NDT METHODS FOR EVALUATINGCARBON FIBER COMPOSITES

RÚNAR UNNÞÓRSSONa, M. T. JONSSONa AND T. P. RUNARSSONb

aDepartment of Mechanical Engineering, University of Iceland, IS-107 Reykjavik, IcelandbScience Institute, University of Iceland, IS-107 Reykjavik, Iceland

Abstract

A review of nondestructive testing methods is presented and their potentiality for monitor-ing composites, undergoing fatigue testing, in realtime is discussed. Fatigue testing of com-posites can take very long time, up to several weeks. The purpose of monitoring is to detectincipient faults and to predict the remaining fatigue strength of the composite. If the predictednumber of cycles-to-failure is unacceptable, the test can be stopped and considerable time saved.Two approaches stand out: methods based on acoustic emission and the electrical properties ofcomposites. The potential of these two methods is discussed in greater detail.

Keywords: NDE, NDT, Fatigue testing, Carbon Fiber Composites, Acoustic Emission, Electricalproperties.

1 Introduction

Testing methods can be roughly divided into two groups depending on their purpose. The firstgroup consists of methods that are used for determining material properties and they are mostoften destructive. The second group, on the other hand, contains nondestructive methods whichare used for condition evaluation.

Material properties are important for engineers. In order to be able to select the right materialand carry out calculations during the design process, reliable properties must be available. Manymaterial properties are found by damaging samples of a material in standardized tests. Examplesof such properties include: hardness, tensile strength, Young’s modulus, Impact strength, fatiguecharacteristics, corrosion resistance, UV-stability and more. Many properties are dependant ontemperature and therefore their behavior at different temperatures must also be known. Eventhough this information is vital for the design process, it isn’t always sufficient for determininghow the product will perform when put to use. The reasons for this can be due to theoreticallimitations and the fact that standardized tests do not completely determine the material prop-erties. This is especially relevant for new material types such as composites. Therefore, duringthe design process prototypes must be built and tested in order to see if the design fulfills therequirements.

Even though a product passes all tests and requirements, quality inspection tests must also beperformed. The need for monitoring does not end after the product has been sold, in fact it needsregular inspection while in service. The inspection methods can vary, depending on the productand its usage. Most are satisfied with inspecting their home appliances by trying them. However,few are willing to fly in an aeroplane that is tested by such inspection methods. It can thereforebe said that the extent of the inspection is proportional to risk and cost.

This paper provides a review of methods for nondestructive testing. The review is aimedat identifying testing methods that can be used for realtime condition monitoring of compositesundergoing fatigue testing. The purpose of condition monitoring is to detect incipient faultsand predict the remaining fatigue strength of the composite, i.e. the number of cycles-to-failure.Because many nondestructive testing methods utilize some material properties, an overview ofcarbon fibers and carbon fiber composites is necessary. Section 2 gives an overview of carbonfibers, their properties and the types of faults/damages that can be found in composites. Section3 starts with a brief review of the main nondestructive testing methods for carbon fiber compositesand then two NDT methods who are believed to be more suitable for use with fatigue testing are

1

discussed in greater detail. These two methods measure Acoustic Emission signals and electricalproperties of the composites. The paper concludes with some final remarks and comments.

2 Carbon Fiber Composites

Approximately 80 years passed from when Sir Joseph Wilson Swan, in 1878, and Thomas AlvaEdison, in 1879, independently produced incandescent lamps with carbon filaments [1] to whencarbon fibers became a commercial product. The need for lighter and more heat resistant buildingmaterial for the military and space applications was the main reason for increasing interest incarbon fibers [2, 3]. The interest was an impetus for improving production methods [4] which inturn made it possible to manufacture cheaper fibers. As a result of this, general demand increased.According to the AMPTIAC1 [5] newsletter, figures released from SACMA2 show that the worlddemand increased from 5.850 tons in the year 1991 to 10.368 tons in the year 1998. The figures arebased on production numbers from SACMA members who are believed to stand for 85% of thedemand in North America, Western Europe, and the Far East. Currently, the world demand isaround 20.000 tons and is expected to increase 7-8% per year in the next years [6]. Carbon fibersare used in all types of vehicles: aeroplanes, cars, ships, bicycles and more. Carbon fibers havealso been used extensively for the production of recreational products such as: golf clubs, tennisracquets, fishing rods and many more. What is it that makes carbon fibers so appealing? Whatare its weaknesses? The answers to those questions will be provided in the following section.

2.1 Carbon Fiber

Carbon fibers can be made from many types of precursor materials, but rayon, PAN and pitchhave been used when mechanical properties are important [4]. Rayon was used in the first gen-erations of carbon fibers, but due to expensive production methods and high mass losses it isn’tpopular today [7].

PAN, or polyacrylonitrile, is the most used precursor material for making carbon fibers [4].The production of a fibre starts with stretching of the PAN until it becomes a very thin fibre (6-10µm) and at the same time it is oxidized in air at 200-300◦C [8]. During the oxidization the fiberbecomes black and its material structure changes; hydrogen atoms disappear and oxygen atomsare added. When the core of thick fibers doesn’t oxidize the fibers will become hollow later inthe production. Hollow fibers have higher strength to weight ratio. The oxidization process is thekey limiting factor in the production, even though the time required has been reduced from hoursto minutes [4]. The next step is carbonation, where the fibers are heated up to 1000-3000◦C [3],but now in an inert gas. Nitrogen is often used for temperatures up to 2000◦C and argon forhigher temperatures [4]. The exact temperature depends on the properties required. In general,the stiffness increases with higher temperature.

According to Landis et al [9], volatile materials are released during the heating process result-ing in a material that is almost pure carbon, or 92-99%. If the temperature goes over 3000◦C thenthe fibers turn into graphite. Graphite is a soft material and has been used as an dry lubricant orpencil lead. The difference between carbon and graphite fibers lies in how the carbon atoms areconnected. In graphite fibers the atoms are arranged in sheets. In each sheet they are connected inhexagonal structure, similar to chicken wire. The bonds between the atoms in each sheet are verystrong but, the bonds between adjacent sheets are very weak. The weak bonds will break understress and the sheets slide past one another. This is what makes graphite a good lubricant. Thestructure of the atoms in carbon fibers is more complicated. It looks like the sheets have been laidin the direction of the fiber and then crumpled. This results in complicated interaction betweenthe sheets and hence it is more difficult for the them to slide.

After the carbonization process the fibers are surface treated and sized. These two last stepsinfluence how the fibers will perform in a composite. By surface treatment the surface area of thefibers can be increased which results in better bonding [10]. The purpose of sizing is to protect

1Advanced Materials and Processes Technology Information Analysis Center.2The Suppliers of Advanced Composite Materials Association

2

Property Steela Aluminumb E-Glassc Carbon Fibersd

Density [g/cm3] 7.85 2.6-2.8 2.54-2.6 1.75-1.8Tensile strength [MPa] 276-1882 230-570 3448 3530-6370Elongation at Break [%] 10-32 10-25 4.8 0.7-2.1CTE [10

−6/K] 11-16 20.4-25.0 5.4 (-1.1) - (-0.38)Thermal conductivity [W/m-K] 24-65 237 1.3 10-150Specific heat [J/g-◦C] 0.45-2.1 0.9-0.96 0.81 0.71-0.75Melting point [◦C] 1500 477-660 1725 3650Resistance [ohm-cm] 1.74×10

−5 5.2×10−6 4×10

12 0.9-1.6×10−3

aCarbon Steel, www.efunda.com & www.weldtechnology.combwww.matweb.comcwww.matls.comdwww.toray.com & www.goodfellow.com

Table 1: Comparison of carbon fiber properties with its rival materials.

them from corrosion, handling and to also improve the resin bonding [11, 3]. If too strong bondingis achieved then the impact resistance of the composite will decrease [12].

