1-s2.0-s1359835x05003957-main

Upload: ze-mari

Post on 08-Mar-2016

13 views

Category:

Documents


0 download

DESCRIPTION

artigo

TRANSCRIPT

  • la

    , H

    , M

    rosp

    Th

    m 2

    indpr

    by a large group of researchers. The main focus is put on the types of heat generation mechanisms during the induction heating processand the parameters that govern the welding process (frequency, power, pressure, residence time), as well as on the secondary phenomena

    interest for the industry. Recently developed matrix mate-

    the limited deformation allowed for the reinforcing bres,currently produced thermoplastic components have rather

    structure that can result in weakening of the properties.

    generally dicult to control in industrial environment, andadhesives used (usually epoxies) have long curing cycles. Itcan also be dicult for the chemically inert thermoplasticmatrix to bond [17].

    Fusion bonding is a joining method that uses the prop-erty of thermoplastic matrices to ow when heated above

    * Corresponding author. Address: Netherlands Institute for MetalsResearch, Mekelweg 2, 2628CD Delft, The Netherlands.

    E-mail address: [email protected] (T.J. Ahmed).

    Composites: Part A 37 (20rials used for manufacturing thermoplastic composites(TPCs) yield materials with basic mechanical properties(strength, stiness) much the same, if not better than thethermosets (TS) [1]. Additionally, TPCs also show a num-ber of advantages when compared to the TS, among whichimproved toughness, better environmental resistance (hightemperature, moisture, aggressive uids), shorter process-ing times, non-ammability and innite shelf life [2,3].One of their most important advantage lies in the possibil-ity for a low-cost, rapid production [4]. However, due to

    Traditional joining methods for metals and thermosets(mechanical fastening and adhesive bonding) are feasible,but not ideal for TPCs. Mechanical fastening has a numberof disadvantages: introducing stress concentrations in thematerial, delamination during drilling, dierent thermalexpansion of the fasteners relative to the composite, waterintrusion into the joint, possible galvanic corrosion, weightincrease and extensive labour and time requirements.Adhesive bonding also presents some diculties whenapplied on TPCs. It requires extensive surface preparation,that can inuence the quality of the weld. An overview of the experimental procedure is also presented, with an emphasis on the exper-imental set-up. Finally, a brief overview of the modelling of the heat generation mechanisms and the induction welding process ispresented. 2005 Elsevier Ltd. All rights reserved.

    Keywords: A. Thermoplastic resin; B. Mechanical properties; E. Joining

    1. Introduction

    As a result of their growing potential for high perfor-mance applications, continuous bre-reinforced thermo-plastic composites (CFRTPCs) are becoming of greater

    simple geometry, which makes joining an indispensablestep in the manufacturing process of TPCs.

    Joining has proved to be a critical step in the process ofmanufacturing thermoplastic composite (TPC) products,because it can initiate a number of irregularities in theInduction welding of thermop

    T.J. Ahmed a,b,*, D. Stavrov b

    a Netherlands Institute for Metals Researchb Design and Production of Composite Structures, Faculty of Ae

    2629HS Delft,

    Received 6 July 2005; received in revised for

    Abstract

    This paper presents a comprehensive overview of the process ofto provide a deeper insight into the nature of the induction welding1359-835X/$ - see front matter 2005 Elsevier Ltd. All rights reserved.doi:10.1016/j.compositesa.2005.10.009stic compositesan overview

    .E.N. Bersee b, A. Beukers b

    ekelweg 2, 2628CD Delft, The Netherlands

    ace Engineering, Delft University of Technology, Kluyverweg 3,

    e Netherlands

    0 October 2005; accepted 20 October 2005

    uction welding of thermoplastic composites. The main objective isocess and to summarise the investigative eort that was put into it

    www.elsevier.com/locate/compositesa

    06) 16381651

  • their glass transition temperature Tg (for amorphous poly-mers) or the crystalline melting point Tm (for semi-crystal-line polymers) and regain their mechanical properties aftercooling down. Known also as welding, it can be generallydescribed as joining of two parts by fusing their contactinterfaces, followed by cooling (consolidating) under pres-sure that enables the bond to be made. It overcomes allproblems connected to the traditional techniques men-tioned above. Fusion bonding is widely considered to bethe ideal bonding technique for TPCs.

    The heat needed for melting the joint interface can beapplied by various means, e.g. hot plates, hot gas, friction,ultrasonic and radio signal, microwaves, Joule eect in aresistor, laser and induction, to mention some of them.From this variety of means, three are considered to havegreatest potential for future development: ultrasonic, resis-tance and induction welding. A large number of researchstudies were performed on these three techniques. Thatresulted in the publication of several mainly collective over-

    2. Induction welding

    Induction welding is a unique process in that it requiresno contact between the induction coil or the heat suscep-tor and can be designed such that no heat is producedoutside of the desired weld area. The process of heatingby induction is not a new technology and since 1916has most frequently been used for heating metals [11].Only within the last two decades has this type of heatingcome into the scope for heating composites and is provingitself to be a very eective method for the high-speed pro-cessing of welding bre-reinforced thermoplastic compos-ites [12]. In addition, the process is extremely versatilewith similar and dissimilar thermoplastics that can bewelded, as well as thermoplastic to non-thermoplasticmaterials [13,14].

    The principle behind the process itself is simple. Whenan alternating voltage is placed across a conductive coil,an alternating current is produced. Subsequently this alter-

    T.J. Ahmed et al. / Composites: Part A 37 (2006) 16381651 1639views of the three welding techniques [510] that oeredfull descriptions of the processes and their advantages.

    This paper presents a comprehensive overview of theprocess of induction welding of TPCs. The main objectiveis to provide a deeper insight into the nature of the induc-tion welding process and the investigative eort that wasput into it by a large group of researchers. After a generaldescription of the induction welding process, an overviewof the experimental procedure is presented, with an empha-sis on the experimental set-up. The main focus is set ontypes of heating that occur during the induction heatingprocess and the parameters that govern the welding process(frequency, power, pressure, residence time), as well as onthe secondary phenomena that can inuence the qualityof the weld. Finally, the modelling of the heat generationmechanisms and the induction welding process is brieydiscussed.Fig. 1. Induction welding process; susnating current induces a time variable magnetic eld whichhas the same frequency as the alternating current causingit. When a magnetically susceptible and electrically conduc-tive material is placed in the vicinity of the coil and its alter-nating magnetic eld, eddy currents are induced, with afrequency matched to that of the magnetic eld. A condi-tion imposed on the material is that closed-loop circuitsmust be present for eddy currents to be induced. In the caseof bre-reinforced thermoplastics, closed-loop circuits inthe form of a conductive network is produced throughweaves or cross plies, for example. The eddy currents aremet with the resistance of the material and energy is lostin the form of heat. There are four mechanisms that resultin heat production and will be further discussed in Section3. Pressure can then be applied during or after heating tocomplete the welding process. A schematic diagram isshown in Fig. 1.ceptor and susceptorless heating.

  • s: P2.1. Welding set-up

    The type of apparatus required to produce a fully work-ing set-up varies for dierent applications, but the equip-ment can easily be divided into four parts [13,15]. Therst is the radio frequency (rf) power generator, which sup-plies the necessary current and voltage to the inductioncoil. The second is the heat station which includes theinduction coil and produces the magnetic eld needed toheat the material. The third constituent is the compositeworkpiece material itself and nally, the fourth involvesthe secondary equipment such as the water cooling system,which is discussed in detail in [11,16], and xtures.

