Analysis of crack propagation resistance of Al–Al2O3 particulate-reinforced composite friction stir welded butt joints

International Journal of Fatigue 31 (2009) 111–121

Contents lists available at ScienceDirect

International Journal of Fatigue journal homepage: www.elsevier.com/locate/ijfatigue

Analysis of crack propagation resistance of Al–Al2O3 particulate-reinforced composite friction stir welded butt joints A. Pirondi *, L. Collini Department of Industrial Engineering, University of Parma, Viale G.P. Usberti 181/A, 43100 Parma, Italy

a r t i c l e

i n f o

Article history: Received 26 November 2007 Received in revised form 8 April 2008 Accepted 12 May 2008 Available online 17 May 2008 Keywords: Particulate metal–matrix composite Friction stir welding Fatigue crack growth Crack closure

a b s t r a c t This work is devoted to the analysis of fatigue crack propagation resistance of particulate metal-matrix composites butt joints obtained by friction stir welding. Two different aluminum alloy matrices reinforced with alumina particles were examined. Tests were conducted on both parent material and welded joint for comparison. Fatigue crack propagation was carried out both within the weld nugget and in the thermo-mechanically altered zone at the side of the weld. The comparison between parent material and joint showed that the welding process affects fracture toughness and fatigue crack growth rate differently depending on the material. The analysis of crack path roughness helped to understand those differences in the fatigue crack growth rate. Therefore, roughness-induced crack closure arguments have been introduced to discuss data obtained under different testing conditions (parent material/joint, R-ratio, crack location, crack growth regime). Both the classical Elber’s approach and more recent approaches based on partial crack closure concept (adjusted compliance ratio, ACR, and 2/p methods) were considered. The results showed that, using partial crack closure, all of the data collapse within a reasonable scatterband. Ó 2008 Elsevier Ltd. All rights reserved.

1. Introduction The interest of automotive, railway, sporting goods, shipbuilding, aerospace and electronics, industries on metal matrix composites produced an expected increase of the market from its estimated 2005 level of 3.6 million kg to 4.9 million kg by 2010, with an average annual growth rate of 6.3% [1]. Particulate metal–matrix composites (PMMCs) are an attractive, alternative solution to short or long fiber-reinforced MMCs especially regarding their lower cost and quasi-isotropic mechanical response [2]. Commonly, Al-, Ti- or Mg-alloys are the matrix materials while high modulus ceramics, such as Al2O3 and SiC, are the reinforcement materials in form of particles. The main improvements given by PMMCs with respect to the matrix alone are higher stiffness, mechanical and wear resistance [2,3]. The very specific and high-tech usage of PMMC requires in principle the knowledge of standard mechanical properties such as tensile or fatigue strength, as well as fracture toughness and fatigue crack growth (FCG) properties to guarantee a reliable in-service durability. Unfortunately, due to their dual-phase nature, the strength of PMMCs is the result of several factors, such as type, size distribution and shape of particles and processing technique. The fracture toughness of a PMMC is generally lower than the respec* Corresponding author. Tel.: +39 0521 905885; fax. +39 0521 905705. E-mail address: [email protected] (A. Pirondi). 0142-1123/$ - see front matter Ó 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.ijfatigue.2008.05.003

tive matrix alloy due to the constraint on plastic deformation caused by particles. On the other hand, at low FCG rates, the PMMC shows a better performance than the unreinforced alloy due to particle-activated shielding mechanisms such as crack deflection or trapping [4–9]. Furthermore, in cast PMMCs the particles are often located at the grain boundary, leading to a crack tip shielding mechanism known as egg-shell [10]. A great influence on FCG properties of PMMCs comes from the R-ratio R = Kmin/Kmax, where Kmin and Kmax are the minimum and maximum values of the applied stress intensity factor during the fatigue cycle. As in unreinforced alloys, the higher the R-ratio the higher FCG rates. The crack closure concept introduced in the 1970s is conventionally adopted to assess the influence of R on FCG rate, also in the case of PMMCs [6], where crack surface roughness or crack bridging induced by particles typically play a prominent role. More recently, several questions have been posed about capabilities of the crack closure concept alone. Maximum stress intensity factor Kmax, residual stresses, cyclic plastic properties, environmental embrittlement of material ahead of crack tip, were found to have also a key role in explaining differences in FCG rates under varying loading sequence, microstructure and environmental conditions [11–17]. A major concern for the industrial application of PMMCs is the joining technology and the resulting strength of the bond. The underlying issue is that reactions between matrix and reinforcement may be promoted by the heating of the pieces typically used to fabricate the joint. Possible outcomes of this reaction are (i)

