Composites: Part A 37 (2006) 18471857 www.elsevier.

com/locate/compositesa

Review

Residual stresses in thermoplastic compositesA study of the literaturePart I: Formation of residual stresses
Patricia P. Parlevliet, Harald E.N. Bersee *, Adriaan Beukers
Faculty of Aerospace Engineering, Delft University of Technology, Kluyverweg 1, 2629 HS Delft, The Netherlands Received 12 October 2005; received in revised form 19 December 2005; accepted 28 December 2005

Abstract Continuous bre reinforced thermoplastic composites are increasingly applied in aircraft structures. These high-performance thermoplastic matrices need processing at high temperatures; therefore, thermal residual stresses arise due to the mismatch in coecients of thermal expansion between the bres and the thermoplastic matrix. Since residual stresses are inherently present in virtually all composite materials and inuence the properties of the composite structures signicantly, it is of utmost importance that the residual thermal stresses are taken into account in both design and analytic modelling of composite structures. In order to understand the eects of residual stresses and nd ways to decrease their magnitude or use them to our advantage, the factors responsible for the residual stress build-up need to be understood. This rst part of a literature review discusses these factors, focusing on thermoplastic composites, and in particular the material properties of their constituents and processing parameters governing stress build-up. 2006 Elsevier Ltd. All rights reserved.
Keywords: A. Polymermatrix composites (PMCs); A. Thermoplastic resin; B. Residual/internal stress

Contents 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.1. Summary of previous literature reviews. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2. Residual stress formation on three mechanical levels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Micromechanical residual stresses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1. Reinforcing fibres . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2. Thermoplastic matrix dominated properties. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2.1. Shrinkage. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2.2. Temperature difference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2.3. Processing conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2.4. Elastic properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3. Fibrematrix interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Macro-mechanical residual stress formation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Global (skin-core) residual stresses. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1. Annealing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2. Influence of processing mould material . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1848 1848 1848 1850 1850 1850 1850 1851 1851 1852 1852 1853 1854 1854 1855

2.

3. 4.

Corresponding author. Tel.: +31 152 788 175; fax: +31 152 781 151. E-mail addresses: P.P.Parlevliet@tudelft.nl (P.P. Parlevliet), H.E.N.Bersee@tudelft.nl (H.E.N. Bersee).

1359-835X/$ - see front matter 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.compositesa.2005.12.025

1848

P.P. Parlevliet et al. / Composites: Part A 37 (2006) 18471857

5.

Discussion and conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1855 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1856

1. Introduction Currently more continuous bre reinforced thermoplastic composites (TPCs) are applied in aircraft structures as is shown by the composite programs for the Airbus A380 and A350 [1]. Arguments for using thermoplastic matrices instead of thermoset matrices are mostly of economical nature: thermoplastic composites can be formed and consolidated in several minutes whereas the thermoset composites need a time-consuming cure cycle. Moreover, thermoplastics are known for their weldability, toughness, long (indenite) storage life, recyclability and chemical resistance [2,3]. However, since this class of materials is relatively new for primary and secondary aircraft structures, some of the properties are still unfamiliar and therefore reluctance exists to apply TPCs. In order to provide TPCs with a wider applicability, more knowledge regarding these materials needs to be generated, summarised, and published. One of the issues that need addressing is residual stress formation in high performance TPCs. After processing and subsequent cooling of composite laminates from the relatively high processing temperature to the service temperature, residual stresses arise due to the signicantly higher shrinkage of the matrix compared with the bres. This type of residual stress formation is present in all bre-reinforced polymers, due to their inherent inhomogeneous nature. Residual stress1 can be dened as a stress that persists in a material that is free of external forces or temperature gradients [4]. In this paper we will refer to one type of residual stress: Thermal stress, which can be dened as a residual stress introduced within a body resulting from a change in temperature [4]. Residual stresses inuence the properties of the composite structures signicantly [58], and it is of most importance that the residual thermal stresses are taken into account in both design and analytic modelling of composite structures. It was already argued in the early nineties [3], that the eects of large residual stresses generated in the processing of high performance TPCs (most often accompanied by high processing temperatures) need to be understood, and inuences on nal properties, quality and durability of the manufactured products need to be studied [9]. To assess the eects of residual stresses on these parameters, a literature study on residual stresses in continuous bre reinforced thermoplastics was carried out and an
In the literature, the terms residual stress and residual strains are often used interchangeably. In some studies, the residual matrix or bre strains are explicitly measured or calculated, and not the residual stresses. It is the eect of such strains on the developed residual stresses in a composite laminate, which may aect the composite properties.
1

attempt was made to identify the knowledge gaps in this eld. This review will consist of several parts of which this is the rst, describing the available literature on sources and mechanisms of formation of residual stresses in TPCs, related to material properties and processing conditions. To nd methods to lower the residual stress magnitude or use the stresses to our advantage, rst the factors responsible for the residual stress build-up need to be understood, not in the least to incorporate them in both design and modelling of composite structures. This part will be followed by a literature study on experimental techniques developed for determination of residual stresses in polymer matrix composites, as well as a review on the eects of thermal residual stresses on the composite material properties. 1.1. Summary of previous literature reviews From a historic point-of-view, Nairn and Zoller [10,11] were, to the best of the authors knowledge, the rst to publish an experiment-based study on thermal residual stresses in bre reinforced thermoplastic composites. They determined the stresses in a polysulfone matrix of a unidirectional graphite bre reinforced tape by means of photoelasticity. In 1988, a review of the then available literature on residual thermal stresses in continuous bre reinforced polymers (thermosets as well as thermoplastics) was published by Favre [5]. This review can be regarded as base literature when studying residual stresses in bre reinforced polymers. Favre [5] covered the origins and magnitudes of the thermal residual stresses, temperature, moisture and time eects, and the eects of residual stresses themselves, such as geometrical distortions, microcracks and delaminations due to the free edge eects. Included were predictive models and experimental techniques available at the time and proposals for reduction of residual stresses in composites. In addition, the work by Eijpe [12] is introduced by a good summary of the available literature, and several publications [5,9,13] summarised most of the work done earlier on the sources and driving forces for internal stresses in composite materials, including composite intrinsic properties, processing conditions and environmental conditions. 1.2. Residual stress formation on three mechanical levels Residual stresses in continuous bre reinforced thermoplastics are present in the laminate or composite structure immediately after processing and subsequent cooling to the service temperature. They can be regarded at three different levels based on their origins [5,9,13,14]:

