural defects or inclusions in multidirectional CFRPup to 8 mm below the surface were detected [28]. As mostof the briefly introduced technologies for permittivitycharacterization, the EC measurement is contact free,

allows a single-sided inspection [27], and does notrequire any special sample preparation [29]. However,usually it is solely used on conductive samples and theinfluence of the sample permittivity is neglected [1], [2].A potentially significant influence of sample permittivityon EC measurement was only reported in the context ofcharacterizing CFRP [23], [24], [30], [31]. The reason forthis effect is seen in the conductive and capacitive networkformed by the carbon rovings (bundles of carbon fibers) [30].

The aim of this paper is to prove that permittivity relatedeffects of low and nonconductive samples should not beneglected when using an HFEC device. They are rather strongenough to serve for permittivity characterization on insulatingmaterials as we will show in the following sections. Startingwith a description of electromagnetic fundamentals, the theo-retical approach is followed by a discussion of results fromfinite-element method (FEM) simulations. Then, this paperwill focus on the experimental monitoring of a curing epoxyresin to demonstrate the feasibility of the presented approach.Finally, the possibility of permittivity imaging is presented,and the results are discussed in comparison with capacitiveimaging technology.

II. THEORETICAL INFLUENCE OF SAMPLE PERMITTIVITYON COIL IMPEDANCE

Maxwell’s equations for constant frequencies demonstratethat sample conductivity and permittivity influence theEC measurement.

Maxwell’s equations describe the fundamentals ofelectromagnetic theory [32]

∇ ·D = ρ (1)∇ · B = 0 (2)

∇ × E = − jωB (3)∇ ×H = J+ jωD. (4)

The above equations are combined with the followingconstitutive equations:

B = μH (5)D = εE (6)J = σE. (7)

Attention shall be drawn into two facts.1) A magnetic field H with the magnetic flux density B,

varying in time with the angular frequency ω (as it iscaused by the ac in the coil) creates a rotating electricfield E independent of the conductivity of the sample (3).This electric field is the source for ECs with the currentdensity J, as well as for electric displacement fields Dand, respectively, displacement currents with the densityJD = jωD. The current density J depends on the con-ductivity of the sample σ (7), whereas the displacementcurrent density JD depends on the permittivity ε of thematerial to be tested (6).

2) Both the ECs and the displacement currents are associ-ated with a magnetic field H also influencing the coil’smagnetic field (4); and therefore, the coil impedance.


GÄBLER et al.: MEASURING AND IMAGING PERMITTIVITY OF INSULATORS 3

Fig. 1. Impedance plot: phase characteristics of electromagnetic effectsoccurring during HFEC measurement.

Although both conductivity and permittivity of the sampleinfluence the coil impedance, they affect the EC measurementsignal in different ways. The reasons are differences in thefrequency dependence and within the phase shift of thetwo factors.

Both the magnitudes of EC density J and displacementcurrent density JJ are frequency dependent. They behaveproportional to the intensity of the electric field E, whichdepends on the rate of change of the magnetic fluxdensity B (3). Whereas EC density J is directly related tothe electric field intensity E (7), the displacement currentdensity JJ depends on the rate of change of the electric field

∇ × H = J+ JD = σE+ jεωE. (8)

Thus, the excitation frequency has much more impacton the ability to measure permittivity using EC technology,than to measure conductivity related effects of the sample.Subsequently, this might explain why displacement currentscan be neglected at the lower frequency ranges, whichare typically used for EC inspection of metallic materials.However, when using higher frequencies permittivity relatedeffects of the sample become relevant, especially inlow-conductive materials.

Even more relevant is the phase shift of the magneticfield in response to ECs compared with the one caused bydisplacement currents. Ideally, without dielectric losses, themagnetic field change resulting from the displacement currentsis delayed by 90° compared with that caused by ECs. As aconsequence, an increase in permittivity empowers the initialmagnetic field, whereas a rise in conductivity weakens themagnetic field created by the coil.

To better understand this effect, it is useful to follow thephase characteristics of each physical phenomenon step bystep during EC measurement (Fig. 1). First, the current flowing

initially through the probing coil IL0 is defined on the real axis,thus having a phase of 0°. The magnetic field with the strengthH0, as well as the magnetic flux density B0 (5), arisingfrom the coil, are in phase with their source, the current.1Another behavior shows the electric field with the strength E0.It is shifted by −90° compared with its feeding alternatingmagnetic field (3). Exposed to a sample, the electric field leadsto ECs with the density J and displacement currents with thedensity JD. As the permittivity of the sample is a complexvalue (9), the displacement current density JD is complex aswell. The component JDP is proportional to the real part ofcomplex permittivity ε′. The component JDl represents thedielectric losses and depends on the imaginary part of complexpermittivity ε′′

ε = ε′ − jε′′ (9)JD = jεωE = jε

′ωE+ ε′′ωE = JDP + JDl. (10)

Dielectric losses JDl and EC density J occur in phase withthe electric field. Together they represent the current densityresulting from the effective conductivity σ + ε′′ω. In contrast,the displacement current DDP that is proportional to the realpart of permittivity is characterized by a phase shift of 90°compared with the electric field.

Each of those effects causes a change in the magnetic fieldstrength HC, which is in phase with its source (8). Thus,compared with the initial magnetic field, the changes resultingfrom the real part of samples permittivity HCP empowerthe initial field as they have the same phase and direction.The magnetic fields resulting from the conductivity HCJ andthe dielectric loss HCl of the sample are shifted in phaseby −90° compared with the coil’s field without sample expo-sure. Consequently, they weaken the magnetic field of the coiland cause a phase shift of the resulting current IL_res, whichis comparable with that resulting from higher ohmic losseswithin the coil [29], [33].

III. FEM SIMULATION

To test our conclusions drawn in Section II and to betterunderstand the influence of complex factors like parasiticand interwinding capacitances within the coil, FEM modelingwas used. It was found that probes with a maximum oftwo layers (in the radial direction) ensure that the permittivitymeasurement with EC technology is not dominated byinterwinding capacitances within the coil. In addition, theFEM results confirm that an increase of the permittivity ofa sample (real part) empowers the magnetic field in thecoil.

A. Model and Solver DescriptionFor FEM simulation, the EC module from ANSYS

Maxwell 3-D 16.0 was used. Compared with thelow-frequency modules from ANSYS Emag, Opera, andFlux 3-D this is the only module, where Maxwell’s equationsare fully implemented. Thus, the displacement currents arebeing considered [34]. Compared with many high-frequencymodules, the EC module of ANSYS Maxwell has the

1Disregarding the imaginary part of complex permeability μ.



advantage of simplified (stranded) modeling and analysis ofprobing coils, reducing the calculation time significantly.

The model consist of a stranded coil (a hollow coppercylinder with an outer radius of 3.9 mm, an inner radiusof 2.35 mm, and 3.2-mm height, excited by an sinusoidalcurrent of 100 mA per winding and 2.8 A in total), a cupcore of ferrite material (μr = 95, outer radius of 4.7 mm,0.7-mm wall thickness, and 4.7-mm height) and a sampleof 60 mm × 60 mm × 3 mm, positioned with a 0.4-mmliftoff to the probing coil. In addition, a 100% region wasused, limiting the virtual space, where Maxwell’s equationsare solved. The Neumann boundaries were assigned to theoutside of this region. Adaptive meshing was applied, using amaximum number of 25 passes to reach the defined error limitof 0.1%. Depending on the scope, the described model wasadapted or expanded slightly, as explained in the followingsections.

B. Relevance of Electric Fields Resulting From InterwindingCapacitances Within the Eddy Current Probe

A coil within an ac circuit does emit not only a primaryelectric field, which is linked to the primary magnetic field,but also a parasitic electric field due to its interwindingcapacitances. The simulation reveals that this parasitic electricfield can be even stronger than the primary electric field,depending on the design of the coil. This capacitive effectinfluences the permittivity measurement, especially at highfrequencies. To narrow the gap between the simulation andthe experiments, the probe coils used have to fulfill certaindesign criteria to ensure that the permittivity measurementis dominated by inductive effects and not by interwindingcapacitances.

To evaluate whether the interwinding capacitance of anEC probe (ECP) has a significant effect on the objective tomeasure sample permittivity using HFEC, the parasitic as wellas the primary electric fields were simulated and compared.The primary field was modeled using the EC solver asdescribed above. As there is no possibility to simulateinterwinding capacitances within the probe using this solver,the model was extended using an electrostatic solver. There,the parasitic electric field was modeled assuming the worstcase of two adjacent wires having the opposite poten-tial (+5 and −5 V). Compared are the magnitudes of theelectric fields in both scenarios 0.4 mm away from the coil/thewire and the direction of the field vectors.