112

A. Pirondi, L. Collini / International Journal of Fatigue 31 (2009) 111–121

cracking at matrix-particle interface; (ii) partial dissolution of the reinforcement and (iii) precipitation of third-phases due to matrix reaction with the reinforcement material [18]. Joining PMMCs by fusion welding techniques such as laser beam welding (LBW) or MIG/TIG gives in general non-optimal microstructures, especially in the case of Al alloys reinforced with SiC particles [19–24]. Values of SE not higher than 70% were attained in [22,24] only with a careful choice of the filler material. The static efficiency of PMMC brazed joints is instead strongly dependent on particle size and content and may be even lower than 50%, as shown in [25]. For these motivations, solid-state processes like friction welding, friction stir welding (FSW) or diffusion bonding have in principle a lower detrimental impact on the strength of PMMC joints than liquid-state processes such as Inert Gas arc welding (TIG or MIG), laser beam and electron beam welding (LBW, EBW) [19–24]. Friction stir welding is a process recently developed and patented by The Welding Institute (TWI) of Cambridge (UK). In FSW, the parts to be joined are tied together while a rotating tool is pressed on them and moved along the seam, as illustrated in the scheme of Fig. 1 [26]. So far, this technique has been employed especially in the naval and aerospace industries thanks to the capability of joining aluminum alloys with higher quality, strength and comparatively low cost with respect to more traditional welding techniques. Microstructure, mechanical strength and their correlation with FSW process parameters have been extensively studied in the past few years in the case of light-weight (especially Albased) alloys [27–36]. Nowadays, the extension of FSW to other materials has become a research topic. As a solid-state joining process, FSW can be particularly effective in the case of materials sensitive to re-solidification after liquid-phase welding. This is indeed the case of particulate metal–matrix composites (PMMC), which suffer from poor weldability with traditional processes due to the presence of ceramic particles. So far only few attempts have been made to produce and characterize PMMC FSW joints [37–42], therefore leaving this field virtually unexplored. This study is devoted to the evaluation of fatigue crack propagation resistance of PMMC butt joints obtained by FSW. The materials considered are two aluminum alloy matrix/alumina particle: AA6061/Al2O3/20p and AA7005/Al2O3/10p. An extensive characterization of microstructure, tensile and fatigue properties of these two PMMCs was carried out in [39,40], comparing their properties with the base material. Static tensile tests have pointed out the high efficiency of the joint: ultimate and yield strength are 80% of the base material in the case of AA7005/Al2O3/10p, while FSW AA6061/Al2O3/20p butt joint exhibited an efficiency of 72%; at the same time the elongation to fracture and the hardening of the joints are higher than the base

Fig. 1. Outline of the FSW process [26].

material. The low-cycle fatigue life is instead slightly reduced with respect to the parent material for both PMMCs. The objective of this work is to present and analyse FCG experiments carried out both on PMMC parent material and on FSW butt joint, with two different values of R = Kmin/Kmax load ratio, where Kmin and Kmax are the minimum and the maximum Stress Intensity Factors at the crack tip, respectively, and two possible crack locations, i.e. within the weld nugget and in the thermo-mechanical affected zone (TMAZ) at the side of the weld. A crack path roughness analysis has been conducted at the SEM to assess the crack-particles interaction during propagation and the role of the load level on crack propagation mechanism. The discussion of different conditions (parent material/joint, R-ratio, crack location, crack growth regime) was done invoking simple crack closure arguments. Both the classical Elber’s approach and more recent approaches based on partial crack closure concept (adjusted compliance ratio, ACR, and 2/p methods) were considered.