    2.1.1. Power source

    The power supply is extremely important and aects thereliability, maintainability, compactness, energy eciencyand cost of the overall system [17]. Each of the systems usesan input power of 230 V or 340 V alternating current (ac)with a frequency of 5060 Hz. The alternating input cur-rent is changed into a direct current (dc) to create a morecontrollable input, which is then changed back into therequired output ac. The frequency, voltage and power ofthe output ac is dened by the induction coil used in inter-action with the workpiece [11,18]. Finally, in order to oper-ate at the coils highest eciency, the current is passedthrough a load matching station.

    Induction power sources come in two forms; solid stateor vacuum tube, and dier in the frequency range that canbe produced. The older vacuum tube power source usesvacuum oscillator tubes for changing the dc to ac. Thesetypes of oscillators are normally used for frequency rangesbetween 200 kHz and 2 MHz and higher power rangesabove 10 kW but because of lower eciency factors,solid-state power supplies are preferred if available [11].The solid-state power source is capable of a frequency upto 1 MHz, are smaller than the vacuum tube power sourcesand have a higher eciency. Conversion eciency factorsof a solid-state power source are greater compared to thatfor vacuum tube power sources [17], with 5560% for avacuum tube compared to 8595% for a solid state [19].

    2.1.2. Heat station

    The heat station is the second segment of the inductionheating apparatus and uses a capacitor and a coil to heatthe workpiece. The design of the capacitor to match boththe power output and the induction coil power require-ments is described in detail elsewhere in [17,16].

    Energy is transferred to the workpiece through theinduction coil. The energy transfer mechanism can be mod-elled using the transformer principle [20] where the induc-tion coil can be seen as the transformers primary coiland the workpiece the transformers secondary coil. In anideal situation where there is a 100% coupling eciency,100% of the power provided by the coil is transferred

    1640 T.J. Ahmed et al. / Compositethrough air to the workpiece. In this case, the relationbetween the induction coil and the workpiece can bederived according to the transformer law Iw = NcIc, wheresubscripts w and c indicate the workpiece and coil respec-tively and Nc is the number of coil windings. Followingthis, the generated energy in the workpiece, Ew is thengiven by Eq. 1 which relates the eects of the coil on theworkpiece.

    Ew Pwt I2wRwt N 2cI2cRwt 1In reality, there is not a 100% coupling eciency betweenthe coil and workpiece since the coupling is through air,and ideally a reduction factor should be introduced intoEq. 1 to reect this. However, the coupling eciency is acomplex factor that depends on many parameters such asthe coilworkpiece distance, the workpiece itself and thecoil geometry. It is unknown how such factors may inu-ence heating eciency quantitatively and this point hasgenerally been avoided in previous studies. Instead, it isaccepted that the coil should be as close as possible tothe workpiece for maximum coupling eciency [20].

    The coil geometry also has a great inuence on the heatgenerated within the workpiece and the importance of thedesign has been discussed in [21,22]. It is possible to designthe induction coil such that the associated magnetic eld isfocused onto the specic weld zone that needs to be heated.When designing the coil geometry several design consider-ations need to be taken into account in order to producethe most ecient and uniform heating eect:

    Due to the higher magnetic ux density near the coil, thecoil should be as close to the workpiece, and as fullyover the weld area as possible to assure maximumenergy transfer [11,23]. The time for welding is alsoaected by the distance and doubling the workpiececoilseparation has been found to increase welding time by300400% [24].

    The shape of the magnetic eld is asymmetric due to thecoil connections to the heat station xtures. Hence theresulting heating pattern of a symmetrical coil is dis-torted [20]. In addition, at the point of the coil connec-tion the magnetic eld is weaker, which is caused bymagnetic eld cancellation of the two parallel connec-tions [20].

    The rest of the coil needs to be designed to prevent mag-netic eld cancellation such that if at any point twolengths of the coil are running parallel to each other,the distance between them should be altered to preventeld cancellations [16,20].

    When working with composites the frequencies used arehigher than the frequencies used for heating metals.Therefore the system has tendencies for overloadingand arcing between coil passages [15].

    For heating application on composites three dierent coiltypes can be considered as shown in Fig. 2 [13]: a singleturn coil, a solenoid coil and a pancake coil, and numerous

    art A 37 (2006) 16381651variations on these basic types. The single turn coil has amagnetic eld that is concentrated around its diameter

  • and is therefore used in applications where circular areas

    The load power dissipated is given by Joules law inwhich the resistance is the sum of Rp and Rs. Secondlythe output current is the output voltage of the converterdivided by the circuit impedance, Z:

    Iout VZ V

    Rp Rs jX Ip X Is X Ig 2

    The formula shows that the output current, and conse-quently power, depends on several factors, such as coilworkpiece geometry, material properties and frequency.To be more specic, the region of optimal eciency isnot only dened by the power source, but also by the work-piece. Therefore for each dierent application dierentmachine settings are used. Each of these parameters is anon-linear function and in turn depend on other factors.

    Fig. 2. Basic coil shapes [10]. (a) Single turn, (b) solenoid and (c) pancake.

    T.J. Ahmed et al. / Composites: Part A 37 (2006) 16381651 1641are to be heated. The solenoid is eectively an enlarged sin-gle turn coil and is able to heat larger cylindrical areas thatare passed through its centre. Finally, the pancake coil isable to heat large at areas and can be used on either sideof a weld zone to give good localised heating. This was con-rmed by Miller et al. [25,23] and Rudolf et al. [24] and wasconcluded that heating the weld from both sides is mosteective for creating a uniform through-the-thickness heat-ing zone. A uniform area of heat can be created with littletemperature variation in the width of the heated zone.

    2.1.3. Workpiece and load matching

    The highest energy transfer between the workpiece andthe coil is achievable through the process of adjusting thethree induction parameters, voltage, power and frequencyin such a way that the induction coil will operate at its max-imum eciency. This is the process of load matching. Themaximum eciency of an induction coil lies at its resonantfrequency. In a normal induction process the induction coiland therefore its input characteristics are designed accord-ing to the desired input of the workpiece. Consequently theoutput characteristics of the power source need to bematched to the induction coil. The circuit representationof the power source, heat station and workpiece is shownin Fig. 3. In this gure Rp is the resistance of the coil, Rsis the reective resistance of the secondary eddy currentpath in the workpiece to the primary circuit, XLp is the pri-mary reactance of the work coil, XLs is the reactance of thesecondary eddy current path reected to the primary circuitand XLg is the reective reactance of the secondary air gapbetween the coil and the workpiece. This indicates that thecurrent in the power source is inuenced by the workpieceitself and the inductor.Fig. 3. Equivalent induction heating system circuit [17].2.1.4. Welding xture

    The nal part of the induction welding apparatus is thetest environment itself and more specically, the apparatuswhere the workpiece is held. To avoid heating of the testequipment, it is important to avoid the use of magneticallysusceptible materials within the vicinity of the magneticeld [15]. In cases where this is not possible, sucient cool-ing is necessary. It is also possible to create distancebetween the xtures and the coil such that the magneticeld has little to no inuence. For example, a concrete slabhas been used to provide this distance between the surfaceon which the workpiece is placed and the coil [26]. In thesecases where the use of metal xtures are unavoidable, theconcrete is also a sucient an insulator that acts to insulatebetween the workpiece and metallic xtures underneath toavoid the metal heating up the workpiece. In this way,uneven heating through the thickness is avoided. However,the surrounding area inuenced by the magnetic coil is rel-atively small and can be localised as will be described inSection 3.

    In order to aid in the consolidation process, pressureneeds to be applied. A number of methods have beenestablished to provide continuous and discontinuous pres-sure and two examples are shown in Fig. 4. In order forFig. 4. Methods of consolidation. Continuous: (i) moving platform withpressure rollers [27,29] and discontinuous: (ii) vacuum bagging [28].