The production of carbon fibers using pitch precursor is much less than using PAN. The pitchcan be either isotropic or liquid crystalline [3]. The oxidization, carbonization and surface treat-ment of pitch based fibers is analogous to PAN based fibers [13]. The main advantage that pitchbased fibers have over their PAN based counterparts is that the precursor material is cheaper [4].The fibers have lower tensile and compressive strength. These properties can be improved, but itis expensive [4].

As was mentioned previously, the demand for carbon fibers has increased during the lastdecades. The main reason for this increase is because of their material properties. Table 1 com-pares the material properties of carbon fibers with their rival materials. As the table shows, carbonfibers possess some desirable properties, and are well suited for applications where strength, stiff-ness, weight and damping are important. Carbon fibers are a good choice when corrosion andheat resistance is required. Finally, when strength to weight ratio is important then carbon fibercomposites are an excellent choice.

When used in composites, carbon fibers are used as an reinforcement material. The com-posite will have more strength than the matrix (cured resin) and the fibers have independently.Carbon fibers have been used to reinforce various materials, e.g. cement, aluminium, PEEK(PolyEtherEtherKetone) and epoxy. The discussion will be limited to carbon/epoxy laminatedcomposites. The function of the epoxy matrix is to distribute stress between the fibers [14]. Thematerial properties of an epoxy resin are vastly different from the fibers, e.g. the CTE of an RS-1epoxy from YLA Inc. is 32 × 10

−6/K and the tensile strength is 80MPa [15].Laminated composites are constructed by adding thin layers of sheets and resin. The fibers

come as chopped strand mats, woven and unidirectional sheets. Chopped strand mats consists ofchopped fibers that are evenly distributed and randomly orientated. The fibers are held togetherwith a binder. The strength of the mats is equal in all directions in the plane of the laminate.Woven sheets are formed from interlacing yarns. They are strongest in the direction of the fibers,i.e. two directions. The orientation of all fibers in a unidirectional sheet is in the same direction.Their strength is therefore almost all concentrated in that direction. The mats and sheets aresometimes pre-impregnated with resin before being stored.

The strength of an composite depends on the ratio of fibers versus matrix. The volume frac-tion of carbon fibers usually lies between 40-60%. Properties of composites usually depend on thedirection in which they are measured. This is called anisotropy. Metals are generally isotropic,which means that their strength and other properties are independent of direction. Compositesoffer therefore more design parameters, which is something that most designers welcome. Thethree different types of sheets offer many interesting possibilities for a designer. The stacking se-quence and orientation of sheets can be used to obtain desired properties. It is possible to designan composite that can be easily bent, but hard to twist and vice versa, depending on the orienta-tion of the fibers. Also, it is possible to design an composite that shows negligible heat expansion

3

over certain temperature interval. This is because the coefficient of thermal expansion for carbonfibers is negative but positive for epoxy matrix [3, 13]. However, carbon fiber composites and theirmanufacturing methods are more complicated than for metals. The number of damage types thatcan occur in composites is also higher, we will next take a look at them.

2.2 Faults and Defects

The reasons why damage occurs can be diverse. But, once a damage has occurred it is morelikely that further damage will occur. This is because the damage will weaken the composite. Thedamage of a composite and its components can roughly be attributed to three different stages intheir life; during the manufacturing of the fibers, during the construction of the composite andduring the in-service life of the composite. In the following, an overview is given of the mostcommon types of damage found in composites. The organization is based on the origin of thedamages, i.e. the three different stages of life.

During the production of carbon fibers, the fibers travel at high speed trough the machineryand occasionally they rub against it [16]. The sizing that is put on the fibers is intended to pro-tect them from abrasion, but if they get scratched then the sizing gets damaged and the fiberwill loose strength. If the sizing fails then the bonding between the fiber and the matrix is morelikely to break. Debonding (Figure 1) of the fiber/matrix interfaces will render the matrix un-able to distribute stresses to the fiber, which results in reduced stiffness and different dampingcharacteristics of the composite [17]. According to Schwartz [18] matrix cracks that intersect afiber and entrapped moisture can also cause debonding. During the construction of a laminated

Figure 1: Shows matrix cracks, broken fibers, debonding and delamination.

composite, it is important to remove wrinkles when new layers are added. Wrinkles (Figure 3)can cause air entrapment and resin buildup [19]. They can therefore weaken the composite andrender the design useless. Air entrapment (Figure 2), sometimes called voids, between the fiber

Figure 2: Shows when air gets trapped between layers.

layers. Voids can occur because air gets trapped between layers during the lay-up process. Ac-cording to Schwartz [18] voids and porosity can be caused by volatile entrapment during curingof the resin. Schwartz also remarks that if they are only partially entrapped then blisters aregenerated. Blisters are generated in the outermost layers. Voids and blisters weaken the com-posite and they can also induce the formation of other types of damage. Foreign objects that getentrapped between layers weaken the composite (Figure 3). Examples of foreign inclusion are:dust, hair, grease and other impurities. According to Adams and Cawley [19] stresses can de-velop around foreign inclusions. Those stresses can cause delamination and other damage. This

4

Figure 3: Shows foreign inclusion and a wrinkle.

has been used in research in order to manufacture composites with damage [20, 21, 22, 23]. It isimportant to use suitable resin for the application, because temperature, stress and moisture cancause damage to the matrix [19]. The ability of the matrix to distribute stresses can change duringfatigue loading, hence affecting the endurance limit of the composite [17]. Density variations, dueto either too little or too much resin, can have serious consequences for the composite. Accordingto Gaylord [24] too little resin results in inadequate bonding between layers and the formation ofvoids and porosity. Too much resin lowers the volume fraction of fibers and increases the risk ofcracks (Figure 1). If the resin cure temperature is too high or the resin cures too fast then crazingcan occur [24]. It is characterized by very fine cracks on the surface and in the matrix. Crazinggenerally occurs where excessive resin or gel coat is located. The orientation of the fibers is oneof the composite design parameters. Fiber misalignment can therefore change the design andgenerate different load distribution than was anticipated by the designer. In some cases this willcause the matrix to take up the load and the composite will break. During the curing of the resin itwarms up and its volume reduces due to polymerization shrinkage but the fibers are unaffected.This generates internal stresses between the matrix and the fibers [25]. When the composite coolsdown, additional stresses are generated because the matrix and the fibers have different coeffi-cient of thermal expansion [25]. The stresses generated are called residual stresses. The residualstresses affect several properties of the composite, e.g. strength, fatigue and chemical resistance.The stresses can be reduced by extending the curing time.

Most composites require some kind of post processing, e.g. drilling and cutting. Those oper-ations can easily cause damage to the composite. The most common reason for damage is dueto forces that are applied perpendicular to the direction of the fibers. But, composites have verylittle strength in that direction [26]. Delamination (Figure 1) is when the bonds between layersbreak. It is the most common damage in composites [27, 28, 29]. The weakest bonding betweenlayers is where voids and too much resin is located. When in service, the weak bonding betweenthe layers can break, which reduces stiffness and buckling resistance [20]. Therefore, both toomuch and too little resin increases the risk of delamination. Delamination can also be caused byimpacts [30]. Carbon fiber composites are sensitive to low energy impacts. [31]. The impacts canbreak fibres, cause delamination and generate cracks [32, 20]. Cracks influence the composite’schemical resistance and strength. Cut or broken fibers weaken the composite. Fiber failure canbe attributed to improper handling, imperfections, and both tensile and compressive stresses.Stresses can develop around the location where the fiber breaks. These stresses can cause otherfibers to break.

In next section, we will look at nondestructive test methods that can be used to detect theabove mentioned damages.

3 Non-Destructive Methods

Many nondestructive testing methods (NDT) exist that can potentially be used for detecting dam-ages in composites. Some of them are cheap and simple, while other are more expensive and com-plicated. Cost is however, not a very good criterion for selecting an method. The testing methodmust be selected after considering several factors, e.g. the cost involved, the time required, theaccess to the composite, and the ability of the method. The cost includes investments in mea-surement equipment, software and the training of personnel. All NDT methods require trained

5

personnel for interpreting the test results. They are also required for setting up, calibrating andoperating the equipment.