2. Crack closure models In 1968 Elber discussed some of his observations indicating that crack closure due to interference of opposing surfaces may occur even during the tensile part of load cycles. This observation led to the definition of a new driving force for crack growth that would account for an opening load higher than the minimum load of the cycle:

DK eff ¼ K max  K open

ð1Þ

The underlying assumption is a rigid contact between crack surfaces and, therefore, for K < Kopen the crack tip is fully shielded. From the experimental point of view, Kopen is determined from the deviation in the linearity of a load vs. opening curve (see for example [43]). The anticipated contact of the crack surfaces is mainly related to the residual plastic deformation (plasticity-induced crack closure, PICC) in the steady-state (Paris) FCG regime, while at threshold closure is related mainly to microstructural asperities of the fracture surfaces (roughness-induced crack closure, RICC) or by oxide layers (oxide-induced crack closure, OICC) that may develop on the fracture surfaces. Anyway, the occurrence of closure due to such mechanisms leads to some criticism about the assumption of a rigid, complete contact of crack surfaces [44]:  the fatigue crack surface may not interfere at the very tip, but only at some distance behind that;  PICC can be hardly invoked under plane strain condition, because plasticity is more limited than under plane stress and therefore there is little material sticking off of the crack surface;  crack closure due to crack face interference can occur by asperities, oxide layers, etc. but such contributions to crack tip stresses are normally small and are important only in threshold region. If one considers instead a compliant crack wake [44] the load transfer between crack faces is progressive and therefore there is a local strain contribution even below Kopen. This means that the value of Kopen and, in turn, of DKeff cannot be simply determined at point of deviation from linearity of the load–displacementcurve. Alternative methods to evaluate DKeff were proposed in [44,45]. The adjusted compliance ratio (ACR) model [44] is based on the hypothesis of a crack driving force proportional to the strain magnitude. A correction is applied to the applied DK is made on the basis of the ratio of the measured strain range to ideally closure-free one, see Fig. 2. DKeff is then obtained as

113

A. Pirondi, L. Collini / International Journal of Fatigue 31 (2009) 111–121

3. Materials and experimental methodology

Fig. 2. Method for ACR closure correction evaluation [44].

DK eff ¼ DK  ACR ACR ¼

ð2Þ

Cs  Ci C0  Ci

ð2bisÞ

where Ci is the specimen compliance before crack initiation, and Cs and C0 are obtained from the load vs. displacement plot for a cracked specimen, see Fig. 2. The ACR parameter is independent from the measurement location. Kopen is then obtained as

K open ¼ K max  DK eff

ð3Þ

At very low FCG rates, it is known that the most important closure mechanisms are RICC and OICC, which may cause contact not immediately at the crack tip but someway behind that, Fig. 3a. According to this, a model was proposed in [45] that corresponds to the presence of a layer of thickness 2h inserted between crack faces at a given distance, d, from the crack tip. The evaluation of the crack driving force for this situation leads to

K max 

2

p

  2 2 K min < DK eff < K max  K open K open  1 

p

p

ð4Þ

It can be noticed that DKeff is independent from h and for low Rratios D Keff ffi Kmax  (2/p). Kopen. The two models, even though quite different in the formulation, rely on the same physical assumption, that is the crack does not always close completely.

Fig. 3. (a) Model of the partial crack closure mechanism; (b) parameter definition [45].

The materials under examination are a 6061 aluminum alloy reinforced with 20 vol.% of Al2O3 particles (W6A20A) and a 7005 aluminum alloy reinforced with 10 vol.% Al2O3 particles (W7A10A) produced by Duralcan (USA). They were supplied as extruded rectangular plates (100  7 mm2 cross-section) in the T6 condition. Table 1 shows the properties of unreinforced and reinforced metal matrices given by the supplier (Metalba, Italy). FSW butt joints were manufactured at the GKSS Research Centre, Geesthacht (D), using a Neos Tricept 805, CN controlled, fiveaxis robot. The FSW tool with a 20 mm diameter shoulder and 8 mm pin, was made of Ferro-Titanit, a highly wear resistant steel (an age-hardenable nickel martensite) reinforced with 30 vol.% TiC, having a service hardness of about 63 HRC. The parameters of the welding process were vertical force 12 kN; rotation speed 600 rpm; welding speed 300 mm/min. The microstructure was characterized by means of optical microscopy (OM) of polished samples, etched with a solution of 85 ml distilled water, 5 ml of HF and 10 ml of H2SO4. Compact tension specimens (CT) for parent material and extended compact tension specimens (ECT) for FSW joint were extracted from plates according to the scheme of Fig. 4. Positions W1 and W2 of Fig. 4 correspond to the weld-line and weld-side crack propagation tests, respectively. Residual stresses were evaluated along top and bottom sides of the ECT specimen by means of an X-ray diffractometer. The scheme of Fig. 5 shows the results of the measures. In the drawings of the ECT specimen, the distribution of longitudinal – Fig. 5a and Fig. 5b – and transverse – Fig. 5c and d – residual stresses on the top and on the bottom surface of the weld are illustrated, where the terms ‘‘longitudinal” and ‘‘transverse” are referred to direction parallel and perpenditively. Distribution of residual stresses significantly changes from