The simulation shows that the parasitic electric vector fieldof the ECP has another direction within the object undertest than the primary electric field, but it has higher fieldstrength (Fig. 2). Based on the model of two adjacent wireswith an opposite potential, the maximum magnitude of theparasitic electric field strength is almost 20 times higher thanthe maximum magnitude of the primary electric field, which ispart of the electromagnetic wave created by the coil. Whereasthe primary electric field is an eddy field sweeping around themagnetic vector field, the parasitic electric field is orthogonalto the sample.

Modeling the interaction of the primary and the parasiticelectric field is not possible in ANSYS, as the electrostatic

Fig. 2. Magnitude of the primary electric eddy field within the objectunder test (top) compared with the magnitude (bottom) and direction(bottom, inset) of the parasitic electric field resulting from the interwindingcapacitance of the coil.

module cannot be combined with the EC module. Thismakes it difficult to conclude whether the interwinding capac-itance improves or disturbs the permittivity measurement withan ECP. The aim of this paper is to demonstrate that thepermittivity measurements are possible using the inductivemechanisms (E × H field) of an ECP, thus the parasitic electricfield needs to be reduced to a minimum.

To reduce the parasitic electric field, the wire length lbetween the two adjacent windings has to be optimized. At agiven wavelength, this is the factor determining the potentialdifference. Consequently, it determines the parasitic electricfield between the adjacent windings, as the voltage with thepeak value û is propagating sinusoidal through the wire

U = u(

sin(ωt0 + 2πl

λ

)− sin(ωt0)

). (11)

The modeled maximum of U = 10 V [+5 and −5 V]between adjacent wires in a probe only appears when thedistance between the adjacent wires l is half of a wavelength λ.For a measurement frequency of 10 MHz, the correspondingwire length between the two adjacent windings is ∼15 m.This would be the case for a coil with 60 windings perlayer and 4 cm in diameter, which is not appropriate for



Fig. 3. Vertical cross section of a coil explaining the influence of the numberof layers on the interwinding capacitances next to the coil.

high-frequency measurements. Nevertheless, here, the voltagewould reach its maximum in one wire, whereas it would bepassing the minimum in the other one. Thus, to minimizethe potential difference U , the wire length between thetwo adjacent windings should be as small as possible. Thiscan be achieved by fulfilling the following two design criteria.

1) Avoiding more than two layers of windings.2) Keeping the coil diameter small.The condition of using ECPs with a maximum of two layers

to minimize interwinding capacitances is related to the way theprobing coils are built up. They are wound by first applying afull layer of windings on the coil former, e.g., from bottom totop, and then applying the second layer from top to bottom,and so on. Thus, a coil consisting of two layers typically hasthe adjacent turns with the maximum difference in potentialnot situated next to the sample, but on the opposite side of thecoil. When adding a third layer to the coil, for the first time,the two adjacent windings next to the sample have a significantwire length l in between. This results in a significant voltagedifference (Fig. 3).

Reducing the coil diameter without removing the third layerreduces the parasitic electric field significantly but not suffi-ciently. Assuming a coil diameter of ∼6 mm and 10 turns perlayer, the wire length l between the two adjacent winding is∼0.38 m (π∗d∗20). At 10 MHz, this is about one-hundredthof the wave length, leading to a maximum voltage differencebetween the adjacent wires <0.4 V. This is more than 20 timesless compared with the +5/−5 V scenario but still sufficientto create an electric field with almost the same strength as theprimary electric field of the ECP.

C. Change of the Magnetic Flux Passing Through theHFEC Probe Due to Permittivity Variations

The simulation results support the conclusions drawnin Section II. With increasing sample permittivity (real part) anincreasing magnetic flux is passing through the HFEC probe.Increasing dielectric losses weaken the magnetic flux similarlyas an increase of sample conductivity.

The previously described model was used in thesimulation. Six homogenous samples were defined, showingdifferences in complex relative permittivity and bulk

TABLE ISAMPLE CHARACTERISTICS AND CALCULATED MAGNETIC FLUX

TABLE IIMAGNETIC FLUX DIFFERENCES DEPENDING ON THE SPECIFIC

PARAMETER THAT IS CHANGED

conductivity (Table I). Relative magnetic permeability μrwas kept constant at μr = 1.

The magnetic flux � passing through the horizontal crosssection S of the coil was calculated at f = 10 MHz using thefield calculator

� =∫∫S

BdS. (12)

The calculated magnetic flux passing through the coildecreases with increasing sample conductivity (Table II,scenario IV-III) and is shifted away from the real axis (Table I,sample IV).

The flux differences caused by a change of sample permit-tivity are quite small with ∼10−14 Wb, which is a challengefor experimental evaluation. It will be important to strictlyeliminate drift effects and to use a measurement setup thatallows a very good signal-to-noise ratio. To eliminate drifteffects like temperature changes regular reference or calibra-tion measurements are needed. To achieve a good signal-to-noise ratio, there are two factors that should be considered.

1) It is important that the processed signal consists mainlyof the change of the signal due to sample permittivity.It should not contain the total signal that is generatedin the coil, as this is the orders of magnitudes bigger.Therefore, a thoroughly adjusted balancing coil isneeded that can compensate for a good portion of thesignal in air [35].

2) It is critical to minimize noise by carefully selecting thecomponents for the analog signal processing like cables,amplifiers, and the analog-to-digital converter. However,from a simulation point of view, the calculated fluxdifferences are significant. The calculations of magneticflux are reproducible up to about 10−30 Wb difference,so the results are about the factor 1016 away from therange of numerical error.



Fig. 4. Change of real part of the magnetic flux passing through the coilshowing a linear dependence on the change of the real part of permittivityfor nonconductive samples.

Fig. 5. Change of imaginary part of the magnetic flux passing through thecoil showing a negative linear dependence on the change of the imaginarypart of permittivity for nonconductive samples.

By analyzing the calculated magnetic flux differences, threeconclusions can be drawn for nonconductive samples.

1) The difference in magnitude of the magnetic flux |�|is negative when the dielectric loss ε′′

r increases celsiusparibus (c.p.), but positive when the real part of per-mittivity ε′

r increases (Table II). Hence, an increasingdielectric loss weakens the magnetic flux � passingthrough the coil where increasing sample permittivityempowers it.

2) The change of the real part of the magnetic flux passingthrough coil Re(�) is mainly influenced by the changeof the real part of permittivity ε′

r (Fig. 4) and shows alinear dependence on this variable within the analyzedlimits. The flux change does not depend on the initialpermittivity.

3) The change of the imaginary part of the magneticflux passing through coil Im(�) shows a negativelinear correlation with the imaginary part of permittivityε′′

r within the analyzed limits (Fig. 5).

IV. MONITORING PERMITTIVITY CHANGE OFA CURING EPOXY RESIN

Based on Maxwell’s equations and on the FEM simulation,it was shown that it is theoretically possible to measure permit-tivity on insulators or composites using HFEC. The following

section gives a first experimental proof. By monitoring thecuring process of an industrial epoxy amine resin system,we demonstrate that permittivity data obtained with HFECdevices are comparable with those obtained with a standardcapacitive sensor.

The cure monitoring was selected, because it is a time-dependent process; i.e., it provides the opportunity to measurea permittivity change within one specific sample, avoidingtypical sources for error in HFEC measurement. Such errorsmight be caused, for example, by differences in samplegeometry and unintended differences in the experimentalsetups like liftoff, sample-to-sensor position, and so on.In addition, the selected task does not require directlymeasuring quantitatively, as the permittivity change duringcure has a characteristic pattern overtime. Thus, a qualitativecomparison with the reference method is possible.

A. Expected Permittivity Change ofEpoxy Resins During Cure

To explain the change of dielectric properties during cureof an epoxy resin, the following topics are briefly introduced:1) factors determining the permittivity of a material; 2) thecuring process of an epoxy; and 3) the resulting permittivitychanges during cure.

Permittivity describes how a material responds toan external electric field. It is a complex quantity resultingfrom the molecular mechanisms of polarization and chargemigration. Charge migration occurs mainly at lowerfrequencies in the form of ionic conduction. Polarization iscaused by dipole orientation, either of permanent dipoles orinduced ones [36], [37]. An extensive overview on molecularand environmental factors that are determining the permit-tivity of a material as well as on existing models can befound in [10]. To understand the expected permittivity changeduring cure, it is important to know that the permittivity ofa material is frequency dependent, which is described bycharacteristic dielectric relaxation times within a variety ofsemiempirical equations [7]. Reaching a certain frequency,a growing number of permanent dipoles cannot move fastenough, hence permittivity is decreasing. At the same time,polarization losses increase, so the imaginary part of thepermittivity is reaching a local maximum [10]. The dielectricrelaxation times increase with increasing molecular weight andincreasing viscosity [38].