This section starts with an overview of the main NDT methods used for damage detectionin carbon fiber composites. Then two NDT methods that are more suitable for use with fatiguetesting are reviewed. These methods use Acoustic Emission signals and the electrical propertiesof composites.

3.1 Overview

Perhaps the simplest nondestructive testing method for composites is coin tapping [33]. Thismethod, which is still used to day, involves tapping the composite with a coin, or a metallic object,and listening for changes in the sound. A change indicates that a damage is presented. Damagessuch as delamination and voids can be detected with this method. The accuracy depends on theoperator and therefore it is not a very reliable method. Also, testing can be very time consumingsince, in order to fully check an composite, the operator must must perform an exhaustive search.

Vibration based methods can be used to measure the frequency response of composites. Themethods can detect stiffness reduction which is evenly distributed over the composite, such stiff-ness reduction is likely to pass undetected by other test methods [34]. A change in natural fre-quencies is an indicator of changed stiffness. In a paper by Hou and Jeronimidis [35] it is statedthat little research has been conducted to investigate the influence of delamination on frequencyresponse. According to Zou et al [36] model based techniques are sometimes used in order to im-prove the method. They also point out that changes in natural frequencies do not always provideenough information to enable identification of damage type. In a paper by Kessler et al [34] it ispointed out locating a damage is difficult from the measured signal. It follows that determiningthe orientation and the severity is also difficult. They mention that the accuracy of the methoddepends on the resolution and the frequency band of the sensors together with the number ofsensors and their locations. They also bring up an interesting idea; use vibrations from the sur-rounding to excite the composite and use it to monitor the composite online. Vibration basedmethods can also be used for realtime monitoring [37].

Visual tests are also simple tests, performed by visually examining the composite. Visual testhave the advantage of not touching the composite and they are easy to execute. However, likethe coin tapping, the accuracy of the method depends on the operator. In order to improve themethod, dye-penetrant is used. A dye is sprayed on the composite and it gets drawn it into thecracks and pores because of surface tension. In some cases cracks close when a load has beenremoved from the composite and the dye will not enter those cracks. After a certain dwell time,the composite is wiped clean and a developer is applied. The developer, usually a dry whitepowder, draws out the penetrant from the cracks and pores. Visual inspection is then performed.

Many instruments and methods have been designed in order to make visual tests more reliableand tractable. Such methods are often called optical tests. During the last few years, two methodshave received increasing attention. They are Digital Speckle Photography (DSP) and ElectronicSpeckle Pattern Interferometry (ESPI). Both use charge-coupled devices (CCD) silicon chips fortaking images. The main reason for the increasing popularity of these methods is the rapid devel-opment of both algorithms and CCD silicon chips. In DSP a pattern is projected on the surface ofthe composite. An image is taken at the start of the monitoring. It is then subtracted from all sub-sequent images. This will generate pattern which is used to detect changes [38]. If two camerasare used, looking at the composite from different angles, then displacements in three dimensionsare obtained [20]. The ESPI method uses a laser light source for the generation of the pattern.The light is pointed at the composite and the reflection generates a pattern. Both methods useimage processing algorithms for analyzing the images. They are relatively cheap, tractable andcan produce results in a short period of time. The methods are sensitive to surface displacements,which can be exploited to detect damages in their early stages. The accuracy is limited to theresolution of the CCD silicon chip [39] and the ability of the image processing software. Opticalmethods use information obtained from the surface of the composite which makes them limitedfor evaluation. Many serious damages, e.g. delamination due to impact, can occur well belowthe surface of the composite [30]. According to Komorowski et al [31] cyclic loading, moisture,temperature and viscoelastic effects can reduce impact dents depths up to 45%. This means that

6

optical methods will in many cases, not be able to detect such damages at their early stages.Thermal Infrared testing involves taking thermal images of a composite, which can provide

information about its inner structure. Infrared cameras with focal plane array (FPA) sensors isone method used for capturing the thermal images. The capturing and post processing of imagesis analogous to DSP and ESPI, except FPA consists of sensitive, semiconducting heat-sensors andthe CCD chip is made up of light sensors. This means that the accuracy of the infrared methodis limited by the same factors as mentioned above. Also, the methods require a vibration freeenvironment. The advantage is that they are non-contact methods. With good equipment, ther-mal infrared inspection can detect voids, foreign inclusions, delamination and impact damage inrelatively short time [40].

In ultrasound testing pulses of high-frequency sound-waves are transmitted into the compos-ite and the time it takes for them to bounce back is measured. The method allows the detection ofvoids and delamination deep inside the composite [33]. The main technical difficulties associatedwith the method are the attenuation, scattering and absorption of the signal and also shadow-ing effect of multiple damage. Those difficulties make it difficult to test specimens which arethick or of a complex shape. According to Hosur et al. [30] the effect of attenuation and scat-tering can be reduced by using the appropriate probe and adjusting the pulse parameters. But,the method is time consuming [41] and the need for coupling medium such as water or gel isan disadvantage [42]. There are other ways of transmitting the sound into the composites, e.g.roller probes [43] and air-coupled transducers. The problem with air-coupled transducers is thedifference how air and the composite transmit sound [42].

With radiographic methods delamination, cracks and foreign inclusions can be detected, andalso density and thickness variations [43]. Both residual stresses and load induced stresses canbe estimated with the method [44]. The resolution of the method lies between 20-200µm [45, 46],depending on what radiographic technique is used, and damages located at depth of tens ofmillimeters can be detected [46]. The major drawbacks are high cost, strict health regulations [33]and the fact that it only detects damage that is parallel to the beams [43]. Some radiographictechniques use dye penetrant for enhancement. One of the main problems involved with usingdye penetrant is when the penetrant does not enter a crack or does not reach the end of it [47].This means that internal cracks and delamination that do not extend to the surface will not beenhanced. Also, in some cases cracks close when a load is removed and the dye will not enter.

Eddy Current based testing methods are sensitive to the volume fraction of fibers, brokenfibers [43] and impact damage [48]. Limitations include: limited penetration depth, surface mustbe accessible, damage parallel to the scan direction are undetectable [49]. Eddy current can beused to estimate damping characteristics of an composite [50]. Zhang and Hartwig [17] recom-mended that damping characteristics be used, instead of stiffness, for evaluation of damage. Theysuggested this because their observations indicated that damping was more sensitive than stiff-ness.

There are many other nondestructive techniques based on the above methods. However, weare interested in realtime evaluation of the fatigue strength of an composite undergoing fatiguetesting. Most of the above mentioned methods require that the test be temporarily halted inorder to take measurement. Vibration based methods can be used in realtime. However, due topractical reasons, i.e. excitation source is needed and noise resides in the same frequency band asthe signal, we consider it unsuitable for fatigue testing of composites. We will now take a detailedlook at two NDT methods are believed to be more suitable.

3.2 Electrical Properties

Electrical conductivity is an interesting property of carbon fibers. When the fibers are used incomposites, the matrix acts as an insulator. However, during the construction of a laminate com-posite the resin will flow and the fibers can move and touch each other [51]. Since the fibers areserpentine [52] they only touch at contact points (Figure 4). In practical applications perfect in-sulation between the fibers is not obtained, hence the composites will conduct electricity in alldirections [52]. Nevertheless, the electrical resistance is much higher perpendicular to the fibersthan along them. Composites are therefore anisotropic conductors. Since the contact points arerandomly distributed the conductivity will vary. According to paper by Louis et al. [53] the num-

7

Figure 4: Shows how fibers touch.

ber of contact points will increase with the volume fraction of fibers. If the fraction goes below acritical value then perfect insulation between the fibers is obtained, but such composite will havelittle strength. Schueler et al. [54] measured the anisotropic resistivity of an composite by usingthe ratio of electrical resistance in the direction of the fibers against the resistance perpendicularto the direction (ρ-ratio). Their measurements for unidirectional composites show that the ratiocan be as high as 2000. According to a paper by Todoroki et al. [52] the intra-ply and inter-plyresistivity perpendicular to the fibers should differ because more resin is located between layersthan fibers in a ply. Their results confirm this. Their measurements showed considerable differ-ence in the intra-ply and inter-ply resistivity perpendicular to the fibers. However, the resultspresented by Louis et al. [53] for unidirectional composite showed no difference. The electricalcircuit of an unidirectional composite can be simplified by representing the individual fibers asresistors in a parallel network. However, because of the random contact points the circuit is muchmore complicated and this simple model will not work.