P.P. Parlevliet et al. / Composites: Part A 37 (2006) 18471857

1849

 On the micromechanical level or constituent level, the mismatch in coecient of thermal expansion between the bres and the matrix is the governing parameter. Unlike a thermoset matrix, the thermoplastic matrix is heated to a processing temperature above its glass or melting temperature and subsequently solidied upon

cooling to the service temperature (often ambient conditions), where no chemical reaction should take place. The cooling involves volumetric shrinkage of the thermoplastic matrix, which is signicantly higher than the bre shrinkage. This represents an important driving force for the development of residual strains in the bres

fibre

matrix

bonded

unbonded

Fig. 1. Schematic view of the eect of cooling on the matrix around a bre. The bonded (constrained) case represents the situation after processing, where bc and cb represent tensile and compressive residual stresses, respectively, after [5].

Fig. 2. Schematic view of residual stress formation in unbalanced cross-ply laminate (a), (b) residual thermal stresses when laminate is constrained, (c) front view of out-of-plane deformation when unconstrained.

1850

P.P. Parlevliet et al. / Composites: Part A 37 (2006) 18471857

and the surrounding matrix [10,15,16]. Assuming that brematrix bonding is present during cooling, Fig. 1 shows that the result is a residual compressive stress in the bre along the longitudinal axis as well as in the radial direction [10,13], and a residual tensile stress in the matrix in the longitudinal and radial direction [8].  Macro-mechanical or lamination residual stresses are present on a ply-to-ply scale due to lamina anisotropy. The residual stresses arise due to a dierence in the transverse and longitudinal ply coecients of thermal expansion [8]. For example with cross-ply composites, the 90 bres impose a mechanical constraint on the 0 bres during cooling and vice versa, because of dierences in thermal shrinkage directions [17], see Fig. 2 (where r represents the residual stresses). The anisotropic shrinkage resembles the behaviour of a bi-metallic strip and, if unbalanced and unconstrained, deformation (curvature) of the laminate may be the result [11].  On a global [13] level, a gradient in cooling rate, temperature or moisture conditions throughout the thickness of the composite laminate or structure may result in a residual stress distribution through the thickness of the laminate. In general, a thick laminate will experience a slower cooling rate in the centre of the laminate than at the surface plies. At a certain temperature, the centre plies may still need to solidify, whereas the surface plies have already become solid. Upon further cooling, the surface plies impose a constraint on the shrinkage of the centre plies. This will often result in a parabolic distribution of compressive residual stresses in the surface plies and tensile stresses in the centre plies (Fig. 3) [9,15]. In literature, these stresses are referred to as thermal skin-core stresses [9], which in, for example, semi-crystalline composites, may result in morphological skin-core stresses. Considering the above, these three levels would result in a complex three-dimensional residual stress state within a composite structure. In order to assess the responsible parameters for residual stresses formation, these levels will

be dealt with separately. In a composite structure however, the distinction between the three is not so clear-cut. According to Favre [5], if the long-term and environmental parameters are omitted, the magnitude of residual stresses in composite structures is dependent on four parameters: temperature dierence, coecients of thermal expansion/shrinkage upon cooling of the composite constituents (or plies), elastic coecients of these constituents (or plies), and bre volume fraction. These parameters are, in turn, largely dependent on the thermoplastic matrix morphology (semi-crystalline or amorphous), type of bres, brematrix interface properties, bre morphology (woven or unidirectional prepreg), and processing conditions. These factors will be discussed separately, starting with the micromechanical level and the bre-dominated properties, followed by the matrix-dominated properties. Most of the governing parameters for residual stress formation apply to all three levels. 2. Micromechanical residual stresses 2.1. Reinforcing bres Studies on residual stresses in thermoplastic composites most often involved carbon bres (CF), with a few exceptional studies on glass bres (GF). Aramid bres were investigated in only one study [10]. Both aramid bres (AF) and carbon bres show highly anisotropic thermal expansion behaviour, with small shrinkage in the bre longitudinal direction when heated [8,18,19]. The thermal expansion behaviour of these bres can be described by separate coecients of thermal expansion (CTEs) in the directions parallel (often negative) and perpendicular (positive) to the bre axis [10,19]. For the temperature range during processing and service life of thermoplastic composites, the expansion behaviour of the bres is linear. The CTE for the reinforcing bres is in general much lower than for the thermoplastic matrices, which results in a large dierence in expansion behaviour between the matrix and the reinforcing bres. However, the perpendicular (radial) bre shrinkage upon cooling was shown to be a relatively insignicant contribution to the formation of residual stresses [13]. 2.2. Thermoplastic matrix dominated properties 2.2.1. Shrinkage The shrinkage in thermoplastic polymers depends on the matrix morphology, i.e. amorphous or semi-crystalline. In semi-crystalline thermoplastics, the volumetric shrinkage results from the densication upon crystallisation (with crystals being of higher density than the amorphous phase) in addition to the shrinkage due to the temperature drop. In amorphous polymers, the shrinkage is only due to the latter. The total shrinkage for semi-crystalline matrices was found to be ten times higher than for amorphous matrices, as can be seen in Fig. 4, where Tg is the glass tran-

Fig. 3. Laminate skin-core residual stress distribution (grey area), after [9].