During the curing reaction of an epoxy resin and a hardener,the concentration of the reactants is decreasing whereas theconcentration of the usually less polar end product [39] isincreasing. Due to the cross-linking process, molecular weightis rising, resulting among others in an augmentation of viscos-ity. As the reaction is exothermic, heat is produced [37].

These micro and macromolecular changes occurring withinan epoxy during cure affect the permittivity in three differentways: 1) ionic conduction decreases; 2) dielectric relaxationtimes of the reactive system decline; and 3) the relativepermittivity drops as well [37], [39]–[41].

The decrease of ionic conduction is only visible in thelow-frequency range, thus this mechanism cannot be used for



Fig. 6. Progress of complex permittivity during cure [38], [39].

HFEC measurements [39]. However, especially, when usingcapacitive setups, scientists often rely on the change of ionicconduction to monitor the curing process [41]–[43].

More relevant for HFEC measurement is the phenomenonof declining relaxation times. It can be observed as a peakof the imaginary part of permittivity measured at a constantfrequency during cure (Fig. 6). The peak indicates the point intime when the current relaxation time is equal to the reciprocalof the frequency (in rad). Thus, at higher frequencies, the peakoccurs earlier than at lower ones (relaxation time has alreadyincreased so that the molecular motion cannot follow a higherfrequency anymore but still can follow a lower one).

The second effect that can be used to monitor the cure ofan epoxy with HFEC devices is the drop of the real partof permittivity. In the early stage of reaction, only a slightdecrease of permittivity is observed, which results from theconcentration change of reactants and the usually less polarend product. Later during cure, due to the increasing relaxationtimes and vitrification of the system, the permittivity decreasesmuch more abruptly. Molecular motion cannot keep up withthe measurement frequency anymore [39]. As explained for theloss peak, with increasing frequency, this phenomenon can benoticed earlier in the curing process (Fig. 6).

Although dielectric cure monitoring is widely used,influences of the electromagnetic field on the sample cannotgenerally be neglected and should be evaluated regardingthe specific material of interest. In particular, liquid-crystalthermosets show strong orientation effects on a microscopiclevel, resulting in a change of macroscopic properties, whencured under the influence of alternating electric fields or strongstatic magnetic fields [44], [45]. In addition, the alignmentof certain filler materials like carbon black [46] or carbonnanotubes [47] can be influenced by applying ac electric fieldsduring cure. In addition, the electric fields can generate heatin polymeric samples due to dielectric losses. Hence, theycan influence the curing process itself. This principle is used,for example, in microwave heating/curing [48]. However, forfrequencies below the microwave range, and for the smallelectric fields that are used for dielectric cure monitoring, thiseffect of heat generation can be neglected.

B. Experimental SetupAn industrial epoxy amine resin system (L20 and EPH 161

distributed by R&G, Waldenbuch, Germany) was manuallymixed following the prescribed weight ratio of 100:25. It wascured at room temperature afterward. During the curing

Fig. 7. Experimental setup. Left: capacitive measurement.Right: HFEC measurement.

process, the change of its permittivity was measured using acapacitive measurement setup and an HFEC device. The fullset of measurements with the two different devices was notconducted on the same day nor at the same sample. Thus, theresults may vary slightly between both approaches, as smalldeviations in stoichiometric composition or room temperatureare possible.

The capacitive reference setup consisted of an LCR meter(HP4275A) connected to a comb electrode (Netzsch IDEX,Model 065S A/D, Ratio 80), built within a Faraday cage.About 20 g of the resin system were filled into a small cup(25-mm diameter and 30-mm high), and placed in the faradaycage immediately after mixing (Fig. 7). The electrode wasdipped into this cup until it was completely covered by resin.Capacitance C and dissipation factor D were measured at thefrequencies 2 and 4 MHz and manually recorded every 5 minfor the first 2 h. Later, the measurement interval was increased.

For the HFEC measurement, the industrial device EddyCUSCF map 4040 from Suragus and a specifically designedpermittivity sensor named RD-P020B was used. The probehas an outer diameter of 24 mm, a coil diameter of 2 mm,and the highest sensitivity at 6 MHz. The manipulator ofthe device was used for a one-point measurement with acontrolled, 1-mm liftoff between the sample and the probe.In addition, the mapping device made it possible to over-come the temperature effects caused by the exothermal curingreaction. The sensor was not permanently located at the mea-surement position, but parked at a reference position betweentwo measurements. When the reference position was reached,a no sample measurement (air) was conducted for calibrationpurpose. Every 60 s, the sensor moved to the measurementposition. There, ∼100 g of the epoxy resin was placed ina plastic cup (112-mm diameter and 15-mm high, Fig. 7).Movement and automated recording of the measurement datawas realized by a software tool developed at Fraunhofer IKTSMD (former Fraunhofer IZFP, Dresden). HFEC measurementswere conducted at 2 and 6 MHz. The LCR meter that was usedfor the capacitive reference setup was restricted to frequenciesof 2, 4, and 10 MHz. Unfortunately, the HFEC sensor showeda low signal-to-noise ratio at 4 and 10 MHz, so 6 MHz wasselected as the second measurement frequency.



Fig. 8. Capacitive measurement during cure of epoxy resin L20: capacityas a measure for real part of permittivity and dissipation factor proportionalto tan δ of permittivity.

For a better comparison of the two different measurementtechnologies, curves were fitted to the tan δ values obtained at2 MHz using the Excel solver (minimizing the sum of squaresof deviations). Following the approach of [38] and [49],a classical Debye model [50] was used for modeling chemicalreactions in solutions. Thus, tan δ can be expressed as afunction of the reaction time t , the angular frequency ω,and the four fitting parameters, namely, static permittivity εs ,high-frequency unrelaxed permittivity ε∞, the material-dependent time constants a, and the material-dependentinverse time constant k

tan δ = εs − ε∞ε∞ + εs − ε∞

1 + ω2τ 2

∗ ωτ

1 + ω2τ 2 (13)

with τ = aekt .Using (13), the time tm , when the peak in the dissipation

factor occurs, can be described

tm = ln(

εs

a2ε∞ω2

)/2k. (14)

As the model is designed for chemical reactions in solutions,the fit only works well up to a certain degree of vitrification.We considered this point by only fitting the measurement dataof the first 4 h.

C. Results of the Capacitive Reference MeasurementThe permittivity change during cure (Fig. 8), which was

observed qualitatively via the comb electrode, shows thetypical characteristics that we described earlier. For quantifi-cation, the permittivity of the substrate material and capacityof the cables must be considered. This could be done byexperimental calibration measurements with a well-knownfluid or alternatively by model-based expost corrections ofthe measured values [51]. The model-based approach requiresthe permittivity value of the substrate material [52], whichwas not available. Likewise, a well-known liquid referencematerial, where complex permittivity is exactly listed for2 and 4 MHz, could not be identified. As both permittivityof the substrate as well as capacity of the cables stay almostconstant during cure, they do not influence the qualitative cure

monitoring at a constant frequency. Only when comparingdifferent frequencies, it is important to keep those influencesin mind.

Focusing at one measurement frequency, it can be seenthat the capacitance decreases during cure. It depends onthe real part of permittivity [10], [51] and shows the typicalcharacteristics (as summarized in Section IV-A) that wewould have expected regarding a permittivity change ofan epoxy resin during cure. At first, the decrease is slow(about −0.05 pF/min). About 90 min after mixing (which isalso the pot time of this epoxy system), permittivity dropsmuch faster for the next 3–4 h (Fig. 8, about −0.2 pF/min).Then, the capacitance decrease slows down again. Whencomparing the acquired data from 4 to 2 MHz (Fig. 8),the expected time shift between both capacitance curves(Fig. 6) cannot be seen clearly. Probably, the difference infrequency is too small. Instead, at higher frequencies, a highercapacitance is measured; within the resin as well as withoutsample. However, this does not necessarily mean that the initialpermittivity differs between 2 and 4 MHz. The cables and theelectrode, which are not designed for such high frequencies,may influence the signal. This argument is supported by thefact that even in air, the measured capacitance rises withincreasing frequency.

Looking at the dissipation factor tan(δ) (Fig. 8), a rathergood agreement with the expected behavior is observed.It shows a clear maximum during cure, which is reachedearlier at higher frequencies. For 2-MHz, fitting parameterswere estimated with εs = 5.72, ε∞ = 4.82, a = 1.06E−08 s,and k = 1.91E−04 s−1 at a mean squared error of 0.4E−05between the measured and estimated tan δ. The time when thepeak occurs was estimated at tm = 10 978 s.

D. Results of the HFEC MeasurementUsing an HFEC device, it is possible to qualitatively monitor

the permittivity change of an epoxy resin during cure. A verygood agreement was found between the HFEC data and thoseobtained with the capacitive reference approach.