  • continuous welding to take place, the workpieces must beheated by induction rst, after which, delamination occurs.In order to reconsolidate and remove voids, the weldregion is passed through pressure rollers. The consolidatedstructure is then allowed to cool under controlled condi-tions [27]. There are a number of variations for continuousinduction welding but are all based on coil/pressure rollercombinations.

    rent produces its own magnetic eld which is able to cancelthe magnetic eld in the deeper regions of the workpiece[23]. This extent of cancellation depends on the size ofthe induced current at the surface nearest to the coil. Thecurrent can only move along electrically conductive paths,or the conductive bres. In prepreg stacks lying perpendic-ular to each other, this means that the mirror image is morerectangular in shape. Woven plies however, have been

    1642 T.J. Ahmed et al. / Composites: Part A 37 (2006) 16381651Vacuum bagging moulding has been used as a discontin-uous method of applying pressure [28], which allows for auniform distribution of pressure to be applied over theentire surface of the weld zone. However, this method islimited to thermoplastics with extremely low viscosity atelevated temperatures.

    3. Heating process

    Heat is produced within the composite workpiece as aresponse to the magnetic eld, but where and how the heatis produced depends on the material properties. The mate-rial itself can act as the susceptor [12,23,25,27,3032] and isdesirable because of the absence of a contaminating insertfrom the bond line which can weaken the mechanicalstrength of the bond [12]. Alternatively an insert can beincluded at the weld [3237] as heat can be generated andconcentrated within the weld zone, even when the adher-ends are also magnetically susceptible. It is unclear in sus-ceptor versus susceptorless welds, as to which produces thebetter quality weld and consequently much research eorthas been spent in this area. Furthermore, how the work-piece generates heat has been the source of much debateover the past decade, and the arguments put forward foreach theory is described in this section.

    3.1. Heating of bre-reinforced thermoplastic composites

    As previously mentioned in Section 2, a conductive loopneeds to be present in order for eddy currents to be inducedin the workpiece. Heat energy, E, is produced according toJoules law, E = I2Rt where I is the current, R is the resis-tance and t is the time of exposure to the magnetic eld.The rate of heating is dependent on the frequency andintensity of the eddy current and the electrical resistance,specic heat and magnetic permeability of the material[11]. Miller et al. [23] have shown that the eddy currentsinduced in the workpiece form a global loop that is the mir-ror image of the coil. A consequence of this is that the cur-Fig. 5. Induction heating mechanisms [45]. (a) Fibre heatfound to produce a more similar image to the coil and thishas been attributed to the high incidence of electricalcontact within the weave [23].

    Three categories of heating mechanisms have been iden-tied, namely Joule loss, junction heating and hysteresisloss. These mechanisms dier in where exactly the heatingtakes place within the workpiece, and are summarised dia-grammatically in Fig. 5. There have been diering views asto which is the predominant mechanism, with junctionheating gaining the majority of attention. A number ofstudies have used various surface temperature monitoringtechniques and have found Joule loss to be a secondarymechanism [30,38,39]. However, it has recently been sug-gested that the structure of the workpiece itself determinesthe nature of the heating mechanism that takes place [40].

    3.1.1. Joule lossbre heating

    Fibre heating is the result of Joule losses due to theinherent resistance of the bres and is therefore dependenton bre length, resistivity and cross-sectional area. Mit-chang et al. [27] singled the bre heating mechanism asthe primary source of heat. Using infra-red camera obser-vations, negligible dierences were found in the tempera-tures obtained between carbon bre weave/PPS matrixand virgin carbon bre woven fabric. This led to the con-clusion that the matrix provided little contribution to heat-ing and therefore bre heating was prevalent. Similarresults have been found by Lin et al. [25]. However, forbre heating to occur, it has also been found that thereneeds to be a very low contact resistance between perpen-dicular bres, which occurs when there is a very high inci-dence of bre contact [40].

    3.1.2. Junction heatingdielectric hysteresis heating

    This rst type of junction heating is based on the obser-vation that the bres of the prepreg or preconsolidatedlaminates at the bond line are separated by very thin layersof matrix material. Upon the application of an alternatingelectric eld, a potential dierence is created between theing, (b) dielectric hysteresis and (c) contact resistance.

  • where contact resistance is higher. Weaves and knitted fab-ric will show bre heating dominance due to a larger areaof contact between the bres and thus a lower contact resis-tance. Moreover, the processing parameters can shift theheating mechanism from junction heating to joule loss[40]. As the workpiece heats, the viscosity of the matrixlowers and, upon the application of sucient pressure,squeeze-out of the matrix occurs. This results in higherbre contact and bre-dominated heating results [40].

    3.1.5. Hysteresis loss

    The nal heating process that can occur is due to hyster-esis losses within magnetic materials [11,12]. When a mag-

    es: Part A 37 (2006) 16381651 1643bres, and a capacitor eect is created. Dielectric heatingoccurs due to the movement of charge and rotation ofthe molecules between the bres. The inuence of the inter-sections on the heating behavior has been researched byFink et al. [30]. Dielectric heating can be modelled as a con-ductive loop with a resistance and a capacitor placed inparallel. The resistance between the bres can be calculatedby [40]

    Rdh hwe0kd

    2f tan d

    3

    where h is the distance between the bres, e0 is the permit-tivity of a vacuum, k and tand are the dielectric constantand the dissipation factor of the polymer respectively,and df is the diameter of the bre. From this, and fromthe support of developed models, Gillespie et al. [41] haveconcluded that to maximise the dielectric heating eect ofcross ply or angle ply laminates, the ply thickness aboveand below the interface and the bre volume fractionshould be maximised, and bre diameter and interply resinthickness minimised. As further support to the dielectricheating mechanism, Fink et al. [42] have observed that dif-ferent polymers heat to dierent degrees and this dierenceis attributed to the dielectric properties of the polymers.

    3.1.3. Junction heatingcontact resistance heatingThe nal heating mechanism arises when the incidence

    of bre-to-bre contact is high. In the case of higher brevolume fraction angled plies, contact resistance heatingmay be dominant and depends on the contact resistanceat the bre junction and the voltage drop across it [40].Squeeze ow of the matrix out of the laminate during con-solidation and bre waviness contributes to this incidence[43]. As a result of the contact, there is a large temperature-and pressure-dependent resistance at bre junctions whichgenerates heat [40]. Direct contact is not necessary for con-tact heating to occur, provided that the distance betweenthe bres is small enough for electrons to pass through thislayer [43]. In order to test for this type of heating mecha-nism, Yarlagadda et al. [39] applied an ac induction eldand a dc voltage to unidirectional strips arranged as asquare loop. Both cases produced eective heating at thebre junctions, thereby discounting frequency-dependentdielectric hysteresis. In addition, it was found that uniformintimate contact is necessary for more uniform heating ofthe workpiece, identifying the importance of surface rough-ness on the extent of contact heating [39,44].

    3.1.4. Joule loss versus junction heating

    Although many studies have aimed to show which is theoutright dominant heating mechanism, it is most likely thatthis depends on a number of parameters. Yarlagadda et al.[40] numerically veried that bre heating is dominant onlyin cases where the contact resistance between the bres islow. This not only depends on the type of bres that are

    T.J. Ahmed et al. / Compositbeing heated, but also on the workpiece architecture. Pre-pregs or cross plies will show a junction heating dominancenetic material is exposed to the alternating magnetic eld,the magnetic dipoles of the material change to realign withthe eld. Hysteresis indicates that energy is needed to turnaround the small internal magnets of the material to alignwith the alternating magnetic eld, as shown in Fig. 6. Asthe magnetic dipoles rotate, they vibrate and energy is lostin the form of heat due to friction. Of the composite bresthat are currently used, none are magnetic and thereforehysteresis loss is not applicable. However, if a metallic sus-ceptor is introduced into the workpiece, hysteresis lossbecomes a source of heating.