During the last years, a considerable amount of research has been conducted on the electricalproperties of composites. Abry et al. [55] showed that the electrical resistance, of an unidirectionalcomposite in the direction of the fibers, increases linearly with increasing length. But, the relationis nonlinear when measured perpendicular to the direction. Wang and Chung [56] investigatedthe idea of using carbon fiber composites for temperature and light sensing. In a paper by Luoand Chung [57] the idea of using carbon fiber composite as a capacitor was investigated. Theyvisualized the possibility of using the body of solar powered vehicles for storing energy, hencesaving space. In an article by Deborah D.L. Chung [58] it is mentioned that carbon fiber com-posites can be used as thermocouples and their sensitivity and linearity is very good. However,what we are interested in is the answer to the following question: Can the electrical propertiesof carbon fiber composites be used for estimating its condition? Few of the advantages of usingthe composite itself as a sensor are identified in a paper by Wang and Chung [59] such as: lowercost, whole structure sensing, and no reduction in mechanical properties, which can occur whensensors are embedded between laminates.

The determination of the electrical resistivity of a composite is often conducted by voltagemeasurements, i.e. a constant current is sent into the composite and the voltage drop is measured.The resistivity is then calculated by using Ohm’s law. Silver paint is often used to get goodconductivity between the composite and the probe. The resistivity of the composite changes iffibers break and the temperature changes. According to Abry et al. [60] different coefficients ofthermal expansion for the fibers and the matrix will cause the contact points of parallel fibers tochange with temperature. Angelidis et al. [61] explain that the contact points also change whenthe composite is subjected to strain and this change will locally change the path of the current inthe composite. Therefore, local change in voltage can sometimes be attributed to different currentpath. Matrix cracks and delamination can also change the resistance.

It is not enough to calculate the resistance alone. What is needed is a method that can be usedboth to monitor strain over time and also to estimate damage. In order to be able to monitorstrain over time, then the resistance must chance in accordance with the change in strain. But, inorder to be able to estimate damage, the resistance increase due to damage must be irreversible.The first condition is a requirement that the composite shows a piezoresistive behavior. Carbonfibers alone do not exhibit such behavior, but with certain production methods it can be attainedin composites [62]. As we saw in section 2.2 internal stresses, called residual stresses, are gener-ated during curing of the resin. By increasing the temperature, these stresses can be increased.Then, under tensile loading, the resistivity of the composite reduces because the residual stressesreduce. When the tensile load is removed, the residual stresses increase and also the resistivity

8

of the composite. A damage will increase the resistivity of the composite, it will still show piezo-resistive behavior but around higher mean value. This makes realtime monitoring possible.

The results from electrical measurements depend largely on the location of electrodes and postprocessing methods. If the electrodes are located at the ends of an unidirectional composite, thenonly average damage will be detected. Average damage will also be detected if the electrodes areall located at the same side of the composite. Because the current takes the easiest path available,which is most often in the top layers, this arrangement is not good for detecting internal damage.In order to be able to detect local damage, the approach and the location of the electrodes mustbe different.

Schueler et al [54] measured the conductivity of an composite and used it to build a finite ele-ment model conductivity map. They used Electrical Impedance Tomography (EIT) for measuringthe conductivity. They connected electrodes to the edges of the composite and sent a current intoit through two of them. The voltage was then measured between all adjacent electrodes. This wasrepeated for all electrodes and the information was used to map the conductivity. They arguethat for unidirectional composites, it is sufficient to place the electrodes on edges parallel to theorientation of the fibers. The assumption was made that the conductivity of the composite in thedirection of the fibers was infinite and damage could be represented by adding resistors perpen-dicular to the orientation of the fibers. This assumption enabled them to model the compositeas a simple resistor network. They state that this approach can locate and estimate the size of adamage in realtime, however, they point out that for composites with low ρ-ratio the electrodesmust be placed at all edges. Todoroki et al. [52] used a finite element model to investigate theeffect of different spacing between electrodes on the ability of locating and determining the sizeof a damage. Their results show that for composites with low volume fraction of fibers, largespacing has no effect on the ability of locating a delamination damage. But, with higher fractionthen the ability decreases with increasing spacing. Independent of volume fraction, the ability ofdetermining the delamination length decreases with increasing spacing between electrodes.

Abry et al. [60] and Kupke et al. [63] used both AC and DC measurements for the detec-tion of damage in carbon fiber composites. The research results of Abry and coworkers showthat DC measurements are more suitable for detecting fiber breakage and AC measurements aremore suitable for delamination and crack detection. The findings of Kupke and coworkers forDC measurements were similar, but they point out that if contact points are also considered thendelamination can be detected and even distinguished from fiber breakage. AC measurements arebased on measuring capacitance. According to Kupke et al. capacitance is the result of high resis-tivity of the matrix and decreases during loading because the matrix gets thinner. By introducingcapacitance, modeling of the composite electrical circuit gets more complicated. Since the contactpoints between fibers are random, then the location of capacitors is also random. The results pre-sented by Kupke and coworkers show that the capacitance of the composite decreases graduallywith each load cycle. They suggest that this could be used for monitoring damage accumulation.Ezquerra et al. [64] investigated the effect of frequencies on the electrical conductivity of carbonfiber composites. Their results show that the conductivity is almost independent of frequencywhich they attribute to high anisotropy.

Monitoring of carbon fiber composites using their electrical properties is very interesting, butfurther research must be conducted before the method can be considered reliable for conditionmonitoring.

3.3 Acoustic Emission

When microstructural changes occur in composites, energy is released and transient stress wavesare generated. These stress waves are called called Acoustic Emission (AE) [65] and approxi-mately 90% of their activity is located on the 10 to 550kHz frequency band [66]. The amplitudeincreases as the energy released increases [65]. The stress waves travel through the composite andwhen they reach the surface it will vibrate. The AE in composites is often audible [67]. They canbe heard if the energy that is released is high enough [65].

If a load is applied to a composite and damage occurs, then an AE is generated. The formationof a damage in a composite involves microstructural changes in the material, which will result inAE. If the load is removed and then reapplied, then no AE should be generated unless the load is

9

increased. This phenomenon is called Kaiser effect and is valid for most materials. But, anotherphenomenon also exists for composites; the Felicity effect. The Felicity effect is when an AE isgenerated at load level which is lower than the previous maximum. Many sources contributeto this effect; the opening and closing of cracks, rubbing of delamination surfaces and rubbingof parts. AE is also generated under loading because of the different material properties of thefibers and the matrix [67]. AE that comes from other sources, e.g. machinery and electricity, areconsidered to be disturbances or noise.

AE can be measured by measuring the small surface displacements generated by the vibrationof the composite. According to a paper by Duesling [68] piezoelectric sensors are most popular.They are either broadband or resonance. The paper also identifies other types of sensors; capac-itance, electromagnetic and optical. The last two are non-contact, but electromagnetic sensorsare considerably less sensitive than piezoelectric sensors. Optical sensor, e.g. laser, are free ofresonance and can be absolutely calibrated by measuring the correct amplitude of the AE [69].Measurements using piezoelectric sensors are sensitive to how the vibration is transferred to thesensors. The main factors that affect this are; the surface of the composite, sensor pressure againstthe composite and the coupling medium [70].