Table 1 Strength of the composites under test compared with the metal–matrices alone Material

E (MPa)

ry (MPa)

ru (MPa)

Elongation at failure (%)

AA6061 W6A20A AA7005 W7A10A

68,000 97,000 72,000 84,000

330 360 325 345

380 375 375 395

18 4 12 7

Fig. 4. Sketch of specimens extraction from welded plates (not in scale). Measures are given in mm.

114

A. Pirondi, L. Collini / International Journal of Fatigue 31 (2009) 111–121

Fig. 5. Outline of residual stress measurement locations by X-ray diffraction on W6A20 FSW joint; (a) longitudinal stress on weld top side, measured on the specimen axis; (b) longitudinal stress on weld bottom side, measured on the specimen axis; (c) transverse stress on weld top side, measured on the weld axis and at its lateral sides L and R; (d) transverse stress on weld bottom side, measured on the weld axis.

the top to the bottom side of the weld, i.e. through the specimen thickness. However, the value of stress goes from 64.7 ± 6.2 MPa (transverse residual stress rT on the weld axis, close to the specimen side) to 51.3 ± 3.0 MPa (longitudinal residual stress rL on the weld axis, close to the HAZ). These values are significantly lower than other results reported in the literature, such as in [33]. Such low residual stresses are probably the result of specimen extraction from the plates with consequent stress relaxation, further promoted by the small specimen size. For this reason, they have not been accounted for in the analysis at this stage. Fracture and FCG tests were then carried out on a MTS servohydraulic machine. FCG tests were performed at R = Kmin/Kmax equal to 0.1 and 0.5, respectively. The crack was placed along the weld line (WL), that is across the weld nugget, or at the weld side (WS) in the TMAZ (see Fig. 4). FCG tests were run in lab air at a frequency of 10 Hz, under constant load amplitude (DK-increasing) or with continuous load-shedding (DK-decreasing). The experiments were conducted in agreement with ASTM E647standard [43]. The experimental setup is outlined in Fig. 6. The crack propagation was monitored from the specimen compliance using a strain gage placed on the back face of the specimen. In this technique, a strain gage is glued to the back face of the specimen and a compliance C is evaluated by linear least squares fitting of (eW) vs. load (P) (e back-face strain). The crack length ratio a/W is calculated then by a polynomial relationship:

a ¼ a0 þ a1 u þ a2 u2 þ a3 u3 þ a4 u4 þ a5 u5 W

Metallographic and fractographic image analysis was made with a commercial image analysis software in order to study metal–matrix grain and particle size, distribution and crack path roughness. This latter is evaluated conventionally by superimpos;ing a grid on the image of the profile and counting the

ð5Þ

as a function of

u¼

1þ

1 pffiffiffiffiffiffiffiffiffi BEC

ð6Þ

The coefficients a0, . . . , a5 were calibrated by finite element analysis and are in good agreement with the literature regarding CT geometry [46].

Fig. 6. Experimental setup.

A. Pirondi, L. Collini / International Journal of Fatigue 31 (2009) 111–121

P

profile-grid intersections, Pi, along a given length, L. The method is more deeply described in [47]. The roughness Rv can then be expressed as

Rv ¼

yX Pi L

ð7Þ

This simple technique can obviously give only a mono-dimensional estimate of roughness. The analysis of crack opening data was carried out automatically by the test control software (790.40 FCG by MTS Systems Corp.) according to the compliance offset method described in [44].