P.P. Parlevliet et al. / Composites: Part A 37 (2006) 18471857

1851

load-bearing capability of the crystalline phases below that temperature [15]. The cooling rate aects the peak crystallisation temperature, which in turn inuences the stress-free temperature [13,15,23,26,28]. It can therefore be concluded that the thermal transition properties of the polymer matrix are the governing parameters for the SFT. For all polymer matrix composites applies that the further the cool-down or service temperature from the SFT, the more residual stresses reside in the composite due to the mismatch in coecient of thermal expansion (CTE) between the polymer matrix and the bres. The SFT was often determined by heating the composite until the residual stresses were found to be zero [10,13]. 2.2.3. Processing conditions The most important processing condition that aects residual stress formation is the cooling rate. The eects of this parameter will be explained later in this section. Few studies mention pressure [6,10,29] and processing environment [5,30] in conjunction with residual stress formation in thermoplastic composites. Pressure may aect the polymer thermal properties [31], which in turn inuence the formation of residual stresses. However, it was concluded that pressure levels in composites processing are relatively low [10], therefore limited eects can be expected on the matrix properties. The bres remain unaected by pressure. In addition, processing environment may prove to have a signicant eect on formation of residual stresses. Oxidation during heating may cause chemical changes to the polymer matrix or brematrix interface. This will inuence the thermal and mechanical properties of the polymer and hence, the residual stresses [5]. An extensive study by Augh et al. [30] showed that longer dwell times and higher temperatures during processing aect the glass transition temperature and hence the stress free temperature. This was found to be dierent for every processing atmosphere. 2.2.3.1. Cooling rate eects in amorphous matrix composites. As stated in the previous section, due to the pronounced visco-elastic behaviour of thermoplastic matrices [32], matrix relaxation accounts for signicant relief of residual stresses, especially at low cooling rates, since the polymer simply has more time to relax at high temperatures. The concept of free volume was formulated to explain the materials thermal expansion and shrinkage response [31,33,34]. This concept describes how the total volume in an amorphous polymer consists of the volume occupied by the polymer molecules and the free volume between the molecules. When the polymer is cooled at a high rate, more free volume is trapped or frozen-in below Tg, hence the total volume of the polymer will be higher after cooling down, see Fig. 5. However, the free volume theory is being disputed and new theories are proposed in terms of excess congurational entropy [33]. Nonetheless, graphs similar to Fig. 5 can still be used for illustrative purposes. As can be seen in this gure, the glass transition temperature is lower with lower cooling rates (CR), leaving a

Fig. 4. Specic volume of polysulfone (PSU, amorphous) and polyethyleneterephthalate (PET, semi-crystalline) as a function of temperature at atmospheric pressure [11].

sition temperature, Tm is the melting temperature, and Tc is the crystallisation temperature on slow cooling from the melt [11]. The portion of the curve that includes Tc is a representative plot during cooling. For the temperature range experienced by the thermoplastic matrix, the expansion and shrinkage behaviour is not linear, i.e. the CTE is subject to change at dierent temperatures [20,21]. 2.2.2. Temperature dierence Upon cooling amorphous thermoplastic composites between the processing temperature and the glass transition temperature (Tg), the amorphous matrix is in the visco-elastic state. This means that possible stress buildup due to thermal shrinkage can still be relaxed, because the molecular chains possess enough energy for freedom of movement at high temperatures, especially at low cooling rates. When the Tg is reached, the amorphous polymer becomes glassy (thermo-elastic state) and residual stresses are formed and frozen in upon subsequent cooling of the composite. Therefore, for amorphous matrix composites such as polyetherimide (PEI), thermoplastic polyimide (TPI), polyethersulfone (PES), and polysulfone (PSU) reinforced with carbon bres, the temperature at which thermal stresses start building-up (also dened as the stress-free temperature) was found to be around the glass transition temperature [13,2225]. For semi-crystalline matrices such as polyetheretherketone (PEEK), the stress-free temperature (SFT) can be found near the peak crystallisation temperature [10,26,27], because of the

1852

P.P. Parlevliet et al. / Composites: Part A 37 (2006) 18471857

CR1 > CR2

Total volume Volume

Free volume

CR1

CR2

Occupied volume
T g1 Tg 2

Temperature
Fig. 5. Eect of cooling rate on free volume and glass transition temperature in an amorphous polymer (CR = cooling rate), after [31]. The free volume is represented by the shaded area.

smaller temperature range at which stresses can build up [35]. It was calculated [23,35] that the cooling rate in the glass transition region mainly controls the residual stress formation and upon lower temperatures, the residual stress variation with temperature is mainly independent of cooling rate. Therefore, for amorphous polymer matrices, the faster the cooling rate in the glass transition region, the higher the residual stresses will be. Studies have shown that for amorphous matrix (CF/PEI) composites, the (radial) residual stresses indeed increase upon increasing cooling rate [23,35]. 2.2.3.2. Cooling rate eects in semi-crystalline polymer matrices. For semi-crystalline thermoplastics, a higher cooling rate will give a lower peak crystallisation temperature and lower crystallinity levels [31,36,37]. This leads to a lower stress-free temperature and less shrinkage [14]. Therefore, it was concluded that residual stresses due to crystallisation could be reduced by increasing the cooling rate [36,38]. However, when taking the relaxation behaviour of the (increased) amorphous content into account, two competing mechanisms will be the result: higher residual stresses due to fast cooling of the amorphous phases, and lower residual stresses because of lower stress-buildup temperature and less crystallisation shrinkage. In this respect, it was stated [39] that a relation needs to be established between visco-elastic behaviour and crystallinity kinetics in order to fully understand the development of residual stresses. In addition, low crystallinity content is often unwanted, due to poor solvent resistance [2,38]. The crystallisation kinetics are dierent for every semicrystalline polymer. Some semi-crystalline polymers crystallise so readily upon cooling, that for every practical cooling rate a constant level in crystallinity will be found (polyethylene), whereas other polymers can even be amor-