In order to conduct good measurements, it is important to understand how stress wavestravel through the composite. The different material properties of the fibers and matrix resultin anisotropic speed of propagation, i.e. the stress waves travel much faster in the direction ofthe fibers [68]. According to Prosser and coworkers [71, 72, 73] classical plate theory, higher orderReissner-Mindlin plate theory and laminated plate theory have been used together for investigat-ing the stress wave behavior in composites. The investigations, based on samples that possess theproperties of thin plates, have improved the understanding on how stress waves behave in com-posites. Prosser et al. explain that the waves propagate through the composite in two modes: aslowest order symmetrical and anti- symmetrical Lamb modes. Those two modes are also knownas extensional and flexural plate modes, respectively. They also name few of the mode properties;the largest component of the extensional mode is in the plane of the plate and is symmetric aboutthe midplane, but the largest component of the flexural mode is out of the plane and generallycontains lower frequencies and slower velocity. In a research conducted by Seale et al. [74] fatiguedamage was estimated by using Lamb wave velocity. The idea was to use Lamb wave velocity toestimate the modulus of the composite, hence allowing estimation of the fatigue damage.

The amplitude of the stress waves decreases as they propagate. The main reasons for the at-tenuation are; geometric spreading, dispersion, internal friction and scattering [75]. Theoretically,geometric dispersion in two dimensions will cause the amplitude to decrease proportionally withone over the square root of the distance. Viscoelastic properties of the composite will result indifferent velocities for different frequencies. Because of the velocity difference high and low fre-quency waves will separate as they propagate and the amplitude will decrease. This separationis called dispersion. Flexural mode components are very sensitive to dispersion but extensionalmode components are not [76, 73]. Internal friction of the material As mentioned above, the stresswaves vibrate the composite. However, not all of their energy can be used for this, part of theenergy will be used to overcome internal friction and also to generate sound. Scattering is whenthe waves change direction. This can be because of changes in material properties, damages,flaws, and structural elements. Composites are inhomogeneous medium which makes all wavebehavior predictions very difficult. The attenuation of stress waves has much influence on themeasurements. Geometric spreading alone can cause the amplitude to drop by 20 dB during thefirst few millimeters of propagation [67]. When the waves come across inhomogeneities or reachthe surface of the composite, they will reflect and mix with other waves and distort the measuredsignal. Reflection is a bigger problem for small composites than large ones. In large compositesattenuation will help reducing the distortion due to reflection.

If attenuation was nonexistant, then sensors could be placed anywhere on the composite andthe waves would propagate to them. The placement of sensors is however very important inorder to get good measurements. In order to determine the placement of sensors and calibratingthe measuring equipment, the method of breaking pencil leads, a Hsu-Nielsen source, on thesurface of the composite is often used. For this procedure a mechanical pencil and Pentel 2H 0.3mm lead is used. In a paper by Higo and Inaba [70] it is reported that, in the year 1975 when Dr.Hsu suggested this procedure, 0.5 mm lead was used. The reason for changing the thickness is

10

because the properties of the lead changed. Change in properties will result in different AE signalwhen they break. In order to produce consistent AE signal a teflon guide collar is sometimes puton the tip of the pencil. Hsu-Nielsen source has also been used for generating AE in research [75,77, 78, 79, 72]. According to Prosser and coworkers [72] pencil lead breaks can simulate impactdamage and delamination AE.

By measuring AE signals we can possibly detect and monitor changes in the composite. Be-cause micro damage generate AE, the method is promising for detecting damages and monitoringthem long before they get serious. Furthermore, this information could be used for predicting theservice life of the composite which could be used for maximizing the service life without riskingfailure. However, the data can be hard to interpret [79, 65]. Based on the idea that each AE sourceemits a characteristic AE, many research projects have been conducted, over the years, to extractuseful information from the signal. Four approaches have been predominant; activity-, feature-,frequency-, and modal-analysis.

Activity analysis uses changes in AE activities to detect damage. Load is often used for ref-erence. Two ratios are illustrative for this approach; Felicity and Shelby ratio. Felicity ratio is theratio of the lowest load that generates certain amount of AE activity, during loading, against thehighest load in last cycle [80]. The ratio was described by Dr. Timothy Fowler and has been usedas a indicator of damage with mixed results [80]. Instead of looking at the loading of an compositethe Shelby ratio looks at the unloading. Shelby ratio is therefore analogous to the Felicity ratio.The Shelby ratio is defined as the ratio of the lowest load that generates certain amount of AEactivity, during unloading, against the previous maximum load [67]. If the level of AE activityisn’t reached then the ratio is defined to be 1.0. The Shelby ratio, described by Dr. Marvin A.Hamstad, is useful for detecting damage that generate AE from friction. This ratio hasn’t beenused much in the literature, but an AE generated from friction can provide important informationabout the condition of an composite [81]. Some researchers have attempted to filter the AE signalgenerated by friction from the total signal in order to better detect AE from damage, but this canbe difficult [82]. The fact that a damage only emits AE once but friction many times suggests thatapproaches based on friction will be more robust.

Feature analysis works with features extracted from the AE signal. Conventional featuresare (see Figure 5); amplitude, event, duration, energy, number of peaks above certain thresholdand rise time, which is the time from the detection of an event until the highest amplitude is ob-tained. Many methods have been devised to extract information about the composite using these

Figure 5: Shows conventional AE features for one event.

features, e.g. using cumulative values, plotting features as function of each other, distributionanalysis and more. Different and in some cases contradictory results have been published usingthese features [77, 83, 84, 68]. In papers by Qi [66] and Prosser et al. [72] this was pointed out andProsser et al. also discovered that the same type of damage can emit AE signals with different

11

amplitudes. Godin et al. [85] claimed that conventional AE analysis cannot distinguish betweendifferent AE sources and suggested that more advanced methods, such as multivariate analysisand classifiers, should be used. Similar comments were made by Tsamtsakis et al. [82]. They sug-gested that new parameters, like force or displacement, should be added. The location of an AEsource by using conventional triangularization is accomplished by using arrival times. However,due to attenuation, anisotropic speed of the waves and dispersion this method is not robust.

Frequency analysis looks at the frequencies in the AE signal and associates frequency bands todifferent sources. Frequency analysis is most often conducted using FFT processing. The resultspresented by Bochse [86] determined the frequency bands for three types of damage. Frequenciesdue to matrix cracks were determined to reside in the 100 to 350 kHz band and fiber breakagewas given the interval between 350 to 700 kHz. The sources were classified by using the criterionthat 70% of the signal power had to lie within either frequency band. If not, then the source wasexpected to be debonding. Similar results were presented in a paper by de Groot et al. [23]. Thefrequency spectrum between 50 to 600kHz was analyzed. They determined that matrix cracksemit frequencies between 90 to 180 kHz, fiber pull-out release frequencies between 180 to 240 kHz,debonding produces frequencies between 240 to 310 kHz and fiber breakage gives AE signal withfrequencies above 300 kHz. In the recent years there has been increasing interest in using waveletbased methods for analyzing AE signals. They allow a convenient way to isolate the variousfrequency subbands. Since AE signals are very weak signals the sensors must be sensitive, whichmeans that they are also sensitive to noise. Noise from the environment is generally a strongersignal than AE, hence noise has the potential of being a huge problem. However, noise is generallylocated at frequencies that are much lower than the AE signals, and therefore noise is not as bigproblem as it could be [77].