The right processing of raw data is crucial foran accurate interpretation of the results obtained byHFEC devices (Fig. 10). For each EC measurement value, acorresponding reference value in air was taken by moving thesensor away from the sample, using the gantry of the scanner.These air values are used to reference each measurementvalue to air, and hence to eliminate temperature and long-termdrift effects. Consequently, only the impedance differenceZ between air and the sample is evaluated. Thus, the measuredpermittivity needs to be interpreted as a permittivity differenceto air. This approach is widely used in EC measurement [1].However, it is assuming an additive influence of drift.

In addition to the drift effects, it is also important tocorrect the frequency-dependent phase offset caused by themeasurement system itself [53]. The phase of the measuredimpedance values is important for determining its real andimaginary parts, as well as the right tan δ. A numericalsolution for correcting the phase offset has not yet beenfound. Hence, we made use of the capacitive reference values.



Fig. 9. Measurement with HFEC device during cure of epoxy resinL20: normalized imaginary part of complex impedance as a measure for realpart of permittivity and tan δ of impedance analog to tan δ of permittivity.

Fig. 10. Principle of phase offset correction and final set of data (gray).

The phase offset was defined as follows. The tan δ of thefirst EC measurement was set to be equal to the dissipationfactor measured with the capacitive approach at 2 MHz and tomeet the dissipation factor calculated for 6 MHz by applyinga Debye model [50] to the 2-MHz value. This might beinappropriate for a quantitative measurement (as the capacitivevalues are not exact either and tan δ of air is neglected), butfor a qualitative comparison it should be sufficient.

The data set derived after processing the raw values can beused for analyzing the change of the complex impedance overcure time (Fig. 9). From Maxwell’s equations (1)–(4) and theFEM simulation (Sections II and III), we know that the changeof the imaginary part of impedance represents the change ofthe real part of permittivity, whereas the real part of impedanceis influenced by a changing dielectric loss. Subsequently, whencomparing the HFEC measurement results to the capacitanceand dissipation factor data obtained by the capacitive referencemeasurement, the change of the imaginary part of the compleximpedance Z and the ratio real to imaginary part of compleximpedance (tan δ) should be used. To enhance comparabilityof two measurement frequencies, we additionally normalizedthe imaginary part of the complex impedance Z (using theaverage of the first five measurements).

Now, comparing the results obtained with the HFEC deviceto those measured with the capacitive reference setup, a highsimilarity can be seen. All the typical characteristics of thecapacitive reference measurement are present in the HFECdata as well—such as the typical progression of relativepermittivity or the frequency-dependent peak in dissipationlosses tan δ. Even the relative change of permittivity is veryconsistent. At 2 MHz, the tan δ curve fitting parameters wereestimated with εs = 5.59, ε∞ = 4.83, a = 1.13E−08 s, andk = 1.90E−04 s−1 at a mean squared error of 1.73E−05between the measured and estimated tan δ. The time whenthe peak occurs was estimated with tm = 10 680 s. This is∼5 min earlier than the estimated peak of the capacitive setup.This deviation is <3%, which is quite small regarding possibleexperimental differences in mixing ratio and room temperatureas well as fitting uncertainties. The most significant differenceis the higher scattering of the values measured with theHFEC device, resulting in a higher mean squared error for thefitting function. One explanation might be a slight deviationof the sensor position for each measurement. The probe canrepetitively return to a position with an accuracy <100 μm.Compared with the 1-mm liftoff, this is still a significantdeviation. Improving mechanics for the sensor movement,using an advanced sensor design and measuring at higherfrequencies should reduce the scatter drastically. In addition,enhanced data processing, such as averaging of raw data orcurve fitting, might be another possibility to improve thequality of HFEC cure monitoring.

However, the example of cure monitoring was selected toqualitatively prove the concept of permittivity measurementusing HFEC devices in a favorable setup for this technology(see the introduction in Section IV). For an industrial applica-tion, in this field, HFEC will need to demonstrate its advan-tages compared with the capacitive measurement [49], [51]and microwave cure monitoring [54], [55]. At the moment,HFEC devices are not achieving measurement accuraciesthat are comparable with contacting capacitive or microwavemethods. Nevertheless, for applications where this accuracyis not critical, HFEC measurement could be a robust andsimple alternative [56] for qualitative cure monitoring. A cal-ibration (using a known sample or a capacitive referencesetup) is needed to determine the phase offset for a specificmeasurement frequency of a specific sensor. That calibrationeither could be done at the manufacturer of the HFEC devicefor each system, which is sold for permittivity measurement,or directly at the customer. If it should be done at the customercalibration samples with a known complex permittivity and agood instruction or even software for calibration should besupplied with the sensor. That would make especially sensefor customers, who already own an HFEC device and wantto use it additionally for permittivity characterization. Oncecalibrated, the system can be used for different resins anddoes not require contact to the sample.

V. PERMITTIVITY IMAGING

In addition to the single-point permittivity measurementdescribed in the previous section, a qualitative permittivityimage can be acquired by moving an HFEC probe over



Fig. 11. Specification of sample 1, measures in millimeter.

Fig. 12. HFEC image of sample 1, measurement values in digits.

a dielectric material. A potential application would be thehomogeneity control of the permittivity after the resin is cured.This is a good indicator for the degree of cross linking [10]or the detection of defects below the surface.

In addition to the proof-of-concept for HFEC permittivitymapping on insulating materials, this section should serveas a first experimental comparison with capacitive imaging.Therefore, we produced our samples according to thespecifications mentioned in [12].

Sample 1 is a 240 mm × 50 mm × 25-mm polymethylmethacrylate (PMMA) plate, which contains six 15-mm deepholes of different diameters (Fig. 11). It was characterizedusing the industrial device EddyCUS CF map 4040 fromSuragus combined with the sensor RD-P007B. This probewas particularly designed for permittivity mapping. It hasan outer diameter of 16 mm, a coil diameter of 0.7 mm(allowing a high spatial resolution), and is especially sensitiveat 1.75 and 3.5 MHz. An area of 30 mm×230 mm was scannedat 3.5 MHz and 25 dB, using a pitch of 0.2 mm × 1 mm.To avoid liftoff variations, the sample was scanned in contactwith the sensor.

The resulting HFEC image (Fig. 12) shows all holes, eventhe smallest one with only 1-mm diameter. Compared with theresults reported in [12], the HFEC image shows a significantlybetter spatial resolution and a comparable or even slightlybetter sensitivity.

Sample 2 is a 240 mm×90 mm×10-mm PMMA plate con-taining four 20 mm × 20-mm flat-bottomed holes of differentdepths—2, 4, 6, and 8 mm (Fig. 13). As opposed to the firstsample, this time the sensor RD-P020B was used, which wasalready introduced in Section IV. An area of 195 mm×45 mmwas scanned at 6 MHz, 32 dB, using a pitch of 0.2 mm×1 mm,once from faces A and B.

The resulting HFEC image from face A (Fig. 14 con-tains information on the depth of the holes (grayscale).

Fig. 13. Specification of sample 2, measures in millimeter.

Fig. 14. HFEC image of sample 2, face A. Measurement values in digits.

Fig. 15. HFEC image of sample 2, face B and cutout with adjusted grayscale. Measurement values in digits.

This is another advantage compared with the capacitive imag-ing results reported in [12]. Again, the better spatial resolutionof HFEC measurement is evident.

The HFEC image from face B (Fig. 15) proofs that thetechnology is able to detect hidden defects in insulatingmaterials like PMMA. All holes, up to 8 mm below thesample surface, have been detected. The depth of the defectis visible in its gray shade, but mixed with the informationon defect size (depth). However, with increasing distance tothe surface, it becomes more difficult to see the exact shapeof the defect. In comparison with the capacitive imagingresults reported in [12], it can be seen that HFEC shows abetter visibility of the holes near the sample surface, whereascapacitive imaging achieves a better contrast for defects furtheraway from the sample surface.



In general, capacitive imaging has the advantage that itis not influenced by magnetic properties of the sample.In addition, the detection of planar cracks, parallel to thesample surface might be easier due to the direction of theelectric field [3], [4]. However, an HFEC device might be moresuitable to detect planar cracks, which are orthogonal to thesample surface or for measurement tasks where the samplesurface is covered by a thin conductive layer. In addition,the HFEC device can be used for various other applicationson conductive materials as well, which might be especiallyadvantageous for small laboratories.

VI. CONCLUSION

Our research demonstrates that HFEC devices, operatingin the megahertz frequency range, can not only be used forcharacterization of electrically conductive materials, but alsofor imaging and investigation of insulating samples. In thiscase, it is the complex permittivity of the material, whichis influencing the complex impedance. The experimentalevidence was given by monitoring the permittivity changeduring cure and by mapping defects in insulators. Potentialadvantages of HFEC permittivity measurements comparedwith capacitive imaging were identified (higher spatialresolution and defect depth indication). However, all thepresented results were qualitatively and only attained oninsulating samples. Maxwell’s equations and FEM simulationsgive the indication that sample permittivity also influences theHFEC measurement of low conductive samples. In addition,there is the hypothesis that the sample permittivity is part ofthe HFEC response even when characterizing CFRP [23], [30].