115

4. Metallographic and fractographic investigation A cross-section of the W6A20A FSW joint is reported in Fig. 7a and in the detail of Fig. 7b. Looking at Fig. 7a, three different areas can be clearly identified: (i) a weld nugget with a typical ‘‘onion rings” structures generated by the spiral flow on the trailing side of the tool; (ii) a TMAZ, were the stirring of the base material begins; (iii) a zone not involved in the flow caused by the tool action but generally affected by heating (heat affected zone, HAZ). In unreinforced alloys, zone (i) is characterized by dynamic recrystallization of the metal-matrix grains, with reduction of the grain size, grain shape smoothing and fragmentation of precipitates [48,49]. On the other hand zone (ii) shows deformation of grains but they preserve the elongated shape of the extruded or laminated base material. In the (iii) zone it is possible to find a grain size increase due to heating, which corresponds generally to a minimum of hardness and tensile strength. In the case of W6A20A, FSW joints do not clearly show the same minimum of hardness in correspondence of the HAZ as in monolithic alloys [37,40] rather a generalized decrease all along the weld seam. In the case of W7A10A the situation is opposite, with a maximum in the TMAZ and similar values in HAZ and nugget [39]. The stirring of a PMMC causes also a redistribution and reshaping of the particles. Figs. 7b and 8 highlight the particles dispersion and fragmentation in the case of the W6A20A composite within: (a) parent PMMC and (b) FSW nugget. Two aspects can be highlighted, that are in agreement with the observations made in [39,40]:  the parent PMMC shows a homogenous dispersion of particles ranging from 10 to 20 lm; smooth (non-elongated) metal– matrix grains with sizes up to about 50 lm are detected;  particles are fragmented and clustered in the nugget, while metal–matrix grain size is much lower. A quantitative study of those modifications has been conducted by a commercial software of digital image analysis. The following microstructural parameter are considered: (i) area, perimeter, major axis and shape factor of the reinforcement particles, and (ii) grain size of the aluminum matrix. The major axis of a particle is defined as the maximum distance between two points inside its perimeter, while the shape factor, SF, has the following definition:

SF ¼ Fig. 7. (a) Micrograph of the W6A20A FSW joint cross section; (b) detail of the microstructure at the nugget border.

4p  A P2

ð8Þ

where perimeter P and area A of each single alumina particle were used. The shape factor varies between 0 in the case of a straight line

Fig. 8. Microstructure of W6A20A: (a) parent PMMC and (b) FSW nugget. Etched with 85 ml distilled water + 5%HF + 10%H2SO4 solution.

116

A. Pirondi, L. Collini / International Journal of Fatigue 31 (2009) 111–121

Table 2 Summary of microstructural parameters of PMMCs Average particle area (lm2)

Average particle major axis (lm)

Average particle shape factor

Average grain size (lm)

W6A20A Unwelded FSW Nugget

102.1 69.3

10.8 7.9

0.60 0.64

24.9 12.8

W7A10A Unwelded FSW Nugget

120.4 73.7

15.6 9.0

0.61 0.63

– –

Position

W7A10A composites. Especially in the case of W7A10A composite – that contains at 10 vol.% of reinforcement – an increase of the frequency in the lower classes, i.e. a reduction of particle area, is clearly visible, Fig. 9c. A higher uniformity in the particle dimension after the FSW process is found, because FSW in PMMC is an invasive process for the reinforcement particles, that crush or crumble, and change in shape and distribution. 5. Results and discussion 5.1. Fracture toughness

and 1 for a perfect circle, therefore it indicates the roundness of a particle, independently from its size. To start the measure of these parameters, the microstructural image obtained at the optical microscope is digitalized and the contrast in the gray scale emphasized. Each microstructural parameter is then automatically evaluated by the software, by the count of each pixel darker than a color threshold value imposed by the user. A set of measures for each image is obtained; the number of counted particle typically varies from 1400 to 2500 per image. The statistical average values of these parameters are summarized in Table 2. It is evident that the FSW process refines both grain and particle size (e.g. from 25 to 13 lm and from 11 to 8 lm respectively in the case of W6A20A), while the shape factor of the particles, is increased a little (from 0.60 to 0.64 always in the case of W6A20A). At the same time, the statistical analysis of particle size distribution shows a lower standard deviation of the values, as illustrated in the histograms of Fig. 9 for W6A20A and

The average value of two repetitions is shown in Fig. 10. In the case of W6A20A, fracture toughness tests were carried out only for the parent PMMC due to the limited amount of material, therefore KIC values were extrapolated from FCG tests. It is evident that in the case of W7A10A the fracture toughness of the FSW joint, is comparable or even higher than the corresponding PMMC. This is also the case of W6A20A if side-grooved PMMC specimens are compared with non-side-grooved FSW joints data from FCG tests. On the other hand, when comparing data extracted only from FCG tests, the FSW joint has only about 25% lower fracture toughness. The position of the crack within the joint has a sensible effect only in the case of W7A10A, where a lower value is found at the WS. A higher fracture toughness for the W7A10A FSW joints can be explained in terms of microstructure refinement (both grain size and particles) caused by the stirring action of the tool. Regarding W6A20A instead, the apparently lower fracture toughness finds a motivation comparing Fig. 8a and b: the FSW joint – Fig. 8b – shows finer metal–matrix grains, but also a finer particle dispersion, that means a shorter particle-to-particle path trough the

Fig. 9. Frequency distribution of (a, c) particle area and (b, d) particle shape factor in parent material and FSW nugget of W6A20A and W7A10A.