Modal analysis involves looking at the waveforms and their modes of propagation. For thisplate wave theory is used. According to Surgeon and Wevers [78] plate wave theory makes it pos-sible to theoretically study the effects of attenuation and dispersion. They also report that severaladvantages of the method have been identified in the literature; better at detecting, distinguish-ing and locating AE sources in addition to be better at handling noise. Prosser [73] was able toeliminate noise by analyzing the waveform of different damages. He found out that cracks cre-ate mainly extensional mode waves and both delamination and impacts primarily create flexuralmode waves. The noise signal he was dealing with was due to grip damage, which was mainlymade of flexural mode waves. He was therefore able to distinguish between noise and matrixcracks. By using the Waveform can be used to locate AE source than is possible with conven-tional triangularization methods. The location was obtained by matching the similar phase pointof the AE waveforms [76, 73]. This resulted in extremely accurate source location. Prosser [87]was able to provide a direct measurement that the cracks initiated at the edge. The downsideof the modal analysis approach is that the modeling and signal processing methods can becomevery complex.

Acoustic Emission is a passive method since it only listens for a signal. An active variant canbe created by sending an ultrasound pulse into the composite, allowing it to travel through thematerial, and measuring it using AE techniques. This method is called Acousto-Ultrasonic andcombines AE and Ultrasound methods. The main advantage of this method is that instead ofwaiting for an AE to be generated, data can be generated manually. Changes in the signal arethen used to analyze the composite. According to a paper by Floyd and Phillips [88], the methodshould be able to detect changes in tensile strength and interlaminar shear strength together withdetecting voids, porosity and impact damage. Furthermore it should be able to locate where ten-sile failure will occur. However, the results presented were unfavorable to the method. They triedto explain the disappointing results by naming some possible reasons; the type of matrix (PEEK)could be incompatible with the method and the instrument settings may have been wrong.

4 Summary and Conclusion

A review of nondestructive testing methods has been presented and their potentiality for moni-toring composites, undergoing fatigue testing, in realtime discussed. Two approaches stand out:methods based on acoustic emission and the electrical properties of composites. The potential of

12

these two methods is discussed in greater detail.The main advantage of these methods is that they do not require the test to be temporarily

halted in order to take measurements. Research results show that both methods can be used forthe detection and discrimination of damage types such as; delamination, matrix cracks and bro-ken fibers. Furthermore, the results also show that damage can be monitored and the compositecondition estimated. Due to the many AE sources in composites the interpretation of the signalcan be difficult. For this reason, many methods have been devised, and used with varying results,for the analysis and interpretation of AE signals. Due to attenuation and the fact that the samesource can emit AE signal with different amplitude, methods based on thresholds are not reliable.Different and in some cases contradictory results show that further research is needed in orderto gain better understanding on both measurement techniques and on the behavior of the stresswaves in composites. Most research is based on simple composites. Because of this, and alsobecause of inconsistent findings, methods must be tested and their results compared before theyare applied to more complex shaped composites.

In the recent years, there has been increasing interest in smart materials. The electrical prop-erties of carbon fiber composites have been investigated with their potential use as a "smart ma-terial" in mind. Using electrical properties for estimating the condition of composites is a veryinteresting option. These properties haven’t been researched as extensively as the AE signal.Because carbon fiber composites with piezoelectric behavior can be manufactured, it would beinteresting to investigate whether their stiffness can be controlled with electricity. This could beexploited for many interesting uses, e.g. to reduce or stop vibration in a composite.

Fatigue testing of composites can take very long time, up to several weeks. It is thereforevaluable for an engineer, that wants to test a prototype, to be able to predict the remaining fatiguestrength of the composite. If the predicted number of cycles-to-failure is unacceptable, the test canbe stopped. Considerable time can be saved by this. By shortening the required time, productscan be; designed and developed in less time and at a lower cost than previously possible. In thiscontext there are several interesting topics to investigate:

• Can the AE signal generated during the first time loading of a composite, the CharacteristicDamage State [67], be used for predicting the fatigue strength?

• Are the first one hundred cycles in a fatigue test sufficient to predict fatigue strength [89]?

• Can AE be used for estimating damping characteristics, which then in turn can be used forpredicting fatigue strength ?

• Finally, can capacitance be used for predicting the fatigue strength [63]?

5 Acknowledgements

This work was supported by a research grant from RANNIS, the Icelandic Research Council 3.

References

[1] “Incandescent Lamp. Encyclopædia Britannica. retrieved august 23, 2004, from Encyclopæ-dia Britannica Online..” internet, 2004. http://search.eb.com/eb/article?eu=43191.

[2] “Materials Science. Encyclopædia Britannica. retrieved august 23, 2004, from EncyclopædiaBritannica Online..” internet, 2004. http://search.eb.com/eb/article?eu=118253.

[3] R. J. Diefendorf, “Carbon/graphite fibers,” Composites - Engineered Materials Handbook, vol. 1,pp. 49–53, 1987.

[4] S. Chand, “Review carbon fibers for composites,” Journal of Materials Science, vol. 35, no. 6,pp. 1303–1313, 2000.

[5] SACMA, “Releases industry statistics,” The AMPTIAC Newsletter, vol. 3, no. 3, p. 14, 1999.[6] TORAYCA, “Press release,” April 2004.

3RANNIS’s website: http://www.rannis.is

13

[7] N. W. Hansen, “Carbon fibers,” Composites - Engineered Materials Handbook, vol. 1, pp. 112–116, 1987.

[8] R. R. Hegde, A. Dahiya, and M. G. Kamath, “Carbon fibers.” internet, 2004.http://www.engr.utk.edu/mse/pages/Textiles/CARBON

[9] C. Landis, T. Cleveland, M. Cloninger, and D. Pollock, “Buckyballs, Diamond andGraphite. Department of Chemistry at the University of Wisconsin.” internet, 2004.www.chem.wisc.edu/ newtrad/ CurrRef/BDGTopic/BDGtext/BDGIntro.html.

[10] D. Cho., S. H. Yun, J. Kim, S. Lim, M. Park, G.-W. Lee, and S.-S. Lee, “Effects of fiber surface-treatment and sizing on the dynamic mechanical and interfacial properties of carbon/nylon6 composites,” Carbon Science, vol. 5, no. 1, pp. 1–5, 2004.

[11] W. D. Bascom, “Fiber sizing,” Composites - Engineered Materials Handbook, vol. 1, pp. 122–124,1987.

[12] S.-L. Gao, E. Mader, and S. Zhandarov, “Carbon fibers and composites with epoxy resins:Topography, fractography and interphases,” Carbon, vol. 42, no. 3, pp. 515–529, 2004.

[13] P. J. Walsh, “Carbon/graphite fibers,” Composites - Engineered Materials Handbook, vol. 21,pp. 35–40, 2001.

[14] J. Flitcroft and R. Adams, “A study of shear crack propagation in glass and carbon fibrereinforced plastics using acoustic emission monitoring,” Journal of Physics D (Applied Physics),vol. 15, no. 6, pp. 991–1005, 1982.

[15] YLA Inc., “Data sheet for rs-1 epoxy.” internet, 2004.http://www.ylainc.com/products/data_sheets/rs_1.htm.

[16] “Composite basics.” internet, 2004. http://www.compositesone.com/basics.html.[17] Z. Zhang and G. Hartwig, “Relation of damping and fatigue damage of unidirectional fibre

composites,” International Journal of Fatigue, vol. 24, no. 7, pp. 713–718, 2002.[18] M. M. Schwartz, Composite Materials - Properties, Nondestructive Testing, and Repair, vol. 1.

Prentice-Hall, 1997.[19] R. Adams and P. Cawley, “Review of defect types and nondestructive testing techniques for

composites and bonded joints,” NDT International, vol. 21, no. 4, pp. 208–222, 1988.[20] R. Olsson, J. Iwarsson, L. Melin, A. Sjogren, and J. Solti, “Experiments and analysis of lam-

inates with artificial damage,” Composites Science and Technology, vol. 63, no. 2, pp. 199–209,2003.

[21] W.-C. Wang, C.-H. Day, C.-H. Hwang, and T.-B. Chiou, “Nondestructive evaluation of com-posite materials by espi,” Research in Nondestructive Evaluation, vol. 10, no. 1, pp. 1–15, 1998.