A focus of future research, therefore, needs to address thefollowing two topics.

1) The evaluation of measurement accuracy regardingquantitative permittivity characterization using a calibra-tion curve. It is important to know the potential of HFECto evaluate the tradeoff between the saving of samplepreparation but lower measurement accuracy.

2) To evaluate the potential of HFEC for permittivitycharacterization of low conductive, highly anisotropicmaterials like multidirectional CFRP. On insulationmaterials, the technology is able to compete withcapacitive imaging but not to compete againstspatial resolution of microwave or terahertz systems(e.g., evanescent microwave probe imaging has a 0.4-μmlateral spatial resolution at 1 GHz [4]). MultidirectionalCFRP is the application where most existing permittivitycharacterization methods fail. Contacting capacitivemethods suffer from electrode polarization [7], [10],capacitive imaging only allows surface characterizationdue to accumulating charges on the sample surface [12]and high-frequency methods, which are operatingin the microwave or terahertz frequency range, onlyallow penetration into unidirectional CFRPs [5], [15].When evaluating the use of HFEC for permittivitymeasurement in CFRP, a major part of the work isnot only evaluation of the possibility of permittivitycharacterization but also the separation of the differentsample properties that are influencing the HFEC signal.

REFERENCES

[1] Non-Destructive Testing—Eddy Current Testing—General Principles,ISO Standard 15549, 2008.

[2] A. Yashan, “Über die Wirbelstromprüfung und magnetische Streu-flussprüfung mittels GMR-Sensoren,” Ph.D. dissertation, Dept. Mater.Sci., Univ. Saarlandes, Saarbrücken, Germany, 2008.

[3] G. G. Diamond and D. A. Hutchins, “A new capacitive imagingtechnique for NDT,” presented at the Eur. Conf. NDT, Berlin, Germany,Sep. 2006.

[4] M. Tabib-Azar, P. S. Pathak, G. Ponchak, and S. LeClair, “Nonde-structive superresolution imaging of defects and nonuniformities inmetals, semiconductors, dielectrics, composites, and plants using evanes-cent microwaves,” Rev. Sci. Instrum., vol. 70, no. 6, pp. 2783–2792,Jun. 1999.

[5] S. Kharkovsky and R. Zoughi, “Microwave and millimeter wave nonde-structive testing and evaluation—Overview and recent advances,” IEEEInstrum. Meas. Mag., vol. 10, no. 2, pp. 26–38, Apr. 2007.

[6] Nomenclature of the Frequency and Wavelength Bands Used inTelecommunications, document Rec. ITU-R V.431–7, 2000.

[7] J. Mijovic, J. Kenny, A. Maffezzoli, A. Trivisano, F. Bellucci, andL. Nicolais, “The principles of dielectric measurements for in situmonitoring of composite processing,” Compos. Sci. Technol., vol. 49,no. 3, pp. 277–290, 1993.

[8] A. A. Nassr and W. W. El-Dakhakhni, “Non-destructive evaluation oflaminated composite plates using dielectrometry sensors,” Smart Mater.Struct., vol. 18, no. 5, p. 055014, 2009.

[9] Alpha-A High Resolution Dielectric, Conductivity, Impedance andGain Phase Modular Measurement System, Novocontrol Technologies,Montabaur, Germany, 2004.

[10] A. R. Blythe and D. Bloor, Electrical Properties of Polymers, 2nd ed.New York, NY, USA: Cambridge Univ. Press, 2005, pp. 27–185.

[11] R. A. Pethrick, “Non-destructive evaluation of composites: Dielectrictechniques for testing partially or non-conducting composite materials,”in Non-Destructive Evaluation (NDE) of Polymer Matrix Composites:Techniques and Applications, V. M. Karbhari, Ed. Cambridge, U.K.:Woodhead Pub., 2013, pp. 116–135.

[12] X. Yin, D. A. Hutchins, G. Chen, and W. Li, “Investigations intothe measurement sensitivity distribution of coplanar capacitive imagingprobes,” NDT & E Int., vol. 58, pp. 1–9, Sep. 2013.

[13] A. W. Kraszewski, “Microwave aquametry—A review,” J. Microw.Power, vol. 15, no. 4, pp. 209–220, 1980.

[14] M. S. Venkatesh and G. S. V. Raghavan, “An overview of dielectricproperties measuring techniques,” Can. Biosyst. Eng., vol. 47, no. 7,pp. 15–30, 2005.

[15] L. Diener, “Microwave near-field imaging with open-ended waveguide—Comparison with other techniques of nondestructive testing,” Res.Nondestruct. Eval., vol. 7, nos. 2–3, pp. 137–152, 1995.

[16] R. Zoughi, Microwave Non-Destructive Testing and Evaluation Princi-ples. Dordrecht, The Netherlands: Kluwer, 2000.

[17] R. Zoughi and S. Kharkovsky, “Microwave and millimetre wave sensorsfor crack detection,” Fatigue Fract. Eng. Mater. Struct., vol. 31, no. 8,pp. 695–713, 2008.

[18] J. H. Hinken, “Mikrowellenbasierte zerstörungsfreie Prüfung,”ZfP-Zeitung, vol. 104, pp. 37–42, Apr. 2007.

[19] T. Beller, J. H. Hinken, and M. Voigt, “Hochauflösende Mikrowellen-Defektoskopie,” in Proc. DGzfP Zerstörungsfreie Materialprüfung,Fürth, Germany, 2007, p. V53.

[20] M. Pastorino, “Recent inversion procedures for microwave imaging inbiomedical, subsurface detection and nondestructive evaluation applica-tions,” Measurement, vol. 36, nos. 3–4, pp. 257–269, 2004.

[21] M. Abu-Khousa, W. Saleh, and N. Qaddoumi, “Defect imaging andcharacterization in composite structures using near-field microwavenondestructive testing techniques,” Compos. Struct., vol. 62, nos. 3–4,pp. 255–259, 2003.

[22] A. Redo-Sanchez, N. Karpowicz, J. Xu, and X. C. Zhang, “Damageand defect inspection with terahertz waves,” in Proc. 4th Int. WorkshopUltrason. Adv. Methods Nondestruct. Test. Mater. Characterization,2006, pp. 67–78.

[23] M. H. Schulze, H. Heuer, M. Küttner, and N. Meyendorf, “High-resolution eddy current sensor system for quality assessment of carbonfiber materials,” Microsyst. Technol., vol. 16, no. 5, pp. 791–797, 2010.

[24] H. Heuer, M. H. Schulze, and N. Meyendorf, “Non-destructive evalua-tion (NDE) of composites: Eddy current techniques,” in Non-DestructiveEvaluation (NDE) of Polymer Matrix Composites: Techniques andApplications, V. M. Karbhari, Ed. Cambridge, U.K.: Woodhead Pub.,2013, pp. 33–55.



[25] D. J. Hagemaier, “Eddy-current standard depth of penetration,” Mater.Eval., vol. 43, no. 11, pp. 1438–1441, 1985.

[26] Z. Mottl, “The quantitative relations between true and standard depth ofpenetration for air-cored probe coils in eddy current testing,” NDT Int.,vol. 23, no. 1, pp. 11–18, 1990.

[27] M. P. De Goeje and K. E. D. Wapenaar, “Non-destructive inspec-tion of carbon fibre-reinforced plastics using eddy current methods,”Composites, vol. 23, no. 3, pp. 147–157, 1992.

[28] C. Beine et al., “NDT for CFRP aeronautical components—A compar-ative study,” presented at the 2nd Int. Symp. NDT Aerosp., Hamburg,Germany, Nov. 2010.

[29] C. Hellier, Handbook of Nondestructive Evaluation. New York, NY,USA: McGraw-Hill, 2003.

[30] R. Lange and G. Mook, “Structural analysis of CFRP using eddy currentmethods,” NDT & E Int., vol. 27, no. 5, pp. 241–248, 1994.

[31] S. K. Brady, “Exploiting changes in dielectric dissipation to detectthermal degradation in polymers and composites,” Rev. Progr. Quant.Nondestruct. Eval., vol. 29, no. 1, pp. 1087–1094, 2010.

[32] J. C. Maxwell, “A dynamical theory of the electromagnetic field,”Philosoph. Trans. Roy. Soc. London, vol. 155, pp. 459–512, Jan. 1865.

[33] J. García-Martín, J. Gómez-Gil, and E. Vázquez-Sánchez, “Non-destructive techniques based on eddy current testing,” Sensors, vol. 11,no. 12, pp. 2525–2565, 2011.