A. Pirondi, L. Collini / International Journal of Fatigue 31 (2009) 111–121

117

Fig. 10. Fracture toughness of the FSW joints compared with the base material.

metal matrix and a larger number of stress raisers. Therefore plastic deformation within the matrix is more constrained than in the parent material, see Fig. 8a, limiting the fracture toughness. This does not occur in the case of W7A10A, where the volume fraction of particle is half the one of W6A20A. 5.2. Fatigue crack growth Two typical crack paths recorded during FCG tests are illustrated in Fig. 11, for (a) W6A20A and (b) W7A10A (parent material). The first image depicts a typical branching mechanism, due to crack deviance caused by some big reinforce particle or cluster of particles; the second image is a clear example of RICC in which a premature contact of the crack surfaces is due to a alumina particle. These aspects allow a deeper analysis and elaboration of data points under the light of crack closure arguments, see Section 6. The results of W6A20A are reported in Fig. 12. The two crack positions, WL and WS, show a very similar FCG rate. Furthermore, they both fall within the 95% confidence limits of the base PMMC mean behavior. The FSW joint exhibit even a higher FCG strength at low crack growth rates (near-threshold regime), that can be attributed to a higher crack shielding effect related to the finer microstructure of the joint with respect to the parent material. The substantiation of the relationship between microstructural changes due to FSW and FCG rate has been given by the examina-

Fig. 12. FCG rate of W6A20A parent material and FSW joint.

tion of crack path roughness. The results summarized in Table 3 are elaborated from pictures of crack path taken at a given crack growth rate. The crack growth rate is similar for all of the conditions considered in Table 3 (parent or FSW PMMC), and belonging to near-threshold or Paris regime, respectively. The Rv found under Paris regime is higher with respect to the near-threshold value probably due to higher crack tip plastic deformation. Nevertheless, it does not produce proportionally higher crack closure effects since under Paris regime crack closure is less effective than at near-threshold. The difference in Rv between parent PMMC and FSW joint at low crack growth rates is attributed to the competition between two effects: (i) the grain refinement due to dynamic recrystallization and (ii) the particle fragmentation due to stirring. With small grains, Stage II (duplex slip) type of crack propagation in the metal–matrix is more likely to occur even at low crack growth rates. The planar with ripple aspect of Stage II FCG surfaces [50] would give a lower Rv with respect to the base PMMC but this is compensated by the higher number of crack deflections caused by fragmented particles, that increases Rv.

Fig. 11. Roughness induced crack closure in PMMC: (a) crack branching in W6A20A; (b) particle interaction in W7A10A.

118

A. Pirondi, L. Collini / International Journal of Fatigue 31 (2009) 111–121

Table 3 Crack surface roughness (Rv) at different regimes of fatigue crack propagation Material

Crack location

R

Rv (lm) (near-threshold regime)

Rv (lm) (Paris regime)

W6A20A

Base PMMC FSW (WL) FSW (WS)

0.1 0.1 0.1

0.22 0.41 0.50

0.51 0.48 0.64

W7A10A

Base PMMC FSW (WL)