[22] G. Tracy, P. Feraboli, and K. Kedward, “A new mixed mode test for carbon/epoxy compositesystems,” Composites Part A: Applied Science and Manufacturing (Incorporating Composites andComposites Manufacturing), vol. 34, no. 11, pp. 1125–1131, 2003.

[23] P. De Groot, P. Wijnen, and R. Janssen, “Real-time frequency determination of acoustic emis-sion for different fracture mechanisms in carbon/epoxy composites,” Composites Science andTechnology, vol. 55, no. 4, pp. 405–412, 1995.

[24] M. W. Gaylord, Reinforced Plastics - Theory and practice 2nd ed. Cahners books, 1974.[25] N. Hancox, “Thermal effects on polymer matrix composites: Part 1. thermal cycling,” Mate-

rials and Design, vol. 19, no. 3, pp. 85–91, 1998.[26] S. Ravishankar and C. Murthy, “Characteristics of ae signals obtained during drilling com-

posite laminates,” NDT and E International, vol. 33, no. 5, pp. 341–348, 2000.[27] V. V. Bolotin, “Mechanics of delaminations in laminate composite structures,” Mechanics of

Composite Materials, vol. 37, no. 6, pp. 367–380, 2001.[28] J. Andersons, M. Hojo, and S. Ochiai, “Empirical model for stress ratio effect on fatigue

delamination growth rate in composite laminates,” International Journal of Fatigue, vol. 26,no. 6, pp. 597–604, 2004.

[29] A. Sjogren and L. E. Asp, “Effects of temperature on delamination growth in a carbon/epoxycomposite under fatigue loading,” International Journal of Fatigue, vol. 24, no. 2-4, pp. 179–184,2002.

[30] M. V. Hosur, C. R. L. Murthy, T. S. Ramamurthy, and A. Shet, “Estimation of impact-induceddamage in cfrp laminates through ultrasonic imaging,” NDT and E International, vol. 31, no. 5,pp. 359–374, 1998.

[31] J. P. Komorowski, R. W. Gould, and D. L. Simpson, “Synergy between advanced composites

14

and new ndi methods,” Advanced Performance Materials, vol. 5, no. 2, pp. 137–151, 1998.[32] C. Baley, Y. Grohens, F. Busnel, and P. Davies, “Application of interlaminar tests to marine

composites. relation between glass fibre/polymer interfaces and interlaminar properties ofmarine composites,” Applied Composite Materials, vol. 11, no. 2, pp. 77–98, 2004.

[33] N. Avdelidis, A. Moropoulou, and Z. Marioli Riga, “The technology of composite patchesand their structural reliability inspection using infrared imaging,” Progress in Aerospace Sci-ences, vol. 39, no. 4, pp. 317–328, 2003.

[34] S. S. Kessler, S. M. Spearing, M. J. Atalla, C. E. Cesnik, and C. Soutis, “Damage detectionin composite materials using frequency response methods,” Composites Part B: Engineering,vol. 33, no. 1, pp. 87–95, 2002.

[35] J. Hou and G. Jeronimidis, “Vibration of delaminated thin composite plates,” Composites PartA: Applied Science and Manufacturing (Incorporating Composites and Composites Manufacturing),vol. 30, no. 8, pp. 989–995, 1999.

[36] Y. ZOU, L. TONG, and G. P. STEVEN, “Vibration-based model-dependent damage (delam-ination) identification and health monitoring for composite structures - a review,” Journal ofSound and Vibration, vol. 230, no. 2, pp. 357–378, 2000.

[37] H. Y. KIM, “Vibration-based damage identification using reconstructed frfs in compositestructures,” Journal of Sound and Vibration, vol. 259, no. 5, pp. 1131–1146, 2003.

[38] L. Melin, J. Schon, and T. Nyman, “Fatigue testing and buckling characteristics of impactedcomposite specimens,” International Journal of Fatigue, vol. 24, no. 2-4, pp. 263–272, 2002.

[39] U. Icardi, “Through-the-thickness displacements measurement in laminated composites us-ing electronic speckle photography,” Mechanics of Materials, vol. 35, no. 1-2, pp. 35–51, 2003.

[40] Y. A. Plotnikov and W. P. Winfree, “Visualization of subsurface defects in composites usinga focal plane array infrared camera,” Proceedings of SPIE - The International Society for OpticalEngineering, vol. 3700, pp. 26–31, 1999.

[41] R. J. Ball and D. P. Almond, “The detection and measurement of impact damage in thick car-bon fibre reinforced laminates by transient thermography,” NDT and E International, vol. 31,no. 3, pp. 165–173, 1998.

[42] M. Castaings and B. Hosten, “Lamb and sh waves generated and detected by air-coupledultrasonic transducers in composite material plates,” NDT and E International, vol. 34, no. 4,pp. 249–258, 2001.

[43] Y. M. Tarnopol’skii and V. L. Kulakov, “Tests methods for composites. survey of investi-gations carried out in the pmi of latvian academy of sciences in 1964-2000,” Mechanics ofComposite Materials, vol. 37, no. 6, pp. 431–448, 2001.

[44] R. Meske and E. Schnack, “Particular adaptation of x-ray diffraction to fiber reinforced com-posites,” Mechanics of Materials, vol. 35, no. 1-2, pp. 19–34, 2003.

[45] P. Munshi and S. P. Singh, “Hamming signature of three composite specimens,” Research inNondestructive Evaluation, vol. 10, no. 2, pp. 109–122, 1998.

[46] J. P. Dunkers, F. R. Phelan, D. P. Sanders, M. J. Everett, W. H. Green, D. L. Hunston, and R. S.Parnas, “The application of optical coherence tomography to problems in polymer matrixcomposites,” Optics and Lasers in Engineering, vol. 35, no. 3, pp. 135–147, 2001.

[47] K. Maslov, R. Y. Kim, V. Kinra, and N. J. Pagano, “A new technique for the ultrasonic detec-tion of internal transverse cracks in carbon-fibre/bismaleimide composite laminates,” Com-posites Science and Technology, vol. 60, no. 12-13, pp. 2185–2190, 2000.

[48] G. D. D. Graham, J.C. Macfarlane and C. Carr, “Nde of damaged carbon fibre compositesusing hts squids: from impact detection to neural networks,” 6th European Conference onApplied Superconductivity, 2003.

[49] Collaboration for NDT Education, “Basic principles of eddy current inspection.” inter-net, 2004. http://www.ndt-ed.org/EducationResources/CommunityCollege/ EddyCur-rents/Introduction/IntroductiontoET.htm.

[50] P. A. Zinov’ev, A. A. Smerdov, and G. G. Kulish, “Experimental investigation of elastodis-sipative characteristics of carbon-fiber-reinforced plastics,” Mechanics of Composite Materials,vol. 39, no. 5, pp. 393–398, 2003.

[51] S. Wang and D. Chung, “Apparent negative electrical resistance in carbon fiber composites,”Composites Part B: Engineering, vol. 30, no. 6, pp. 579–590, 1999.

[52] A. Todoroki, M. Tanaka, and Y. Shimamura, “Electrical resistance change method for mon-

15

itoring delaminations of cfrp laminates: effect of spacing between electrodes,” CompositesScience and Technology, vol. – In press –, no. x, pp. xx–xx, 2004.

[53] M. Louis, S. Joshi, and W. Brockmann, “An experimental investigation of through-thicknesselectrical resistivity of cfrp laminates,” Composites Science and Technology, vol. 61, no. 6,pp. 911–919, 2001.

[54] R. Schueler, S. P. Joshi, and K. Schulte, “Damage detection in cfrp by electrical conductivitymapping,” Composites Science and Technology, vol. 61, no. 6, pp. 921–930, 2001.

[55] J. Abry, S. Bochard, A. Chateauminois, M. Salvia, and G. Giraud, “In situ detection of damagein cfrp laminates by electrical resistance measurements,” Composites Science and Technology,vol. 59, no. 6, pp. 925–935, 1999.