[34] ANSYS. (Mar. 2012). Maxwell 3D User’s Guide: Software Version 15.[Online]. Available: http://www.mae.ncsu.edu/buckner/courses/mae535/Maxwell3D.pdf

[35] J. Hansen, “The eddy current inspection method. Part 2: The impedanceplane and probes,” Insight, vol. 46, no. 6, pp. 364–365, 2004.

[36] H. Block, “The nature and application of electrical phenomena inpolymers,” in Electric Phenomena in Polymer Science (Advancesin Polymer Science), vol. 33, H.-J. Cantow, Ed. Berlin, Germany:Springer-Verlag, 1979, pp. 93–167.

[37] G. W. Ehrenstein, E. Bittmann, and L. Hoffmann, Duroplaste:Aushärtung—Prüfung—Eigenschaften. München, Germany: Hanser,1997, pp. 1–19 and 69–75.

[38] S. Wu, S. Gedeon, and R. A. Fouracre, “The measurement and modelingof the dielectric response of molecules during curing of epoxy resin,”IEEE Trans. Electr. Insul., vol. 23, no. 3, pp. 409–417, Jun. 1988.

[39] J. Mijovic and B. D. Fitz. (1998). Dielectric Spectoskopy of Reac-tive Polymers. Novocontrol Application Note Dielectrics 2. [Online].Available: http://www.novocontrol.de/pdf_s/APND2.PDF

[40] A. Schönhals, “6.3 Elektrische und dielektrische Eigenschaften,” inKunststoffprüfung, W. Grellmann and S. Seidler, Eds. München,Germany: Carl Hanser Verlag, 2005, pp. 352–404.

[41] N. F. Sheppard and S. D. Senturia, “Chemical interpretation of therelaxed permittivity during epoxy resin cure,” Polym. Eng. Sci, vol. 26,no. 5, pp. 354–357, 1986.

[42] D. Kranbuehl et al., “Use of the frequency dependence of the impedanceto monitor viscosity during cure,” Polym. Eng. Sci., vol. 29, no. 5,pp. 285–289, 1989.

[43] Y. A. Tajima, “Monitoring cure viscosity of epoxy composite,” Polym.Compos., vol. 3, no. 3, pp. 162–169, 1982.

[44] A. Shiota and C. K. Ober, “Orientation of liquid crystalline epox-ides under ac electric fields,” Macromolecules, vol. 30, no. 15,pp. 4278–4287, 1997.

[45] B. C. Benicewicz et al., “Magnetic field orientation of liquid crystallineepoxy thermosets,” Macromolecules, vol. 31, no. 15, pp. 4730–4738,1998.

[46] M.-K. Schwarz, W. Bauhofer, and K. Schulte, “Alternating electric fieldinduced agglomeration of carbon black filled resins,” Polymer, vol. 43,no. 10, pp. 3079–3082, 2002.

[47] C. A. Martin et al., “Electric field-induced aligned multi-wall carbonnanotube networks in epoxy composites,” Polymer, vol. 46, no. 3,pp. 877–886, 2005.

[48] J. Mijovic and J. Wijaya, “Comparative calorimetric study of epoxycure by microwave vs thermal energy,” Macromolecules, vol. 23, no. 15,pp. 3671–3674, 1990.

[49] J. Delmonte, “Electrical properties of epoxy resins during polymeriza-tion,” J. Appl. Polym. Sci., vol. 2, no. 4, pp. 108–113, 1959.

[50] P. Debye, Polare Molekeln. Leipzig, Germany: Hirzel, 1929.[51] S. D. Senturia and N. F. Sheppard, “Dielectric analysis of ther-

moset cure,” in Advances in Polymer Science. Berlin, Germany:Springer-Verlag, 1986, pp. 1–47.

[52] F. Starzyk, “Parametrisation of interdigit comb capacitor for dielectricimpedance spectroscopy,” Archives Mater. Sci. Eng., vol. 34, no. 1,pp. 31–34, 2008.

[53] D. Joneit, M. Nadler, and H. Heuer, “Correction of eddy cur-rent measurements to obtain accordance with simulation results,”NDT & E Int., vol. 62, pp. 167–171, Mar. 2014.

[54] S. Carrozzino, G. Levita, P. Rolla, and E. Tombari, “Calorimetric andmicrowave dielectric monitoring of epoxy resin cure,” Polym. Eng. Sci.,vol. 30, no. 6, pp. 366–373, 1990.

[55] C. Carlini, P. A. Rolla, and E. Tombari, “Measurement method andapparatus for monitoring the kinetics of polymerization and crosslinkingreactions by microwave dielectrometry,” J. Appl. Polym. Sci., vol. 41,nos. 3–4, pp. 805–818, 1990.

[56] X. E. Gros, K. Ogi, and K. Tagahashi, “Eddy current, ultrasonic C-scanand scanning acoustic microscopy testing of delaminated quasi-isotropicCFRP materials: A case study,” J. Reinforced Plastics Compos., vol. 19,no. 5, pp. 389–405, 1998.

Simone Gäbler received the Diploma degreein business and engineering from the DresdenUniversity of Technology (TU Dresden), Dresden,Germany, in 2010, where she is currently pursuingthe Ph.D. degree in mechanical engineering.

She is currently a Research Assistant with theLeibniz Institute of Polymer Research, Dresden, anda Visiting Researcher with the Fraunhofer Institutefor Ceramic Technologies and Systems-MaterialsDiagnostics, Dresden. Her current research interestsinclude the development of eddy current testing

technology for the characterization of polymers and composites.Ms. Gäbler received a scholarship from TU Dresden. She was a recipient

of the Viktor-Klemperer Certificate of TU Dresden and the Award of theFaculty of Business and Economics, both for outstanding performance duringher studies, and the DPG Award for exceptional A-levels in Physics.

Henning Heuer received the Diploma degreefrom the Dresden University of Technology (TUDresden), Dresden, Germany, and the Ph.D. degreein electrical engineering and microelectronicsfrom Brandenburg University of Technology (TUCottbus), Cottbus, Germany, in 2005.

He became a Junior Professor and the Chair ofthe Sensor Systems for Nondestructive Testing atTU Dresden, in 2012. He is also the Head of theDepartment of Sensors and Sensor Systems with theFraunhofer Institute for Ceramic Technologies and

Systems-Materials Diagnostics (IKTS-MD), Dresden. The department focuseson the development of inspection techniques for new generations of materialsand technical structures. With his strong background in semiconductors andtheir packaging and assembly technologies, his team developed ultrasonicphased array sensor for special applications. Also, new solutions for eddycurrent based inspection systems, the high frequency platform EddyCus weredeveloped. The company SURAGUS GmbH is a spin off from his team atFraunhofer IKTS-MD in Dresden. His current research interests include theeddy current and ultrasonic sensors and sensor systems.

Gert Heinrich received the Diploma degree in the-oretical physics from Friedrich-Schiller UniversityJena, Jena, Germany, in 1973, and the Ph.D. degreein polymer physics and the Habilitation degree fromthe University of Technology at Merseburg, Merse-burg, Germany, in 1977 and 1986, respectively.

He taught theoretical physics, polymer physics,and rubber physics as an Assistant Professor andas an Associate Professor with the University ofTechnology at Merseburg, from 1978 to 1990. Hewas a Senior Research Scientist and the Head of the

Materials Research with the Research and Development, Strategic TechnologyCentre, Continental AG, Hannover, Germany, from 1990 to 2002. He hasbeen a Professor of Polymer Materials Science and Rubber Technologywith the Institute of Materials Science, Dresden University of Technology,Dresden, Germany, and the Director of the Institute of Polymer Materialswith the Leibniz Institute of Polymer Research Dresden, Dresden, since 2003.He has contributed to various aspects of the science, technology and edu-cation of polymer materials and rubbers, in particular, statistical theoryof polymer networks, rubber friction and tire materials physics, predictivetesting, reinforcement and its molecular and mathematical basis, and polymernanocomposites.

Iryna Patsora 1, a), Susanne Hillmann 2, Henning Heuer 2, Bryan C. Foos 3, Juan G. Calzada 3

1 Technische Universität Dresden, Fakultät Elektrotechnik und Informationstechnik, Institut für Aufbau- und Verbindungstechnik, Dresden, Germany

2 Fraunhofer Institute for Ceramic Technologies and Systems, Branch for Material Diagnosis (IKTS-MD); Dresden, Germany

3 Air Force Research Laboratory; Dayton, Ohio a) [email protected]

Coatings based on wet particles containing pastes are currently used in many industries, such as automotive, aircraft and/or wind-power plants, to protect carbon-fiber reinforced plastic against damages caused by electrical effects, such as a lightning strike. In order to understand and control the percolation behavior during the drying, a non-contact Eddy Current based Impedance Spectroscopy can be used. This technique can be applied in the wet state of the coating and it works non-destructively. Percolation behaviors of the wet conductive coatings are strongly affected by the type of particles used as a filling and the thickness of the coating. Experimental results of Eddy Current measurements on wet conductive coatings based on different conductive particles and deposited with different thicknesses are discussed. Based on High-Frequency Eddy Current measurements, a prognosis of the coating parameters after final curing during the wet state becomes conceivable. This, for example, offers a wide opportunity for process control and repairs.