phous when quenched or cooled rapidly enough [40]. For PEEK/CF composites subjected to low cooling rates, it was shown that the crystallisation shrinkage is fairly constant [41] and thus only had a small eect on the total residual strain [35,38,4144]. In eect, it seems that there is a limiting cooling rate for PEEK composites, below which no signicant change in residual stresses with dierent cooling rates can be found [43]. When exposed to higher cooling rates, the residual stresses were found to decrease with increasing cooling rate [4244], although most practical cooling rates lie below this level (600 C/min). In carbon bre/polypropylene (CF/PP) microcomposites higher compressive bre strains were found for slower cooling rates [39]. At high temperatures and slow cooling rates, relaxation of the amorphous phases was observed to take place, which was counteracted by the high crystallisation rate causing the residual strains to be higher. In other studies [36,37] it was observed that in glass bre reinforced polypropylene (GF/PP) a higher level of residual stresses was present for higher cooling rates. For this material system, the relaxation rate of the matrix amorphous phases prevailed over the crystallisation rate upon slow cooling. The dierence in observation may be explained by the dierence in reinforcing bres and the eect of these bres on the crystallisation rate (trans-crystallinity). This eect will be discussed in Section 2.3. 2.2.4. Elastic properties One physical polymer property that inuences the formation of residual stresses in composites, is the matrix Youngs modulus [5]. The modulus is temperature-dependent and for semi-crystalline materials, cooling rate dependent. A higher cooling rate generates a lower level of highly elastic crystals, hence the resulting Youngs modulus will be lower [31,43]. Barnes and Byerly [13] ranked their calculated stress levels for dierent composite material systems and found a certain order, linked to the dierence in transverse stiness of the laminates. This transverse stiness depends on the matrix modulus among other things. In this respect, it may be said that if the matrix Youngs modulus is higher, the residual stresses will increase [45]. 2.3. Fibrematrix interface As was shown in Section 1.2, the signicant dierence in shrinkage behaviour between the bre and the matrix will result in radial as well as longitudinal residual compressive strains in the bre (Fig. 1). The anisotropic aramid and carbon bres will sustain higher residual strains in the longitudinal bre direction [5]. The residual strains in the radial bre direction oer a signicant fraction of the bre matrix interfacial (shear) strength for thermoplastic composites, since it provides one of the mechanisms whereby the matrix grips the bres and allows stress transfer via the brematrix interface [17,46]. Since the bres are loaded in compression along their length due to thermal residual strains, the tendency for

P.P. Parlevliet et al. / Composites: Part A 37 (2006) 18471857

1853

bre buckling increases and interfacial shear stresses are produced. This may lead to interface debonding and eventually microcracking [8]. A strong bond between the bre and the matrix is necessary to prevent this from occurring. Treatments to improve the brematrix bond strength are based on chemical bonding (on a molecular level), with e.g. a bre coating (sizing), and mechanical interlocking [46], where the bre surface may be treated to become rougher. Arguments for the latter treatment are based on the idea that stress transfer is governed by friction, and a linear relation can be assumed between the radial stress and the friction shear stress at the interface [35]. These treatments are synergistically enhanced by the residual radial stresses. A higher brematrix interfacial bond strength by chemical treatment will increase the residual stresses [35,37], because the residual stresses can be little relieved by interfacial debonding. A weak chemical interface would allow for brematrix slippage in the longitudinal bre direction relieving partially the stress build-up. This reduces the residual bre strains [37], preventing the bre from buckling and interfacial debonding. However, this is not very desirable, as in the end this will yield a composite with lower mechanical properties [37,47]. Nonetheless, experiments with AS4/PEEK composites showed [21] that if no chemical type of brematrix interfacial bonding was present, the shrinkage stresses in AS4/PEEK are suciently large to overcome the interfacial shear stresses. This shows that for every brematrix system an optimum in brematrix adhesion strength can be achieved in relation to residual stresses. Another eect regarding the brematrix interface is that in semi-crystalline composites, the bre surface may act as nucleation points for the crystalline spherulites. This can result in epitaxial spherulite growth from the bre surface, which results in a trans-crystalline region around the bre; see Fig. 6. This region was shown to cause additional radial residual stresses [43,48]. A thicker trans-crystalline layer was found to increase the stress transfer eciency between the bres and the matrix, probably due to increased radial compressive stresses owing to the anisotropic nature of the interlayer [48]. However, it was believed by others [49] that the trans-crystalline interface relieved the thermal residual stresses due to preferred orientation of the crystallites with respect to the bre. This was based on ndings of improved static strength and fatigue properties when a trans-crystalline interface was present. It seems that more research is necessary in order to understand the exact mechanisms behind the trans-crystalline layer and its eects on residual stress formation. In accordance with the processing eects, the interfacial shear stress was higher for slow cooled semi-crystalline matrix composites owing to additional crystallisation shrinkage of the matrix and a thicker trans-crystalline region around the bre. For amorphous thermoplastics, the interfacial shear stress increased for increasing cooling rate or lower end temperatures [35].

Fig. 6. Microstructure of polypropylene sample isothermally crystallised around a carbon bre followed by quenching [48].

3. Macro-mechanical residual stress formation Most of the factors described in Section 2 are also responsible for the macro-mechanical residual stress formation in angle-ply composites, which mechanism was explained in Section 1.2. Parameters that additionally inuence the magnitude of residual stresses in TPCs during lamination are shortly listed below. These include the laminate lay-up and morphology of the bres; such as unidirectional prepregs, fabrics, bre volume fraction, etc. [5,8,17].  For all cases, the bre residual strains were found to be compressive and higher for the more-ply composites [39,5052].  As applicable to all levels of residual thermal stresses, the lamination residual stresses also increase as the end temperature is further away from the stress-free temperature [10,11,13,22,23,26,28,43,53].  Regarding the bre volume fraction, a few conclusions were drawn in literature: The residual bre stresses in PP/CF micro- and macro-composites were modelled [54] and it was found that for higher bre volume fractions, the bre residual strains were lower. For some composites, an optimum in bre volume may exist at which the tensile stresses in the matrix in between the bres turn to compressive [8,55]. For cross-ply composites it was stated that the residual interlaminar stresses increase with increasing bre volume fraction [56].  A micromechanical analysis was performed into the eects of bre waviness [57], and it was found that in general, the residual stresses can be aected by bre waviness and hence the failure behaviour. Increased