    This heating process only occurs up to the point of theCurie temperature of the magnetic material; the point atwhich the material becomes non-ferromagnetic. At thispoint, the material can no longer generate heat and sustainsits Curie temperature even when a higher magnetic eldstrength is applied [46]. In this way, good temperaturecontrol of the weld can be maintained [46].

    3.2. Heating elements

    In cases where the adherends are not magnetically sus-ceptible, or controlled and localised heating is necessary,heating elements are used as inserts in the weld. Two maintypes of heating elements are available for the process ofinduction welding which are common to all types of electro-magnetic welding, and come in the form of a powder ormesh. Similarly with the bre case, the prerequisites are thatthey are susceptible to the eects of an electromagnetic eld,Fig. 6. Hysteresis loss [11].

  • have enough electrical resistance to produce heat and forma conductive closed-loop network. Therefore any electri-cally conductive material can be used as a heating element[47]. The use of inserts as a magnetic susceptor has a fewadvantages over using the bres themselves. Firstly, heatcan be provided exactly where is needed and thermalstress build-up is prevented in other areas of the workpieceand assembly [13]. The susceptors may also be coated inresin which help to ll voids in the weld zone and also canbe a blend of two matrices in the case of joining dissimilarthermoplastic materials. Finally, non-conductive bres,such as glass or aramid, are not excluded and can also bewelded. Many studies have focused on the use of metallicpowder interdispersed in thermoplastic resin [13,15,46].Generally, the frequency needed to heat the weld with suchinserts are up to one order of magnitude higher than bre

    inserts. Finally, depending on the area to be welded, thereis also the importance of the eect of the weight penaltystructures where weight reduction is a premier design goal[49].

    3.2.1. A comparison of weld congurationsThe simplest way to compare a weld with a susceptor to

    one without is to use lap shear strength (LSS) data. LSSdata has most commonly been used in the literature as anindicator of the quality of the weld. Table 1 displays thestrength values for various laminates, weld and workpiececongurations obtained from previous studies. It shouldbe indicated that bond strength depends on the weldparameters, which will be described in Section 4, and mate-rials used to produce the joint, accounting for the range ofvalues listed. It is also possible that LSS data does not give

    1644 T.J. Ahmed et al. / Composites: Part A 37 (2006) 16381651inserts and the more costly vacuum tube power source areneeded.

    Metal meshes have emerged as eective susceptors[33,35,47] but there are a few important parameters aect-ing the eectiveness of these inserts. For good bonding ofthe mesh inserts to the adherends, sucient resin must beavailable and hence embedding the mesh in the requiredpolymer is necessary. Studies that have focused on metalmesh susceptors have found problems with uniform heat-ing of the composite laminate aggravated by the non-uni-formity of the magnetic eld generated by the inductioncoil [48]. Yarlagadda et al. [48] went some way to solvingthis by selectively removing segments of the mesh. How-ever, the insert could act as a contaminant, inducing stressconcentrations and residual stresses due to dierences inthermal expansion, and environmental degradation ofthe weld [12]. Mahdi et al. [28] found that there was pooradhesion between the metal mesh insert and resin used.Although not a largely signicant loss in shear strengthwas observed during lap shear testing, the eects of suchpoor adhesion under dierent loadings, such as fatiguecycles, could prove to be detrimental. This highlights theimportance of adequate surface preparation of such

    Table 1Comparison of lap shear strength values

    Reference LSS (MPa)

    Border and Salas [12] 27Cogswell et al. [31] 31Schwartz [5] 3848Mitschang [27] 30van Wijngaarden [32] 25Cogswell et al. [31] 36Border and Salas [12] 44Todd et al. [7] 33Williams et al. [37] 46Nagumo et al. [35] 1722Hodges et al. [33] 4148Whitworth [51] 27van Wijngaarden [32] 18

    Suwanwatana et al. [46] 20van Wijngaarden [32] 10a clear picture about the quality of the weld and for thisreason, various studies have instead used a number of othertests such as double cantilever beam, fatigue and exure[21,28,50,51]. Such research has found that induction weld-ing produces comparable, if not better, joints in compari-son to oven-cured or bolted joints.

    3.3. Edge eects

    One of the major issues associated with induction weld-ing is an eect arising from the geometry of the weld zone.This so-called edge eect results from a coils proximity toan edge of the workpiece. As an example, if a simple circu-lar pancake coil is considered, eddy currents induced in theworkpiece create global current loops that are circular innature. Fig. 7(i) shows the eddy current path producedby such a coil and an example of the corresponding temper-ature prole taken across line AA, for a workpiece that islarger than the coil. At the edges, and especially at the cor-ners, there is a large area for eddy currents to ow. Thisresults in lower current densities in these regions and lessheat is generated, as shown by the lower temperatureproles at the edges of the workpiece [23].

    Laminate type Weld conguration

    Carbon/PEEK No insertCarbon/PEEK No insertCarbon/PEEK No insertCarbon/PPS No insertCarbon/PPS No insertCarbon/PEEK PEEK lm insertCarbon/PEEK PEEK lm insertCarbon/PEEK PEI/PEEK lmCarbon/PEEK Woven carbon bre insertCarbon/PEEK Metal meshCarbon/PEEK Metal mesh and PEEK insertCarbon/PEKK PEKK lm insertCarbon/PPS Expanded metal foil

    Glass/PPS Nickel/PSU lm insertGlass/PPS Expanded metal foil

  • be shortly addressed in this section. When designing aninduction heating set up it can be a help to know the causeof these eects for problem solving.

    A main consideration for design is how to control andconcentrate the magnetic eld onto the workpiece. How-ever, due to eld interactions inherent to the coil, theresulting heat zone is not symmetrical. The generated heatis directly related to the power inside the workpiece. Tomake a good prediction of transferred power inside the

    es: Part A 37 (2006) 16381651 1645If the size of the workpiece is reduced, as shown inFig. 7(ii) and (iii) the currents are unable to follow theshape of the coil. In order to create closed-loop paths,the eddy currents are then forced to travel along the edgeof the laminate in closest proximity to the coil [23]. Highercurrent densities and higher temperatures in these regionsresult, as indicated by the temperature proles.

    It is the higher temperatures that arise at the edge of theworkpiece that is the most dicult to eliminate and therehave been some eorts to minimise edge eects, or to avoidthem altogether. The simplest and most common method isto use models to predict where excessive edge heating mayoccur. Changes to the coil design can then be made tocounteract this eect [25,27]. However, the procedurebecomes more complex as the coil design, workpiece geo-

    Fig. 7. An example of edge eects resulting from changes in workpiecegeometry.

    T.J. Ahmed et al. / Compositmetry and layup becomes more complex.Another method has been aimed at preventing edge

    eects of susceptor materials placed at the weld line. Theprocess involves redirecting eddy current ow paths inmetal mesh susceptors by selectively cutting patterns inthe mesh. Once again, models can be used to rstly predictheat generation for a given mesh conguration and thenoptimising cut patterns to create more even heating in areaswhere overheating may occur [48]. Along similar lines,mesh susceptors with solid foil edges, i.e. edges with zeromesh opening, can also work to reduce the eect of edgeeect. In this way, wherever the current density is higher,the resistance is lower and therefore the temperature canbe reduced [52].