Protection against damages caused by a lightning strike for CFRP materials is an important task nowadays, due to their use as aircraft components instead of heavy metals. Until now, the protection is done by an integration of copper mesh over the CFRP’s structure, which weighs down the structure. Moreover, the process is complicated by the fact that it is difficult to apply copper mesh on bending or curved shapes. Wet conductive coatings are a good alternative to the copper mesh, because they are liquid after production, and can be easy applied to any structure. Their final conductivity can be controlled in accordance with percolation theory by variation of the concentration of the conductive particles used as a filler. The main question for usage of wet conductive coatings as protection, is the control of their parameters during drying, such as conductivity and thickness, which change over curing. High-Frequency Eddy Current based Impedance Spectroscopy is chosen as a controlling method, as this technique can be applied in the wet state of the coating and it works non-destructively.

The variation of the parameters of the wet conductive coatings during curing strongly depends upon their chemical composition, coated area, environmental conditions, substrate material, etc. In addition, the measurement is affected by interference from the Eddy Current itself, such as the distance between the sensor and the layer (lift-off).

The aim of this work is to investigate drying behaviors of wet conductive coatings based on different conductive particles by variation of their thickness. Other parameters are selected, so that they do not make an additional

41st Annual Review of Progress in Quantitative Nondestructive EvaluationAIP Conf. Proc. 1650, 414-423 (2015); doi: 10.1063/1.4914637

© 2015 AIP Publishing LLC 978-0-7354-1292-7/$30.00


contribution to the measurements. Thus, an isolation substrate is used and coated area and lift-off are constant, and the coatings are cured at room temperature.

This paper presents experimental results of the High-Frequency Eddy Current measurements on wet conductive coatings based on two types of filling material having different thicknesses. Correlations are shown between the High-Frequency Eddy Current measurements and the final parameters of the coatings, which allow characterization of the drying behaviors of wet conductive coatings at different curing times. Based on the results, a prognosis of the coating parameters after final curing during the wet state becomes conceivable. This offers a wide opportunity for process control and repairs.

Electrically conductive pastes consist of a matrix of the carrier material (polymer matrix) filled with electrically conductive particles (mostly silver, but other electrically conductive materials are also used). The formation of the conductive areas in defined structures with a random distribution of the conductive particles in the matrix is described by the percolation theory [1]. From a certain amount of conductive particles in the polymer matrix, the individual particles are in contact with each other and form individual interconnections. The electrical conductivity of the system is realized when the conductive particles come into contact with each other, building a network of conductivity passing through the entire volume. After depositing, the coating is soft and mostly non-conductive. The polymer matrix is hardened during the drying of the coating, whereby the filler particles are fixed in their position. Through the evaporation of the solvent and crosslinking processes, the layer can shrink during the drying. The shrinkage process improves final electrical conductivity of the coating, because particles are more strongly pressed together.

Schematic of the percolation behaviors in wet conductive coatings

The drying process for wet conductive coatings is shown in Fig. 1 and can be described (simplified) as follows.

First, the particles are uniformly distributed in a volume. At this time, the layer has a maximum thickness, while forming no or a very low electrical conductivity. During the drying of the coating, the polymer matrix shrinks depending on its composition, more or less. Drying decreases the layer thickness and increases the conductivity of the layer, because conductive particles are pressed together by the shrinkage of the layer and come into better contact with each other. At that time, the drying is complete. The layer has reached its maximum electrical conductivity and its minimum thickness. During the drying of conductive coating, two parameters are changing, i.e. the thickness of the coating decreases and its conductivity increases. Both parameters affect the Eddy Current signal. According to the functional capability of the wet conductive coatings, the final sheet resistivity is a combination of their final thickness and final conductivity. As the extension of the layer is only possible in a liquid state of the coating, the challenge of this project is to perform Eddy Current measurements on coatings in a liquid state and to use these measured values to predict their final sheet resistivity and thickness [2, 3, 4, 5].


By the High-Frequency Eddy Current based Impedance Spectroscopy the penetration depth is derived as follows:

(1)

where is the angular frequency , µ is the permeability and is the electrical conductivity. The induced voltage is:

(2)

where is the rate of change of magnetic flux. Equation (1) states, that increasing the frequency causes decreasing of the penetration depth and equation (2)

shows that increasing of the frequency increases the signal of the amplitude. This means that the lower the conductivity or thickness of the specimen, the higher the frequencies needed for higher sensitivity [6].

Drawing of the High-Frequency Eddy Current based testing system

Figure 2 illustrates, that High-Frequency Eddy Current based on the EddyCus system operates in frequency

range between 100kHz to 100MHz.

Polymer matrix: Epotek301; Coupling Agent: 3-glycidoxypropyltrimethoxy silane; Solvent Agent: Alcohol 94% and n-Butyl acetate 99%; Conductive filling: silver-coated copper and

silver-coated glass particles (allocated by Potters Industries). Substrate: ceramic. Silver-coated copper-based layers provide high final conductivity and silver-coated glass based layers a low final

conductivity. Thus, a wide range of conductivities are investigated and analyzed using the High-Frequency Eddy Current spectroscopy.

Ceramic substrate is chosen because it is non-conductive, so the High-Frequency Eddy Current based impedance measurements are influenced only by conductivity and thickness changes of the wet conductive coatings during curing.

After preparation, wet conductive coatings are deposited on a ceramic substrate using the frame printing technique (Fig. 3). The thickness and coated area of the coating depends only on the thickness and form of the frame. The used copper frame has the same size as a ceramic substrate of 70mm×70mm. In the center of the copper frame, a rectangular opening of 40mm×40mm is etched; with this, the coated area is provided and is the same for all


deposited coatings. The thickness of the frames are 80µm, 160µm and 240µm. Wet conductive coatings are applied to the substrate using a scraper.

Schematic of the Frame Printing Technique

After application, wet conductive coatings are placed under the High-Frequency Eddy Current sensor and are

cured 24 hours at room temperature and under normal conditions. At the same time, Eddy Current measurements are being performed at 30 second intervals over 24 hours (from beginning to the end of the drying); the lift-off is 80µm.

Reference measurements were performed to analyze the wet conductive coatings during drying, because the thickness of the coatings deviates from a desired value by a manual application process used in this work.

Silver-coated copper-based layers on ceramic substrate with coated area of 4×4cm

Thin d = 80µm d1-1 = 56-61µm RF1-1 = 71.67-138 m / Middle d = 160µm d1-2 = 97-117µm RF1-2 = 28.05-43.18 m / Thick d = 240µm d1-3 = 138-160µm RF1-3 = 22.7-27.16 m /

Silver-coated glass-based layers on ceramic substrate with coated area of 4×4cm

Thin d = 80µm d2-1 = 46-69.6µm RF2-1 = 436.3-934.8-138 m / Middle d = 160µm d2-2 = 94-98µm RF2-2 = 287.6-338.3 m / Thick d = 240µm d2-3 = 105-119µm RF2-3 = 164.7-267.1 m /

The reference measurements were performed on the wet conductive coatings after curing using the Laser

Profilometer for thickness measurements and the four-point probe method (multimeter Keithley) for resistivity measurements [7] and are given in Tables 1 and 2 above.

High-Frequency Eddy Current based impedance measurements were performed using a laboratory system based on the EddyCus® Systems developed at Fraunhofer IKTS-MD that consists of a special sensor for contacting to liquid layers, an electronic system for controlling the measurement and software for data analysis. The sensor allows automatic data acquisition over multiple frequencies from 100kHz to 20MHz. The sensor is integrated into a precision positioning table that allows an exact adjustment of the lift-off with an accuracy of 10µm.

Obtained data was analyzed using MatLab and are represented in different dependences:

Real part of the complex voltage over the drying time; Imaginary part of the complex voltage over the drying time; Real part of the complex voltage over imaginary part of the complex voltage.


Drying curves rotate depending on the frequency used by Eddy Current based impedance measurements. It was found by experiments that some of the frequencies used in a sweep provide an opportunity for separating coatings with different thicknesses in the wet state.

Figure 4, for example, illustrates High-Frequency Eddy Current based impedance measurement results represented in a logarithmic scale as a dependency of the real part of the complex voltage of the drying time at a frequency of 10MHz for silver-coated copper-based coatings deposited in three different thicknesses: thin, middle and thick. Coatings marked in the same color were deposited using the frame with the same thickness. It is visible that all curves are shifted from each other, which is caused by a deviation of the thickness from a desired value as confirmed by reference measurements given in Table 1.