1854

P.P. Parlevliet et al. / Composites: Part A 37 (2006) 18471857

bre waviness lowers the dierence in CTE in the transverse and longitudinal direction, consequently for unidirectional as well as angle-ply laminates the residual stresses are lower. The bre waviness additionally induces longitudinal shear stresses.  It was found that during consolidation, residual strains accumulate at dierent rates for each lay-up [58]. In addition, the optimum processing cycle, giving the lowest residual stresses and maximising the mechanical properties of the composite, was found to be dierent for each lay-up [16] or varying thickness ratios of 90 0 plies in PEEK/CF [26] and PEI/CF [53] cross-ply laminates.  It was shown for cross-ply polyphenylenesulde/carbon bre (PPS/CF) laminates that a minimum processing temperature and pressure are necessary in order to prevent debonding of the plies due to interlaminar residual stresses [59,60].  At bre angles of x  30, interlaminar shear stresses are maximum [61,62]. Studies on thermoplastic composites and the eects of their bre morphologies on the formation of residual stresses are not numerous. However, with respect to lamination residual stresses, the parameters governing residual stress build-up are little dierent between thermoset and thermoplastic matrix composites. Currently, more frequently fabric reinforced thermoplastic composites are applied [1], although few studies exist that compare the dierences in residual stress formation in unidirectional prepreg laminates and fabric reinforced laminates. 4. Global (skin-core) residual stresses As explained in Section 1.2, cooling rate plays a signicant role in the formation of global residual stresses. High performance thermoplastic composites are inherently manufactured at high temperatures. Therefore, the complex and high cooling rates within thick laminates are characteristically unavoidable [15]. As explained in Section 2.2.2, most of the stresses induced initially relax when temperatures are still high, whereas the stresses induced in the later stages of cooling remain when the matrix is partly crystallised or glassy and thus locks the material. Result is that the process leaves the surface regions of a plate in a state of permanent compression balanced by interior tension (thermal skin-core eects) [14], see Fig. 3. It was shown that with higher cooling rates, the residual stresses are higher and the distribution is more signicant [9,14,15,38,6365]. Most of the fast cooled or quenched composites such as PEEK/CF and PET/CF, show almost a parabolic residual stress prole [9,15,16,38,63,64,66,67], with the surface plies under compression of a higher (negative) magnitude than the tensile residual stresses in the centre plies. This residual stress gradient was found to be of a much higher magnitude in cross-ply laminates as in unidirectional laminates [15,16,63,64].

Crystallinity of semi-crystalline thermoplastic matrices plays a signicant role in skin-core eects, when a non-uniform cooling rate through the thickness of the laminate exists. Therefore, many studies regarding residual stress distributions on this level focus on crystallinity distributions immediately after processing and subsequent annealing of the composite (morphological skin-core stresses) [15,63,68]. Annealing is typically executed on thick laminates to reduce the skin-core residual stresses through relaxation of the stresses. The eects of annealing on residual stresses will be considered later in this section. In some studies on semi-crystalline matrix composites, even though no gradient in crystallinity level was present after cooling, still thermal skin-core residual stresses exist, due to the temperature gradients during processing [15,38]. No skincore residual stresses were measured in 8- and 20-ply composites, because of the low thickness [38], nor for very low cooling rates in 40-ply laminates, such as 0.2 C/min [38,63]. It was shown that fast cooling rates can lead to signicant temperature and crystallinity gradients in the vicinity of free edges and hence signicant localised residual stress distributions [65]. 4.1. Annealing In isotropic materials, residual stresses may be removed by annealing since their elastic expansion behaviour with temperature is uniform [23]. However, for anisotropic composites this is not so straightforward. In order to relax a portion of the residual stresses, to lower the residual stress gradients through the thickness of the laminate, or simply to obtain optimal levels of crystallinity, annealing of the composite can be carried out, although this requires additional time and sources [69]. Annealing can be accomplished by raising the composites temperature above the glass transition temperature of the matrix and allowing relaxation processes to take place. In consequence, the stress-free temperature may eectively be altered by annealing [13,29]. For amorphous systems on a micromechanical ply scale, annealing does not signicantly relax the residual stresses [10]. Moreover, on a macro-mechanical level it was concluded that in multi-axial laminates, the ability for selfrelieving of residual stresses is very limited within timescales representative for processing time scales. The amount of relaxation itself is not very signicant and even if signicant stress relaxation could be achieved at these high temperatures, during cooling to the service temperature the stresses reappear. Therefore, most annealing studies were performed to lower the residual stresses on the global level, e.g. the residual stress gradient through the thickness [9,13,38,43,70]. Increasing the annealing time for semi-crystalline thermoplastics below the glass transition temperature (Tg), decreases the residual stresses by means of relaxation [38]. The crystallinity level will be similar before and after annealing, since annealing is carried out below Tg. For