[56] S. Wang and D. Chung, “Temperature/light sensing using carbon fiber polymer-matrix com-posite,” Composites Part B: Engineering, vol. 30, no. 6, pp. 591–601, 1999.

[57] X. Luo and D. Chung, “Carbon-fiber/polymer-matrix composites as capacitors,” CompositesScience and Technology, vol. 61, no. 6, pp. 885–888, 2001.

[58] D. D. Chung, “Composites get smart,” Materials Today, vol. 5, no. 1, pp. 30–35, 2002.[59] W. Xiaojun and D. Chung, “Self-monitoring of fatigue damage and dynamic strain in carbon

fiber polymer-matrix composite,” Composites Part B: Engineering, vol. 29, no. 1, pp. 63–73,1998.

[60] J. Abry, Y. Choi, A. Chateauminois, B. Dalloz, G. Giraud, and M. Salvia, “In-situ monitoringof damage in cfrp laminates by means of ac and dc measurements,” Composites Science andTechnology, vol. 61, no. 6, pp. 855–864, 2001.

[61] N. Angelidis, C. Wei, and P. Irving, “The electrical resistance response of continuous carbonfibre composite laminates to mechanical strain,” Composites Part A: Applied Science and Man-ufacturing (Incorporating Composites and Composites Manufacturing), vol. 35, no. 10, pp. 1135–1147, 2004.

[62] X. Wang and D. D. L. Chung, “Piezoresistive behavior of carbon fiber in epoxy,” Carbon,vol. 35, no. 10-11, pp. 1649–1651, 1997.

[63] M. Kupke, K. Schulte, and R. Schuler, “Non-destructive testing of frp by d.c. and a.c. electri-cal methods,” Composites Science and Technology, vol. 61, no. 6, pp. 837–847, 2001.

[64] T. Ezquerra, M. Connor, S. Roy, M. Kulescza, J. Fernandes-Nascimento, and F. Balta-Calleja,“Alternating-current electrical properties of graphite, carbon-black and carbon-fiber poly-meric composites,” Composites Science and Technology, vol. 61, no. 6, pp. 903–909, 2001.

[65] M. Wevers, “Listening to the sound of materials: acoustic emission for the analysis of mate-rial behaviour,” NDT and E International, vol. 30, no. 2, pp. 99–106, 1997.

[66] G. Qi, “Wavelet-based ae characterization of composite materials,” NDT and E International,vol. 33, no. 3, pp. 133–144, 2000.

[67] M. A. Hamstad, “Acoustic Emission Primer, Lecture delivered by Dr. Hamstad at theacoustic emission working group meeting on august 4, 2003 (aewg-46).” internet, 2004.http://www.engr.du.edu/profile/frame_bio.html.

[68] L. Duesing, “Acoustic emission testing of composite materials,” Reliability and MaintainabilitySymposium, 1989. Proceedings., Annual, pp. 128–134, 1989.

[69] H. Kim and H. Park, “Laser interferometry system for measuring displacement amplitudeof acoustic emission signals,” Journal of Physics D (Applied Physics), vol. 17, no. 4, pp. 673–5,1984.

[70] Y. Higo and H. Inaba, “General problems of ae sensors,” ASTM Special Technical Publication,no. 1077, pp. 7–24, 1991.

[71] W. H. Prosser, The Propagation Characteristics of the Plate Modes of Acoustic Emission Wavesin Thin Aluminum Plates and Thin Graphite/Epoxy Composite Plates and Tubes. PhD thesis,Johns Hopkins University, november 1991. techreports.larc.nasa.gov/ltrs/PDF/phd-91-prosser.pdf.

[72] W. Prosser, K. Jackson, S. Kellas, B. Smith, J. McKeon, and A. Friedman, “Advancedwaveform-based acoustic emission detection of matrix cracking in composites,” NDT andE International, vol. 30, no. 2, p. 108, 1997.

[73] W. H. Prosser, “Waveform analysis of ae in composites,” Proceedings of the Sixth InternationalSymposium on Acoutic Emission From Composite Materials, vol. 3700, pp. 61–70, 1998.

[74] M. D. Seale, B. T. Smith, W. H. Prosser, and J. E. Masters, “Lamb wave response of fatigued

16

composite samples,” Review of Progress in Quantative Nondestructive Evaluation, vol. 13B,pp. 1261–1266, 1994.

[75] W. H. Prosser, “Applications of advanced waveform-based ae techniques for testing com-posite materials,” Proceedings of the SPIE conference on Nondestructive Evaluation Techniques forAging Infrastructure and Manufacturing: Materials and Composites, pp. 146–153, 1996.

[76] W. Prosser, “Advanced ae techniques in composite materials research,” Journal of AcousticEmission, vol. 14, no. 3-4, pp. S1–11, 1996.

[77] M. Giordano, A. Calabro, C. Esposito, A. D’Amore, and L. Nicolais, “An acoustic-emissioncharacterization of the failure modes in polymer-composite materials,” Composites Scienceand Technology, vol. 58, no. 12, pp. 1923–1928, 1998.

[78] M. Surgeon and M. Wevers, “Modal analysis of acoustic emission signals from cfrp lami-nates,” NDT and E International, vol. 32, no. 6, pp. 311–322, 1999.

[79] G. Farrow and M. Darby, “Study of simulated acoustic emission in unidirectional carbonfibre reinforced plastic,” Journal of Physics D: Applied Physics, vol. 25, no. 2, pp. 234–241, 1992.

[80] G. Ramirez, P. Ziehl, and T. Fowler, “Nondestructive evaluation of frp design criteria withprimary consideration to fatigue loading,” Transactions of the ASME. Journal of Pressure VesselTechnology, vol. 126, no. 2, pp. 216–28, 2004.

[81] J. Awerbuch and S. Ghaffari, “Monitoring progression of matrix splitting during fatigueloading through acoustic emission in notched unidirectional graphite/epoxy composite,”Journal of Reinforced Plastics and Composites, vol. 7, no. 3, pp. 245–64, 1988.

[82] D. Tsamtsakis, M. Wevers, and P. De Meester, “Acoustic emission from cfrp laminates duringfatigue loading,” Journal of Reinforced Plastics and Composites, vol. 17, no. 13, pp. 1185–1201,1998.

[83] S. Barre and M. Benzeggagh, “On the use of acoustic emission to investigate damage mech-anism in glass-fibre-reinforced polypropylene,” Composites Science and Technology, vol. 52,no. 3, pp. 369–376, 1994.

[84] K. Ono, “Acoustic emission behavior of flawed unidirectional carbon fiber-epoxy compos-ites,” Journal of Reinforced Plastics and Composites, vol. 7, no. 1, pp. 90–105, 1988.

[85] N. Godin, S. Huguet, R. Gaertner, and L. Salmon, “Clustering of acoustic emission signalscollected during tensile tests on unidirectional glass/polyester composite using supervisedand unsupervised classifiers,” NDT and E International, vol. 37, no. 4, pp. 253–264, 2004.

[86] J. Bohse, “Acoustic emission characteristics of micro-failure processes in polymer blends andcomposites,” Composites Science and Technology, vol. 60, no. 8, pp. 1213–1226, 2000.

[87] W. Prosser, K. Jackson, S. Kellas, B. Smith, J. McKeon, and A. Friedman, “Advancedwaveform-based acoustic emission detection of matrix cracking in composites,” NDT andE International, vol. 30, no. 2, p. 108, 1997.

[88] R. Russell-Floyd and M. Phillips, “A critical assessment of acousto-ultrasonics as a methodof nondestructive examination for carbon-fibre-reinforced thermoplastic laminates,” NDTInternational, vol. 21, no. 4, pp. 247–57, 1988.

[89] M. Fuwa, B. Harris, and A. Bunsell, “Acoustic emission during cyclic loading of carbon-fibre-reinforced plastics,” Journal of Physics D (Applied Physics), vol. 8, no. 13, pp. 1460–71,1975.

17