    3.4. Additional heating eects

    The magnetic eld for induction heating applicationseld is created by an induction coil which can be of almostany shape to t the application. The more complex theshape of the coil, the more instances where the magneticeld of dierent parts of the coil interact. These eects willworkpiece it is important to know each of the dierentcauses for an asymmetrical power density and thus asym-metrical heating zone.

    3.4.1. Proximity eect and its derivativesWhen a single conductive wire is considered to carry an

    alternating current the current distributes itself equallyover the surface. When a second conductive wire is placedin the vicinity of the rst the electric eld of the two cur-rents inuence each other and the eld distributionbecomes asymmetrical. A change is therefore generatedon the associated magnetic eld. In the case of oppositecurrent directions the magnetic ux lines are concentratedin between the wires as a result of a higher current density.On the opposite sides, the magnetic eld strength is lessstrong than in the single wire condition. In the case oftwo identical current directions the magnetic ux linesare driven out of the center between the two wires andthe magnetic eld becomes more stretched as can be seenin Fig. 8. The ring eect due to ring-shaped coils is oneof the more known examples of the proximity eect. Themagnetic ux density inside the coil is higher than theexpected magnetic ux density because of interactions ofthe magnetic eld between sections along the ring [17].For induction heating application this means that the mostecient heating occurs inside the induction coil.

    The proximity eect also occurs between the current inthe induction coil and the induced eddy currents in theworkpiece. A large part of the current in the coil is forcedto ow along the surface that is closest to the workpiece.The induced current in the workpiece is always the mirrorimage of the coil geometry, thus the same shape, but ow-ing in an opposite direction. This is a favourable eect,because the currents are drawn to each other. To enhancethe proximity eect between the induction coil and theworkpiece a C-shaped ux concentrator can be placed overthe top of the inductor [17]. A ux concentrator is aFig. 8. Inuence of current direction on magnetic eld lines. (a) Singlewire, (b) same current directions and (c) opposite current directions.

  • magnetically conductive material that is able to provide aneasier path for the magnetic ux to travel and conduct themagnetic elds more eciently and eectively than air [17].Three groups of materials may be used for ux concentra-tors [53]; laminations of silicon steel, ferrites and magneto-dielectric materials, which are made from magneticparticles dispersed in an electric insulator. Flux concentra-tors can direct, control, and focus the magnetic elds into aspecic area of the work coil while keeping it away fromareas that do not need heat. The result is that almost allthe current is drawn towards the open end of the concen-trator and the coil area closest to the workpiece. The e-

    it is due to the alternating magnetic eld that eddy currentsare induced in the laminate. As already described in Section3, the frequency also has an eect on the reference depth;the higher the frequency, the lower the reference depth.However, Rudolf et al. [24] conrmed through experimen-tation that the time to heat the composite laminate to thedesired temperature decreases quadratically with increasingfrequency. Thus for a greater generation of energy withinthe laminate to be produced, a higher frequency is desired.Consequently this also leads to a more shallow referencedepth and these two conditions need to be balanced.

    4.2. Power

    The power input is one of the most important parame-ters of the process because the amount of heat generated

    1646 T.J. Ahmed et al. / Composites: Pciency of the coil is increased [11] and localised heatingcan be enforced. Fig. 9 describes diagrammatically thiseect where the darker shaded areas are regions of highercurrent intensity.

    3.4.2. Skin eect

    When a direct current ows through a conductive mate-rial the current distribution over a cross-sectional area ofthis conductive material is uniform. When an alternatingcurrent is applied to the same conductive element the distri-bution becomes non-uniform. The induced current tends toow outwards at the surface of the material rather thanpenetrate the cross section with the same intensity. Thiseect is called the skin eect [17]. The result of the skineect is that most of the heat is generated in a specicregion on the surface. To have some predictive values ofthe skin eect, the reference depth is used and dened asthe depth in which the eddy current density has decreasedby 1/e, or 37% [54]. The reference depth depends on mate-rial properties such as electrical resistivity, q, and magneticpermeability, lr, of the workpiece, and the eld frequency,f, which is the same frequency as the magnetic eldfrequency of the coil [55].

    d q

    plrf

    r4

    As can be seen from Eq. (4) a higher frequency leads to asmaller reference depth and thus a more shallow skin eect.Also temperature eects need to be taken into account be-cause the magnetic permeability and electrical resistivityFig. 9. Magnetic ux concentrator [11].are both a function of temperature. Fig. 10 shows the eectof the variables represented in Eq. 4 on penetration depth.

    It is possible to divide electromagnetically thick and thinbodies with regard to workpiece materials. The rst is thetype of body from which the penetration depth is less thanthe bodys thickness, while the second is the type of bodywhere the reference depth is greater than the bodys thick-ness. In this type of body there is no inuence of the fre-quency. For induction heating applications it is generallyfavourable to keep the body electromagnetically thin tocreate a heat zone throughout the total thickness of thematerial for good consolidation of the composite part[23], whereas for welding applications, it can be more desir-able simply to heat until the interface [54]. The referencedepth can also be restricted to the surface in highly conduc-tive materials through cancellation of the coils magneticeld in deeper regions of the workpiece.

    4. Induction heating parameters

    4.1. Frequency

    The current frequency is a fundamental parameter, since

    Fig. 10. Variables aecting penetration depth [56].

    art A 37 (2006) 16381651in a specic region of the material is proportional to thepower generated in that same region. The source and the

  • workpiece are therefore heavily coupled and the generatedpower is dened by [24]

    P 2pflrHIA2

    R5

    where H(I) is the magnetic eld intensity, which is depen-dent on the current of the equipment, and A the cross-sec-tional area of the conductive loop in the workpiece. Theamount of heat generated in the workpiece is also propor-tional to the frequency squared. This means that when themagnetic eld intensity drops in distance from the coil tothe workpiece, the drop in generated power can be com-pensated by a rise in frequency. Fig. 11 displays the depen-dence of heating time with input power and coil/workpieceseparation.

    From a processing point of view, the starting point ofthe energy needed to heat an ideally insulated conductiveworkpiece of mass mw, can be the required temperaturerise, DT:

    4.3. Pressure

    Appropriate pressure application is important for highquality consolidation because it allows for good intimatecontact. However, Rudolf et al. [26] found that there is apractical limit to the pressure that can be applied. The con-tinuous welding of carbon bre-reinforced thermoplasticswas investigated and found that a higher pressure resultedin a lower quality of weld. This was attributed to increasedsqueeze-out of matrix at the weld and a compromise mustbe met between adequate intimate contact and polymersqueeze-out. Poor pressure application can result in thefollowing faults.

    4.3.1. Voids

    The appearance of voids have a close relationship withthe deconsolidation of the matrix material and there area number of reasons that are the cause. The release of elas-tic energy as bre bundles deform under pressure and heat,

    T.J. Ahmed et al. / Composites: Part A 37 (2006) 16381651 1647E Pwt mwcDT 6where c is the specic heat of the workpiece. However, dueto eciency losses between the coil and the workpiece thetotal power needed to be provided by the power source ishigher during the time of heating.