P.P. Parlevliet et al. / Composites: Part A 37 (2006) 18471857

1855

annealing temperatures above the glass transition temperature and within the crystallisation range, annealing will increase the crystallinity level of (quenched) specimens (accompanied by an increase in matrix modulus) and hence the degree of residual stresses due to increased (crystallisation) shrinkage [29,43,70]. Mechanical properties and solvent resistance can be improved after annealing [43], although it was believed that relaxation during annealing was the main contributor to increased mechanical properties and not the increased crystallinity level [44]. Quenching of the semi-crystalline composite and subsequent annealing were shown to give similar results as slow cooling [9]. 4.2. Inuence of processing mould material Another processing parameter that inuences residual stress formation and which was investigated by few researchers, is the use of dierent mould materials during processing, such as a rubber mould on one half and a metal mould on the other as in rubber press forming. Two types of interaction between the tools and the composite part can be distinguished: thermal and mechanical, and both can inuence the development of stresses in the composite. The thermal interaction was found to have a signicant inuence on the formation and distribution of global residual stresses (thermal skin-core residual stresses) due to differences in cooling rate on both laminate surfaces [68]. This dierence originates from the variation in heat transfer properties of the mould materials and hence the cooling rates [71]. The residual stress proles for unbalanced cooling were modelled [64,72], by utilising two mould halves under dierent cooling conditions, which led to much higher residual stresses in the unidirectional laminate, when compared to balanced cooling. The mechanical interaction comprises two ways: the rst is the mismatch in CTE between the tool and the composite, and often the CTE of the tooling is higher than that of the composite [71]. Upon cooling, this may induce compressive residual stresses in the surface plies of the composite part at the tool interface [71], which can result in a through-the-thickness stress distribution in the composite part, depending on the interface conditions between the tool and the composite [72]. Ultimately, the CTE mismatch can lead to warpage of the laminate [72] or bre waviness [71]. Interfacial stress transfer is determined by the friction behaviour between the tool and the composite material. It is conceivable that friction of the mould with the polymer matrix in its solidifying state [73], induces additional residual stresses in the surface ply, due to forced shrinkage. In some processes such as in rubber forming, a deformable tool is used in combination with a rigid tool, which comprises the second mechanical interaction. The deformation of the deformable tool may be imposed on the composite part, which eciency depends on the friction behaviour in the contact interface. The composite part hence experiences dierent loading on both surfaces, which may result in global residual stresses [72].

5. Discussion and conclusions In order to increase the applicability of continuous bre reinforced thermoplastics, it was argued already two decades ago that more knowledge regarding residual stresses and their eects needed to be generated. Up to now, still knowledge gaps exist, especially when it comes to studies on structural composites with other high-performance polymer matrices than polyetheretherketone (PEEK). For other semi-crystalline matrices than PEEK, no such extensive research has been performed on the crystallinity kinetics in composites and their eects on residual stress formation. Some studies investigated other high performance thermoplastic matrices such as polyethersulde (PES), polyphenylenesulde (PPS) and polysulfone (PSU) reinforced with carbon bres. Even fewer studies were published that concerned the residual stress formation in continuous bre reinforced engineering plastics, such as polypropylene (PP) [36,37,39] and polyethyleneterephthalate (PET) [11,38]. This can be explained by the fact that the processing temperatures of the high performance thermoplastic composites are, in general, very high and the magnitude of residual stresses will be accordingly. The eects will therefore be more signicant and perhaps more disastrous, hence the need to investigate these high-performance materials is more stringent. Few studies were published regarding residual stresses in fabric-reinforced thermoplastic composites. Since these materials will be applied more often in the future, it is of utmost importance to study residual stress formation in these materials and their dierences with composite laminates based on unidirectional plies. After identication of the most important parameters for residual stress formation with respect to the material properties, it may be concluded that the magnitude of the residual stresses is most dependent on the shrinkage behaviour of the thermoplastic matrix during processing, which varies for dierent processing conditions such as cooling rate, pressure and environment. Only few studies have been conducted on the eects of the latter two factors on residual stress formation. In order to predict the magnitude of residual stresses, the shrinkage coecients of the composite constituents must be known, the quality of brematrix adhesion must be taken into account, and the elastic moduli of the constituents must be identied as well as the bre morphology. The last few years, much eort is dedicated towards understanding the mechanisms of generation of additional residual stresses due to tool-part interaction in thermoset composites [7478]. However, few studies are available that investigate this eect on residual stress formation in thermoplastic composites, therefore it may be concluded that the exact mechanisms remain to be investigated for thermoplastic composites. As a last note, local residual stresses may arise during welding of thermoplastic composites. This may aect the

1856

P.P. Parlevliet et al. / Composites: Part A 37 (2006) 18471857 [21] Barnes JA, Simms IJ, Farrow GJ, Jackson D, Wostenholm G, Yates B. Thermal expansion characteristics of PEEK composites. J Mater Sci 1991;26(8):225971. [22] Cowley KD, Beaumont PWR. The measurement and prediction of residual stresses in carbon-bre/polymer composites. Compos Sci Technol 1997;57(11):144555. [23] Kim KS, Hahn HT, Croman RB. The eect of cooling rate on residual stresses in a thermoplastic composite. J Compos Technol Res 1989;11(2):4752. [24] Peeters LJB, Powell PC, Warnet L. Thermally-induced shapes of unsymmetric laminates. J Compos Mater 1996;30(5):60326. [25] DAmore A, Caprino G, Nicolais L, Marino G. Long-term behaviour of PEI and PEI-based composites subjected to physical aging. Compos Sci Technol 1999;59(13):19932003. [26] Jeronimidis G, Parkyn AT. Residual-stresses in carbon berthermoplastic matrix laminates. J Compos Mater 1988;22(5): 40115. [27] Youssef Y, Denault J. Residual stresses in continuous glass ber/ polypropylene composite thermoformed parts. In: Proceedings of international SAMPE symposium conference. p. 64150. [28] Papadopoulos DS, Bowles KJ. Use of unbalanced laminates as a screening method for microcracking. In: Proceedings of 35th international SAMPE symposium conference. Anaheim (CA), 25 April 1990. p. 15. [29] Chung DDL. Continuous carbon ber polymermatrix composites and their joints, studied by electrical measurements. Polym Compos 2001;22(2):25070. [30] Augh L, Gillespie JW, Finck BK. Degradation of continuous carbon ber reinforced polyetherimide composites during induction heating. J Thermoplast Compos Mater 2001;14(2):96115. [31] Young RJ, Lovell PA. Introduction to polymers. London: Chapman & Hall; 1991. [32] Brinson LC, Gates TS. Viscoelasticity and aging of polymer matrix composites. In: Talreja R, Manson JA, editors. Polymer matrix composites. Oxford: Elsevier Science Ltd.; 2001. p. 33368. [33] Jones R. The glass transition in polymers-bulk and thin lms. RPK BModule polymer physics, PTN, Eindhoven, The Netherlands, 2003. [34] Struik L. Physical ageing in amorphous polymers and other materials. Amsterdam: Elsevier Scientic Publishing Company; 1978. [35] DiLandro L, Pegoraro M. Evaluation of residual stresses and adhesion in polymer composites. Compos Part AAppl Sci Manufact 1996;27(9):84753. [36] Guillen JF, Cantwell WJ. The inuence of cooling rate on the fracture properties of a thermoplastic-based bre-metal laminate. J Reinforced Plast Compos 2002;21(8):74972. [37] Youssef Y, Denault J. Thermoformed glass ber reinforced polypropylene: microstructure, mechanical properties and residual stresses. Polym Compos 1998;19(3):3019. [38] Deshpande AP, Seferis JC. Processing characteristics in dierent semi-crystalline thermoplastic composites using process simulated laminate (PSL) methodology. J Thermoplast Compos Mater 1996;9(2):18398. [39] Nielsen AS, Pyrz R. The eect of cooling rate on thermal residual strains in carbon/polypropylene microcomposites. Sci Eng Compos Mater 1998;7(12):122. [40] Domb MM, Hansen JS. The eect of cooling rate on free-edge stress development in semi-crystalline thermoplastic laminates. J Compos Mater 1998;32(4):36186. [41] Lawrence WE, Seferis JC, Gillespie JW. Material response of a semicrystalline thermoplastic polymer and composite in relation to process cooling history. Polym Compos 1992;13(2):8696. [42] Wang C, Sun CT. Thermoelastic behavior of PEEK thermoplastic composite during cooling from forming temperatures. J Compos Mater 1997;31(22):223048. [43] Unger WJ, Hansen JS. The eect of cooling rate and annealing on residual stress development in graphite bre reinforced PEEK laminates. J Compos Mater 1993;29:10837.