Drying behaviors of the silver-coated copper-based coatings at a frequency of 10MHz represented as the real part of

the complex voltage over the drying time

Schematic of the drying behaviors of the silver-coated copper-based coatings at a frequency of 10MHz represented

as the real part of the complex voltage over the drying time Figure 5 explains in more detail the drying process of silver-coated copper-based coatings. It is clearly seen that

the Eddy Current signal initially rises shortly before it sharply drops, while the conductivity of the layer is increasing


dramatically; this period of the drying is a percolation threshold. This narrow peak just before the sharp drop is called the "characteristic point". It was established that the amplitude and the time of the characteristic point have correlations to the final sheet resistivity and thickness of the coatings. After the percolation threshold (for coatings used in this work it happens after 70 minutes after deposition), properties of the wet conductive coatings are changing slowly until the polymerization process is completed.

Using the reference measurements given in Table 1, correlations were carried out that are presented in Fig. 5 thru 9 below.

Correlations between the time of characteristic point at a frequency of 10MHz and the final thickness for silver-

coated copper-based coatings

Correlations between the amplitude of the real part Re(U) of the complex voltage at the characteristic point and the final sheet resistivity for silver-coated copper-based coatings

From Fig. 6 and 7 it can be obtained that:

The thinner the layer, the earlier a characteristic point is observed; The lower the final sheet resistivity of the layer, the lower is amplitude of the characteristic point.


The final parameters of the wet conductive coatings can be predicted a couple of minutes after depositing, but knowing an exact time of the depositing and performing Eddy Current based impedance measurements continues until the characteristic point has occurred.

Figures 8 and 9 below show correlations for silver-coated copper-based coatings after 70 minutes and at the end of the drying. It is seen, that different frequencies should be used for coating analysis at different curing times. There is also another correlation after a percolation threshold between the amplitude of the real part of the complex voltage and the final sheet resistivity at a frequency of 10MHz for these coatings, but providing a little lower accuracy. This correlation is shown later in “Comparative Analysis”.

Correlations between the amplitude of the real part Re(U) of the complex voltage at a frequency of 13MHz after

70 minutes of curing and the final sheet resistivity for silver-coated copper-based coatings

Correlations between the amplitude of the imaginary part Im(U) of the complex voltage at a frequency of 10MHz after curing and the final sheet resistivity for silver-coated copper-based coatings

Consequently, combining the frequencies and using the correlations mentioned above, the entire drying process

can be controlled while the final properties of wet conductive coatings can be predicted in liquid and wet states. The presence of the characteristic point for silver-coated copper-based coatings provides an opportunity for separating characterization of the thickness and sheet resistivity at the characteristic point.


Drying behaviors obtained by High-Frequency Eddy Current based impedance measurements on silver-coated glass-based coatings having different thicknesses are represented as a real part of the complex voltage at a frequency of 10MHz and are shown in Fig. 10. It is seen that drying curves for silver-coated glass-based coatings have the same characteristics after a percolation threshold as silver-coated copper-based coatings. However, at the percolation threshold they do not provide any characteristic point. This means, that the percolation behaviors for coatings based on copper and on glass particles are different. It was established, that only coatings having a characteristic point can be characterized in a liquid state (at percolation threshold). Thus, it is not possible to characterize and control silver-coated glass-based coatings until the percolation process is completed.

Schematic of the drying behaviors of the silver-coated glass-based coatings at a frequency of 10MHz represented

as real part of the complex voltage over the drying time Nevertheless, there is an opportunity to characterize silver-coated glass-based wet conductive coatings

70 minutes after deposition and until the end of the drying. Figure 11 shows corresponding correlation.

Correlations between the amplitude of the real part Re(U) of the complex voltage at the frequency of 10MHz after

70 minutes of drying and the final sheet resistivity for silver-coated glass-based coatings


Correlations for silver-coated glass-based coatings at the end of the drying look similar to correlations after 70 minutes after deposition. Based on Fig. 11 above it follows that the thinner the layer and the higher the final sheet resistivity, the lower is the amplitude of the real part of the complex voltage. The analysis of the drying behaviors of the silver-coated glass-based wet conductive coatings at each frequency used in a sweep has revealed that only one frequency of 10MHz is good to be used for the characterization of coatings.

A comparative analysis was performed for all wet conductive coatings measured and analyzed in this work. First, the accuracy of High-Frequency Eddy Current-based impedance measurements was evaluated. An absolute deviation and a relative deviation factors are given in Table 3. It is seen that the lower the sheet resistivity of the conductive coatings, the higher is the accuracy of High-Frequency Eddy Current based impedance measurements.

Accuracy of High-Frequency Eddy Current based Impedance Measurements

At the characteristic point No characteristic point

70 Minutes

At the end of the drying

Using the High-Frequency Eddy Current based Impedance Spectroscopy, the prognosis of the final sheet

resistivity for coatings based on silver–coated copper and silver coated glass particles in the same plane becomes conceivable, because they both have correlations between the final sheet resistivity and the amplitude of the real part of the complex voltage at a frequency of 10MHz at the end of the drying (Fig. 12).

Correlations between the amplitude of the real part Re(U) of the complex voltage at a frequency of 10MHz at the

end of the drying and the final sheet resistivity for both, silver-coated copper-based (red) and silver-coated glass-based (green) coatings

The next opportunity provided by High-Frequency Eddy Current based impedance measurements is the

separation of coatings based on different filling materials. In this case, all drying curves are represented in the complex voltage plane at a frequency of 10MHz.

Figure 13 illustrates that drying curves for coatings based on different filling materials are shifted in different ways in the complex voltage plane, whereby they are clearly separated from each other. In this case, we cannot


separate coatings depending on the thickness or on the sheet resistivity but only depending on the type of the filling material.

Drying behaviors of the silver-coated copper-based (red) and silver-coated glass-based (green) coatings at a

frequency of 10MHz represented in the complex voltage plane

Data and correlations obtained and presented in this paper will be used in a future project for the development of an algorithm to allow the prognosis of the final parameters for wet conductive coatings based on different conductive particles, deposited in different thicknesses and on different substrates at any time of the curing.

Drying behaviors of wet conductive coatings based on silver-coated copper and silver-coated glass particles having different thicknesses were analyzed in this work using the High-Frequency Eddy Current based Impedance Spectroscopy. Correlation allowing the characterization of the drying behaviors over drying time are shown and discussed. It was established that:

Based on different filling particles, conductive coatings have different drying behaviors, which can be characterized by the High-Frequency Eddy Current based Impedance Spectroscopy. High-Frequency Eddy Current based Impedance Spectroscopy allows evaluating the final sheet resistivity of the conductive coatings in a wet state. High-Frequency Eddy Current based Impedance Spectroscopy provides an opportunity for the evaluation of the sheet resistivity in a wide range (e.g., 22.7m / – 934.8m / , as shown in this paper). The lower the sheet resistivity of the conductive coatings, the higher is the accuracy of High-Frequency Eddy Current based impedance measurements.

1. Stauffer D. and Aharony A. 1991 Introduction to Percolation Theory (London: Taylor & Francis). 2. I. Patsora; H. Heuer, S. Hillmann, D. Tatarchuk, Bryan C. FOOS, “Experimental setup for the characterization

of the percolation behavior of wet conductive coatings by high frequency Eddy Current spectroscopy”, 978-1-4799-0036-7, 2013 IEEE, 36th Int. Spring Seminar on Electronics Technology.

3. S. Hillmann, M.Klein, and H. Heuer, „In-Line thin film characterization using eddy current techniques“, Studies in Applied Electromagnetics and Mechanics Volume 35, ISBN 978-1-60750-749-9, 2011.

4. S. Hillmann “Evaluation eines zerstörungsfreien Prüfverfahrens zur Ermittlung des Flächenwiderstandes flüssiger, leitfähiger Schichten; Dresden International University, Studienrichtung „Zerstörungsfreie Prüfung“; Masterarbeit, 2013.

5. D. Lu, Q. K. Tong, C. P. Wong, “Conductivity Mechanisms of Isotropic Conductive Adhesives (ICA`s)”, IEEE Transactions on electronics packaging manufacturing, Vol. 22, No. 3, July 1999, pp. 223-227.

6. H. Heuer, M.H. Schulze, N. Meyendorf, „Non-destructive evaluation (NDE) of composites“, Non-destructive evaluation (NDE) of composites: eddy current techniques, pp. 33-35.1.

7. F. M. Smits, “Measurement of sheet resistivities with the four-point probe”, The bell system technical journal, May 1958, pp.711 718.


advanced eddy current imaging - トップページ | tia institute for ceramic technologies and...

Documents