    The power inuences the heating time in general. Forwelding applications to be valuable, short processing times,thus heating times, are needed. However, as will beexplained, this must be oset with the quality of the weldand hence a compromise must be met. When designingthe total system the heat time can be the starting pointfor calculating the required power, because the otherparameters such as the resistivity and specic heat arerelated to the material and therefore vary within vastboundaries.Fig. 11. Inuence of powerexpansion of entrapped gas bubbles, collapsing of air pock-ets, inserts and thermal stresses due to the removal of pres-sure before cooling to below the required temperature, allcontribute to the production of voids [57]. For the entrap-ment of air bubbles, the surface roughness is an importantparameter. To prevent voids it is important to strive for ahigh surface smoothness and apply a high weld pressure tosuppress the occurrence of voids [27]. The high pressureconstraint can cause a contradiction with the low pressureto avoid folds and ashes, described later. A further conse-quence of the creation of voids is delamination. Most of thetime delamination is a result of extreme deconsolidation. Inthis case relatively large air pockets appear in the matrixmaterial that will separate one layer from the other in thelaminate and thus destroy the interaction between thelayers.with heating time [24].

  • 4.3.2. Cracks

    Rudolf et al. [26] reported cracks due to the high degreeof crystallization of the semi-nished product, whichresulted in strong shrinkage of the matrix material. Thesecracks occurred due to thermal stresses due to the thermalexpansion mismatch between the matrix and bres. Crackprevention is possible through the control of pressurethrough the cooling phase. The matrix is thus preventedfrom expanding and shrinking. The pressure can then beremoved when a uniform temperature prole is reachedand the temperature is well below the melt or glass transi-tion temperature.

    4.3.3. Folds and ashes

    where the quality of the weld improves with residence timeand temperature as Tg is reached and exceeded. Thisbecomes the optimum time and temperature range forwelding, and hence the optimum process window. Finally,if temperature within the workpiece exceeds the maximumwelding temperature of the polymer, thermal degradationof the polymer and a consequent degradation of weldstrength results.

    5. Modelling induction welding

    Due to the complex structure of the composite materi-

    1648 T.J. Ahmed et al. / Composites: Part A 37 (2006) 16381651These faults are due to misalignment and poor applica-tion of pressure. When the pressure is unequally dividedover the weld zone it can force matrix material out the weldat the sides, causing ashes, or it can result in the folding ofthe laminate at the edge of the pressure device. The foldingcan cause bres to buckle. To prevent these ashes it isimportant not to apply a very high welding pressure. Veryhigh pressure forces the matrix out of the welding zone [26].To prevent the folds it is important to have a very smoothunder surface, because when applying pressure an unequalplate can force a fold into the laminate.

    4.4. Residence time

    The residence time is the time of exposure of the work-piece to the induction eld and has an inuence on themovement of polymer molecules across the weld interface.In general, allowing for a longer residence time results in ahigher quality of weld because more polymer chains aregiven time to move across the weld interface [29,46]. Ifwelding parameters frequency, power and pressure are con-sidered to be constant, three welding regimes can be estab-lished relating to the residence time and resultingtemperature; non-wetting, uniform fusion and degradationand are shown in Fig. 12. Insucient weld times, and there-fore low temperatures, result in insucient wetting andweld strength is low. A period of uniform fusion follows,Fig. 12. Dierent welding regimes of thermoplastic composites [29].als, the modelling eort was mainly focused on the heatgeneration mechanisms in the carbon bre-reinforced com-posite materials. Several concepts that covered virtually allpossible heat generation mechanisms were proposed,discussed and investigated.

    Miller et al. [23,25] proposed Joule heating as a domi-nant heating mechanism in preconsolidated carbon bre-reinforced thermoplastic materials. They also introducedand proved the existence of the global current loop(between the adjacent plies) as the major path of theinduced current in the laminate. A theoretical model wasdeveloped that assumed near perfect electrical contactbetween crossed plies. The conclusions were that Jouleheating is the primary heat generation mechanism and thatelectrical ply to ply transfer is either real, by arcing, or vir-tual, by displacement currents (as in a capacitor). Themodel is limited to preconsolidated laminates and is notapplicable for cases when good electrical contact betweenthe ply bres cannot be provided.

    Fink et al. [30,38,58,42] proposed an alternative conceptthat indicates dielectric heating in the matrix region at theply junctions as a dominant heat generation mechanism inlaminates in which direct contact between the bres fromadjacent plies does not exist. Their major proposition statesthat the primary heating mechanism in multi-directional,thermoplastic composites subjected to a transverse mag-netic eld alternating at less than 200 kHz is dielectriclosses in the polymer region between bres in adjacentplies, which make up a conductive path of signicantdimension [38]. The theory was developed using a simpleidealized conductive loop, illustrated in Fig. 13, that con-Fig. 13. An idealized conductive loop in a [0,90] cross-ply [38].

  • sists of two parallel carbon bres in one plane and twomore in an adjacent plane. A lengthy theoretical analysisof the dielectric heating produced an equation that pro-vided a comparison between the local contribution of theJoule loss in the bre, Pi, and the dielectric loss in the junc-tion,Wj. The comparison is given by taking the ratio of thetwo sources of heat [38]:

    W jP i

    3:35 1013h 7

    Eq. 7 clearly shows that the thickness of the junctionsshould be of order 1014 [m] for the Joule losses to be com-parable to the losses in the polymer (an electrical break-down would be expected at that thickness), whichsupported the proposed theory that power loss throughdielectric heating in the polymer region of cross-over junc-tions is the dominating heat generation mechanism [38].

    Yarlagadda et al. [39] considered contact heating atbrebre junctions as an important heating mechanismin case of unconsolidated prepreg stacks. Their studyshowed that contact heating is indeed major heating mech-anism in these cases. This statement was supported by acomparative experiment performed on an unidirectionalloop connected to a 30 V dc power source and a loopheated by an induction coil, both producing the same heat-ing rates and temperature proles. Several numerical mod-els were developed based on this concept [39,40,59]. Therst one, a conductive loop network model [39] formulatedfor two-ply heating predictions accounted only for Jouleheating in the bres and junction heating, so a choicebetween dielectric and contact heating mechanisms has tobe made prior to the computation. Modelling the newlyproposed concept of contact heating produced good, satis-factory qualitative results. Later on, a couple of experimen-

    T.J. Ahmed et al. / Composites: Part A 37 (2006) 16381651 1649Several theoretical models for unconsolidated laminateswere developed based on the concept of dielectric heating[30,38,58,42]. An extensive experimental study was per-formed in order to verify the model data. The reported re-sults agreed well with the model predictions [30] andalthough in some cases the results could not provide a di-rect proof of the proposed model [42], they strongly sup-ported the concept of dielectric heating. The Jouleheating in the bres was excluded by observing the heatingpatterns, which showed substantial heating only at thepoints of ply overlap. At relatively smaller thicknesses ofthe polymer region Joule heating through brebre con-tact at the junctions was allowed for, but only as a possiblecontributing mechanism. Experiments with dierent resinmaterials interlayer showed signicant dierence in theheating characteristics [42], discounting by that the theoryof virtual charge displacement and further supporting theproposal of dielectric losses in the polymer. Finally, dielec-tric breakdown was dismissed by performing cycle teststhat showed no change in the heating rate between the rstand later cycles [38,42].Fig. 14. Schematic of the bre heating and junction heatingtal methodologies were developed for estimating thethrough-thickness contact resistance of the bre materials[60,40] that were used as an input to induction heatingmodels. The next step was developing a unied modellingapproach in order to determine the dominating heat gener-ation mechanisms for any type of composite system (dier-ent bre and matrix material and laminate congurations),as well as for dierent processing parameters [40]. This wasprovided by using a representative electrical circuit todescribe the heating mechanisms and non-dimensionalparameters for their comparison. The model accountedfor all three possible heat generation mechanisms, as it isschematically shown in Fig. 14. A comprehensive paramet-ric study was performed that resulted in a design map thatcan determine the expected dominant heating mechanism,given the composite system. The results further showedthat for carbon bre composite systems junction heatingeects mostly dominate compared to Joule bre heating,although it was noted that the bre architecture plays sig-nicant role in determining the dominant mechanism(woven fabrics may in some cases show dominant breat each conductive loop of the cross-plied laminate [60].