mechanical behaviour of welded structures signicantly. To the best of the authors knowledge, this issue was not addressed nor raised before. Hence, future research on welding of thermoplastic composites should incorporate the factor of residual stresses. References
[1] van Barschot J. Airbus en Boeing vliegen met Ten Cate. In: NRC Handelsblad, Rotterdam, 5 January 2005. p. 17. [2] Cogswell FN. Thermoplastic aromatic polymer composites. A study of the structure, processing and properties of the carbon bre reinforced polyetheretherketone and related materials. Oxford: Butterworth-Heinemann; 1992. [3] Johnston NJ, Towell TW, Hergenrother PM. Physical and mechanical properties of high-performance thermoplastic polymers and their composites. In: Carlsson LA, editor. Thermoplastic composite materials. Amsterdam: Elsevier Science Publishers B.V.; 1991. p. 2771. [4] Callister WDJ. Materials science and engineeringan introduction. John Wiley & Sons, Inc.; 1994. [5] Favre JP. Residual thermal stresses in bre reinforced composite materialsa review. J Mech Behavior Mater 1988;1(14):3753. [6] Adams ME, Campbell GA, Cohen A. Thermal-stress induced damage in a thermoplastic matrix material for advanced composites. Polym Eng Sci 1991;31(18):133743. [7] Unger WJ, Hansen JS. Method to predict the eect of thermal residual stresses on the free-edge delamination behavior of bre reinforced composite laminates. J Compos Mater 1998;32(5):43159. [8] Kim J-K, Mai Y-W. Residual stresses. In: Engineered interfaces in ber reinforced composites. Oxford: Elsevier Science Ltd.; 1998. p. 30820. [9] Manson JAE, Seferis JC. Process simulated laminate (Psl)a methodology to internal-stress characterization in advanced composite-materials. J Compos Mater 1992;26(3):40531. [10] Nairn JA, Zoller P. Matrix solidication and the resulting residual thermal-stresses in composites. J Mater Sci 1985;20(1):35567. [11] Nairn JA, Zoller P. The development of residual thermal stresses in amorphous and semicrystalline thermoplastic matrix composites. In: Proceedings of toughened composites conference, Houston (TX), USA, 1315 March 1985. p. 32841. [12] Eijpe MPIM. A modied layer removal method for determination of residual stresses in polymeric composites. Enschede: Mechanical Engineering, University Twente; 1997. [13] Barnes JA, Byerly GE. The formation of residual-stresses in laminated thermoplastic composites. Compos Sci Technol 1994; 51(4):47994. [14] Trende A, Astrom BT, Nilsson G. Modelling of residual stresses in compression moulded glass-mat reinforced thermoplastics. Compos Part AAppl Sci Manufact 2000;31(11):124154. [15] Chapman TJ, Gillespie JW, Pipes RB, Manson JAE, Seferis JC. Prediction of process-induced residual-stresses in thermoplastic composites. J Compos Mater 1990;24(6):61643. [16] Li MC, Wu JJ, Loos AC, Morton J. A plane-strain nite element model for process-induced residual stresses in a graphite/PEEK composite. J Compos Mater 1997;31(3):21243. [17] Harris B. Residual strains. In: Harris B, editor. Engineering composite materials. London: IOM Communications; 1999. p. 7983. [18] Smith PA. Carbon ber reinforced plasticsproperties. In: Talreja R, Manson JA, editors. Polymer matrix composites. Oxford: Elsevier Science Ltd.; 2001. p. 10750. [19] Wardle MW. Aramid ber reinforced plasticsproperties. In: Talreja R, Manson JA, editors. Polymer matrix composites. Oxford: Elsevier Science Ltd.; 2001. p. 199229. [20] Barnes JA. Thermal-expansion behavior of thermoplastic composites 2. J Mater Sci 1993;28(18):497482.