  • s: Pheating, due to low contact resistance). The results alsosupported the notion of contact heating as dominant junc-tion heating mechanism. Finally, a combined numericalmodel was presented [59] to predict in-plane heating gener-ation for unconsolidated prepreg stacks. The modelaccounted for all three heating mechanisms combined,the main process and material parameters and the stackangle. The models capability to predict the bre or junc-tion heating dominance was veried by its comparison withFink and Miller models. The comparison with Fink modelshowed excellent agreement, except at the edges of thedomain (where it was expected for the new model to pro-vide more accurate predictions). The comparison withMiller model also showed excellent agreement in aniso-tropic heating patterns. An experimental validation wasperformed for by comparing three dierent stacking caseswith the model results. Again there was an excellent agree-ment in overall heating patterns and distances between theheated spots, except for very low stack angles, when modelresults deviated signicantly from the experimental ones.

    Rudolf et al. performed an extensive experimental studyon induction heating of carbon bre-reinforced thermo-plastics [24], focused on the inuence of the major processparameters on the heat rate and heat distribution in thematerial. Several dierent matrix materials and fabric typeswere investigated, as well as dierent dierent laminatecongurations, including a single fabric layer. From theheat generation viewpoint it was concluded that Joule heat-ing in the bres is the most likely dominate heat generationmechanism, which supported the theory of Miller et al.[23,25] and excluded the dielectric heating theory of Finket al. [38]. Based on this ndings, a transient thermal modelof the continuous induction welding process was developed[27] in order to ease and improve the optimisation of thewelding process parameters. The induction heating phaseof the process was modelled with a nite element modelthat accounted for the anisotropy of the composite mate-rial and for the temperature dependency of the materialproperties and boundary conditions. Since the initial two-dimensional models did not produce satisfactory results,a full three-dimensional model of a single lap joint wasbuilt. The inuence of the edge eect was eliminated bychoosing suitable dimensions and the laminate materialwas assumed to be monolithic and orthotropic. The com-parison between the model predictions and experimentalmeasurements showed a good agreement of the in-planetemperature distribution and transient temperature pro-les. The cooling phase was modelled using relatively sim-ple nite dierence model based on the Fouriers law ofthermal conduction. The model used dierent thermal con-ductivity coecients in the main directions, but the usedmaterial properties were constant, averaged on the pre-dicted temperature range. The model produced results withaccuracy within the 10% of the measured temperatures,which was considered sucient for predicting the optimum

    1650 T.J. Ahmed et al. / Compositeprocess parameters.6. Conclusion

    Induction heating has already proven to be a worthwhiletechnology for metals. The past two decades have seen theemergence of induction heating as a suitable and eectivetechnology for heating of thermoplastic composites. Thesimplicity of the physical process and the extensive researchinto numerical modelling of the heating process hasallowed for the development of the induction welding pro-cess. A number of studies that have been performed clearlyshowed the potential of induction welding for applicationin thermoplastic composite structures. Produced lap shearstrengths were comparable, if not better, than oven-cured,bolted or resistance welded joints.

    In spite the considerable research eort, the inductionprocess has not seen signicant shift in application inindustry from current bonding and welding methods.Apart from the inherent inertness of the industry whenintroducing new technologies is concerned, there are sev-eral issues, most notably the edge eect and the local heat-ing eect, that prevent embracing induction welding on alarge scale. Addressing these and other important issuesremains as an incentive for further development of theinduction welding method.

    References

    [1] Bersee HEN. Diaphragm forming of continuous bre reinforcedthermoplastics. Delft University Press; 1996.

    [2] Leach DC. Continuous bre reinforced thermoplastic matrix com-posites. Adv Comp 1989:43109.

    [3] Chang IY, Lees JK. Recent developments in thermoplastic compos-ites: a review of matrix systems and processing methods. JThermoplast Comp Mater 1988:27796.

    [4] Harper CA. Handbook of plastics, elastomers and composites. 4thedition. New York: McGraw-Hill; 2002.

    [5] Ageorges C, Ye L, Hou M. Advances in fusion bonding techniquesfor joining thermoplastic matrix composites: a review. Composites:Part A 2001;32:83957.

    [6] Stokes VK. Joining methods for plastics and plastic composites: anoverview. Polym Eng Sci 1989;29:131028.

    [7] Todd SM. Joining thermoplastic composites. In: 22nd InternationalSAMPE Technical Conference, vol. 22, 1990. p. 38392.

    [8] Threadgill PL, Fernie JA, Watson MN. Progress in joining advancedmaterials. Weld Met Fabricat 1991:17984.

    [9] Silverman EM, Griese RA. Joining methods for graphite/peekthermoplastic composites. Sampe J 1989;25(5):347.

    [10] Yousefpour A, Hojjati M, Immarigeon J-P. Fusion bonding/weldingof thermoplastic composites. J Thermoplast Comp Mater 2004;17:30340.

    [11] Haimbaugh RE. Practical induction heat treating. ASM Inter-national; 2001.

    [12] Border J, Salas R. Induction heated joining of thermoplasticcomposites without metal susceptors. In: 34th International SAMPESymposium, 1989. p. 256978.

    [13] Chookazian SM. Electromagnetic welding: an advance in thermo-plastics assembly. Mater Des 1987;8:415.

    [14] Chookazian SM. Electromagnetic welding of thermoplastics andspecic design criteria with emphasis on polypropylene. In: ANTEC94, 1994. p. 13525.

    [15] Stokes VK. Experiments on the induction welding of thermoplastics.

    art A 37 (2006) 16381651In: ANTEC 2001, 2001. p. 125661.

  • es: P[16] Zinn S, Semiatin SL. Elements of induction heating: design, controland applications. ASM International; 1988.

    [17] Rudnev V, Loveless D, Cook R, Black M. Handbook of inductionheating. New York, USA: Marcel Dekker; 2003.

    [18] Simpson PG. Induction heating; coil and system design. New York,USA: McGraw-Hill; 1960.

    [19] Loveless D, Cook R, Rudnev V. Considering nature and parametersof power supplies for ecient induction heat treating. Ind Heat1995:337.

    [20] Zinn S, Semiatin SL. Coil design and fabrication: basic design andmodications. Heat Treat 1998:326.

    [21] Benatar A, Gutowski TG. Methods for fusion bonding thermoplasticcomposites. Sampe Quart 1998;18(1):3441.

    [22] Lin W, Buneman O, Miller AK. Induction heating model for graphitebre/thermoplastic matrix composites. Sampe J 1991;27.

    [23] Miller AK, Chang C, Payne A, Gur M, Menzel E, Peled A. Thenature of induction heating in graphiteber, polymermatrix com-posite materials. Sampe J 1990;26:3754.

    [24] Rudolf R, Mitschang P, Neitzel M. Induction heating of continuouscarbon-bre-reinforced thermoplastics. Comp Part A: Appl SciManufact 2000;31:1191202.

    [25] Lin W, Miller AK, Buneman O. Predictive capabilities of aninduction heating model for complex-shape graphite ber/polymermatrix composites. In: 24th International SAMPE TechnicalConference, 1992. p. T60620.

    [26] Rudolf R, Mitschang P, Neitzel M. Welding of high-performancethermoplastic composites. Polym Polym Comp 1999;7:30915.

    [27] Mitschang P, Rudolf R, Neitzel M. Continuous induction weldingprocess, modelling and realisation. J Thermoplast Comp Mater2002;15:12753.