P.P. Parlevliet et al. / Composites: Part A 37 (2006) 18471857 [44] Cantwell WJ, Davies P, Kausch HH. The eect of cooling rate of deformation and fracture in IM6/PEEK composites. Compos Struct 1990;14(2):15171. [45] White SR. Processing-induced residual stresses in composites. In: Dave R, Loos AC, editors. Processing of composites. Munich: Hanser Publishers; 2000. p. 23971. [46] Kim J-K, Mai Y-W. Engineered interfaces in ber reinforced composites. Oxford: Elsevier Science Ltd.; 1998. [47] Drzal LT, Madhukar MS. Fibermatrix adhesion and its relationship to composite mechanical properties. J Mater Sci 1993;28:569610. [48] Nielsen AS, Pyrz R. A Raman study into the eect of transcrystallisation on thermal stresses in embedded single bres. J Mater Sci 2003;38(3):597601. [49] Klein N, Marom G, Pegoretti A, Migliaresi C. Determining the role of interfacial transcrystallinity in composite materials by dynamic mechanical thermal analysis. Composites 1995;26(10):70712. [50] Filiou C, Galiotis C. In situ monitoring of the bre strain distribution in carbon-bre thermoplastic composites 1. Application of a tensile stress eld. Compos Sci Technol 1999;59(14):214961. [51] Galiotis C, Melanitis N, Batchelder DN, Robinson IM, Peacock JA. Residual strain mapping in carbon-ber peek composites. Composites 1988;19(4):3214. [52] Young RJ, Day RJ, Zakikhani M, Robinson IM. Fibre deformation and residual thermal stresses in carbon bre reinforced PEEK. Compos Sci Technol 1989;34(3):24358. [53] Warnet L. On the eect of residual stresses on the transverse cracking in cross-ply carbon-polyetherimide laminates. Enschede: Mechanical Engineering, University of Twente; 2000. [54] Wagner HD, Nairn JA. Residual thermal stresses in three concentric transversely isotropic cylinders: application to thermoplasticmatrix composites containing a trans crystalline interphase. Compos Sci Technol 1997;57(910):1289302. [55] Jones FR. Durability of reinforced plastics in liquid environments. In: Pritchard G, editor. Reinforced plastics durability. Cambridge, England: Woodhead Publishing Limited; 1999. p. 916. [56] Wang SK, Kowalik DP, Chung DDL. Self-sensing attained in carbon-ber-polymermatrix structural composites by using the interlaminar interface as a sensor. Smart Mater Struct 2004;13(3): 57092. [57] Karami G, Garnich A. Micromechanical study of thermoelastic behavior of composites with periodic ber waviness. Composites Part BEngrg 2005;36:2418. [58] Sorensen L, Gmu r T, Botsis J. Residual strain development in laminated thermoplastic composites measured using bre Bragg grating sensors. In: Proceedings of CompTest 2004 conference, Bristol, UK, 2123 September 2004. p. 1456. [59] Mei Z, Chung DDL. Eect of heating time below the melting temperature on polyphenylene sulde adhesive joint development. Int J Adhesion Adhesives 2000;20(4):2737. [60] Mei Z, Chung DDL. Thermal stress-induced thermoplastic composite debonding, studied by contact electrical resistance measurement. Int J Adhesion Adhesives 2000;20(2):1359. [61] Schwarz G, Krahn F, Hartwig G. Thermal-expansion of carbon-ber composites with thermoplastic matrices. Cryogenics 1991;31(4): 2447.

1857

[62] Zhang L. Time-dependent behaviour of polymers and unidirectional polymeric composites. Delft: Mechanical Engineering, Delft University of Technology; 1995. [63] Gillespie JW, Chapman TJ. The inuence of residual stresses on mode i interlaminar fracture of thermoplastic composites. J Thermoplast Compos Mater 1993;6(2):16074. [64] Sunderland P, Yu WJ, Manson JA. A thermoviscoelastic analysis of process-induced internal stresses in thermoplastic matrix composites. Polym Compos 2001;22(5):57992. [65] Domb MM, Hansen JS. Development of free-edge eect during processing of semicrystalline thermoplastic composites. AIAA J 1994;32(5):102933. [66] Hsiao SW, Kikuchi N. Numerical analysis and optimal design of composite thermoforming process. Comput Methods Appl Mech Eng 1999;177(12):134. [67] Ersoy N, Vardar O. Measurement of residual stresses in layered composites by compliance method. J Compos Mater 2000;34(7):57598. [68] Lawrence WE, Manson JAE, Seferis JC. Thermal and morphological skin core eects in processing of thermoplastic composites. Composites 1990;21(6):47580. [69] Kim BS, Bernet N, Sunderland P, Manson JA. Numerical analysis of the dimensional stability of thermoplastic composites using a thermoviscoelastic approach. J Compos Mater 2002;36(20):2389403. [70] Shih GC, Tseng WW, Lou AY. Evaluation of test methods in the determination of inplane shear modulus of poly(phenylene sulde) matrix composites. Polym Compos 1994;15(1):16. [71] Kugler D, Moon TJ. Identication of the most signicant processing parameters on the development of ber waviness in thin laminates. J Compos Mater 2002;36(12):145179. [72] Wijskamp S. Shape distortions in composites forming. Mechanical Engineering, Division of Design, Production and Manufacturing. Enschede, The Netherlands: University of Twente; 2005. [73] Personal communication of Sorensen L with Parlevliet PP. Inuence of mould friction on formation of residual stresses in thermoplastic composites. Delft, November 2004. [74] Wisnom MR, Gigliotti M, Ersoy N, Campbell M, Potter KD. Mechanisms generating residual stresses and distortion during manufacture of polymermatrix composites structures. CompositesPart A: Appl Sci Manufact; in press, doi:10.1016/ j.compositesa.2005.05.019. [75] Twigg G, Poursartip A, Fernlund G. Tool-part interaction in composites processing. Part I: experimental investigation and analytical model. Composites Part AAppl Sci Manufact 2004;35(1): 12133. [76] Twigg G, Poursartip A, Fernlund G. An experimental method for quantifying tool-part shear interaction during composites processing. Compos Sci Technol 2003;63(13):19852002. [77] Potter KD, Campbell M, Wisnom MR. Investigation of tool/part interaction eects in the manufacture of composite components. In: Proceedings of international conference on composite materials conference. San Diego, July 2003. [78] Ersoy N, Potter K, Wisnom MR, Clegg MJ. An experimental method to study the frictional processes during composites manufacturing. Composites Part AAppl Sci Manufact 2005;36(11):153644.