    [28] Mahdi S, Kim H-J, Gama BA, Yarlagadda S, Gillespie Jr JW. Acomparison of oven-cured and induction-cured adhesively bondedcomposite joints. J Comp Mater 2003;37(6):51942.

    [29] Zach T, Lew J, North TH, Woodhams RT. Joining of high strengthoriented polypropylene using electromagnetic induction bonding andultrasonic welding. Mater Sci Technol 1989;5:2817.

    [30] Fink BK, Gillespie Jr JW, McCullough RL. Experimental vericationof models for induction heating of continuous-carbon-ber compos-ites. Polym Comp 1996;17(2):198.

    [31] Cogwell FN, Meakin PJ, Smiley AJ, Harvey MT, Brooth C.Thermoplastic interlayer bonding for aromatic composite structures.In: 34th International SAMPE Symposium, 1989. p. 231525.

    [32] van Wijngaargen MJ. Welding technologies for a generic carbon berreinforced thermoplastic assembly. In: 25th International SAMPEConference, 2004. p. 5560.

    [33] Hodges WT, Tyeryar JR, Berry M. Bonding and nondestructiveevaluation of graphite/peek composite and titanium adherends withthermoplastic adhesives. In: Society of Manufacturing Engineers,Conference on Fabricating Composites, Hartford, CT, Conneticut,USA, 1985. p. 19852001.

    [34] Lawless GW, Reinhart TJ. A study of the induction heating oforganic composites. In: SAMPE Conference: Advanced Materials:Meeting the Economic Challenge, Toronto, Canada, 1992. p. T37584.

    [35] Nagumo T, Nkamura H, Yoshida Y, Hiraoka K. Evaluation of peekmatrix composite. In: 32nd International SAMPE Symposium, 1987.p. 396407.

    [36] Wedgewood AR, Hardy PE. Induction welding of thermoset com-posite adherends using thermoplastic interlayers and susceptors. In:28th International SAMPE Technical Conference, 1996. p. 85061.

    [37] Williams G, Green S, McAfee J, Heward CM. Induction welding ofthermoplastic composites. In FRC90Proceedings, IMechE, 1990.p. 1336.

    [38] Fink BK, McCullough RL, Gillespie Jr JW. A local theory of heatingin cross-ply carbon ber thermoplastic composites by magneticinduction. Polym Eng Sci 1992;32(5):35769.

    T.J. Ahmed et al. / Composit[39] Yarlagadda S, Kim HJ, Gillespie Jr JW, Shevchenko N, Fink BK.Heating mechanisms in induction processing of carbon ber rein-forced thermoplastic prepreg. In: 45th International SAMPESymposium, 2000. p. 7989.

    [40] Yarlagadda S, Kim HJ, Gillespie Jr JW, Shevchenko NB, Fink BK.A study on the induction heating of conductive ber reinforcedcomposites. J Comp Mater 2002;36(4):40121.

    [41] Gillespie Jr JW, McCulough RL, Fink BK. Induction heating ofcross-ply carbon-ber composites. In ANTEC 92, 1992. p. 21069.

    [42] Fink BK, McCullough RL, Gillespie Jr JW. A model to predict thethrough-thickness distribution of heat generation in cross-ply carbon-ber composites subjected to alternating magnetic elds. Comp SciTechnol 1995;55:11930.

    [43] Wang S, Kowalik D, Chung DDL. Eects of the temperature,humidity, and stress on the interlaminar interface of carbon berpolymermatrix composites, studied by contact electrical resistivitymeasurement. J Adhesion 2002;78:189200.

    [44] Fosbury A, Wang S, Pin YF, Chung DDL. The interlaminarinterface of a carbon ber polymermatrix composite as a resistanceheating element. Composites: Part A 2003;34:93340.

    [45] Kim HJ, Yarlagadda S, Fink BK, Gillespie Jr JW. Through-thicknessheating behavior of carbon ber reinforced prepreg stacks ininduction heating process. In 34th International SAMPE TechnicalConference, 2002. p. 1192204.

    [46] Suwanwatana W, Yarlagadda S, Gillespie Jr JW. Induction bondingof composite materials using nickel/polysulphone lms. In: 34thInternational SAMPE Technical Conference, 2002. p. 102639.

    [47] Stavrov D, Bersee HEN, Beukers A. The inuence of the heatingelement on resistance welding of thermoplastic composite materials.In: Proceedings of ICCM-14 Conference, San Diego, CA, USA, 2003.p. 1581.

    [48] Yarlagadda S, Fink BK, Gillespie Jr JW. Resistive susceptor designfor uniform heating during induction bonding of composites. JThermoplast Comp Mater 1998;11:32137.

    [49] Beevers A. Welding: the way ahead for thermoplastics? Eng AdvComp Eng Suppl 1991;231(10):112.

    [50] Ramulu M, Stickler PB, McDevitt NS, Datar LP, Kim D, JenkinsMG. Inuence of processing methods on the tensile and exureproperties of high temperature composites. Comp Sci Technol2004;64:176372.

    [51] Whitworth HA. Fatigue evaluation of composite bolted and bondedjoints. J Adv Mater 1998;30:2531.

    [52] Selvaged susceptor for thermoplastic welding by induction heating,April 1996. Patent no: US5508496.

    [53] Runi RS, Runi RT, Nemkov VS, Goldstein RC. Advanced designof induction heat treating processes and work coils. Heat Treat Met1999;4:849.

    [54] Karamuk E, Wetzel ED, Gillespie Jr JW. Modeling and design ofinduction bonding process for infrastructure rehabilitation withcomposite materials. In: ANTEC 95, 1995. p. 123943.

    [55] Paul CR, Nasar SA. Introduction to electromagnetic elds. 2ndedition. New York: McGraw-Hill; 1987.

    [56] Depth of penetration and current density. Internet Publication;Accessed 20/08/04. Available from: http://www.ndt-ed.org/Educa-tionResources/CommunityCollege/EddyCurrents/Physics/depthcurr-entdensity.htm.

    [57] Ageorges C, Ye L. Resistance welding of thermosetting composite/thermoplastic composite joints. Comp Part A: Appl Sci Manufact2001;32:160312.

    [58] Fink BK, McCullough RL, Gillespie Jr JW. A model to predict theplanar electrical potential distribution in cross-ply carbon berthermoplastic composites subjected to alternating magnetic elds.Comp Sci Technol 1993;49:7180.

    [59] Kim HJ, Yarlagadda S, Shevchenko NB, Fink BK, Gillespie Jr JW.Development of a numerical model to predict in-plane heat gener-ation patterns during induction processing of carbon ber-reinforcedprepreg stacks. J Comp Mater 2003;37(16):146183.

    [60] Kim HJ, Yarlagadda S, Gillespie Jr JW, Shevchenko NB, Fink BK.

    art A 37 (2006) 16381651 1651A study on the induction heating of carbon ber reinforcedthermoplastic composites. Adv Comp Mater 2002;11(1):7180.

    Induction welding of thermoplastic composites-an overviewIntroductionInduction weldingWelding set-upPower sourceHeat stationWorkpiece and load matchingWelding fixture

    Heating processHeating of fibre-reinforced thermoplastic compositesJoule loss-fibre heatingJunction heating-dielectric hysteresis heatingJunction heating-contact resistance heatingJoule loss versus junction heatingHysteresis loss

    Heating elementsA comparison of weld configurations

    Edge effectsAdditional heating effectsProximity effect and its derivativesSkin effect

    Induction heating parametersFrequencyPowerPressureVoidsCracksFolds and flashes

    Residence time

    Modelling induction weldingConclusionReferences