0.1 0.1

0.86 0.23

1.21 0.51

However, the overall behavior is quite different from the case of W6A20A, since WL crack position gives the higher rates. The mechanism can be again attributed to the change in microstructure at WL. In this case, the metal–matrix grain size refinement effect described before prevails on particle fragmentation influence because of the low particle volume fraction. Therefore FSW joint exhibit lower Rv than base PMMC, as shown in Table 3. It is also worth to underline, that the results presented above show that fracture and FCG strength of the two composites is little affected by FSW. This is in perfect agreement with the results of [37,40] on the same materials (indeed, the same batch), where static joint efficiency SE = ru,FSW/ru,base  100 is in average 77% for W6A20A and 83% for W7A10A. Similar values of SE are shown also in [20,49]. 6. Crack closure analysis The objective of this analysis is to discuss experimental data at different R-ratios, material and crack location with the help of crack closure models. The trend of DKeff recorded according to the method of [43] is shown for the parent materials in Fig. 14 as a function of the Kmax applied during the test. The points in Fig. 14 include DK-increasing as well as DK-decreasing tests. The lines of equation DKeff = Kmax  (1  R) represent a closure-free propagation, i.e. where DKeff = DK. It is easy to draw the conclusion that DKeff is always less than DK, that is closure has developed. Besides, the difference between DK and DKeff:  is higher at low Kmax hence at a low crack growth rate;  is higher at R = 0.1 than at R = 0.5;  is higher in the case of W6A20A than in the case of W7A10A.

Fig. 13. FCG rate of W7A10A parent material and FSW joint.

Also in the case of W7A10A the two crack positions, WL and WS, show a very similar FCG rate (Fig. 13) and fall almost completely within the 95% confidence limits of the parent PMMC mean value.

The higher value of crack closure in a material with higher particle content with respect to a material with a lower particle content enforces the idea that roughness-induced crack closure is a prominent mechanism in this case. The result of the application of crack closure models to selected experiments is shown in Figs. 15 and 16 in the case of the parent materials. Concerning W6A20A (Fig. 15), all of the three models work quite efficiently; the best approximation is obtained globally with the 2/p method while the Elber’s method at low FCG rates probably overestimates the closure (data at R = 0.1 lying above

Fig. 14. Graphs of DKeff vs. Kmax (recorded according to the method of [43]): (a) W6A20A and (b) W7A10A FSW joint. The lines of equation DKeff = Kmax  (1  R) represent a closure-free propagation.

A. Pirondi, L. Collini / International Journal of Fatigue 31 (2009) 111–121

Fig. 15. FCG rate of W6A20A corrected for crack closure (parent material).

Fig. 16. FCG rate of W7A10A corrected for crack closure (parent material).

Fig. 17. FCG rate of parent materials and FSW joints corrected for crack closure.

119

120

A. Pirondi, L. Collini / International Journal of Fatigue 31 (2009) 111–121

R = 0.5). In the case of W7A10A (Fig. 16), the models differ more markedly and it is evident that the quantitatively and qualitatively better correction comes again from the 2/p method. The same analysis conducted on FSW joint led to the same qualitative result, that is the best agreement between data at different R-ratios is obtained, for both PMMCs, using 2/p partial crack closure evaluation method. The comparison of parent material and FSW joint in terms of DKeff has been therefore carried out using only 2/p partial crack closure evaluation method, as shown in Fig. 17. In the case of W7A10A, graph on the right, the data in the intermediate to low crack growth rate fall together irrespective of R-ratio and starter crack position. Also in the case of W6A20A the FCG data at the two different R-ratios and crack position come close to each other when plotted in terms of DKeff, although slightly different trends may be identified between R = 0.5 and R = 0.1. 7. Conclusions This work was devoted to the experimental evaluation and analysis of fracture toughness and FCG behavior of FSW butt joints of two PMMCs, namely 6061 aluminum alloy with 20 vol.% of Al2O3 (W6A20A) and 7005 aluminum alloy with 10 vol.% of Al2O3 (W7A10A). These properties were compared with the ones of the base PMMCs. Differences can be related to the interaction between the joining process and the microstructure. In particular:  the fracture toughness of the FSW joint is about 25% lower than the parent material in the case of W6A20A, while it is 10–20% higher in the case of W7A10A: this is because of the embrittlement effect due to the particle reinforcement;  the influence of FSW joining on FCG strength, in particular at near-threshold rates, is the opposite as for the fracture toughness, i.e. crack propagation rate is lower than in the parent material in the case of W6A20A, while it is higher in the case of W7A10A. The explanation of this behavior rate has been given by examining crack path roughness, Rv. The difference in Rvs between base PMMC and FSW joint at low crack growth rates is attributed to the competition between two effects: (i) the grain refinement due to dynamic re-crystallization and (ii) the particle fragmentation due to stirring. With finer grains, Stage II (duplex slip) type of crack propagation in the metal-matrix is more likely to occur even at low growth rates leading to a lower Rv. Only in the case of a high particle content, as in the case of W6A20A, this is compensated by the higher number of crack deflections caused by fragmented particles, that increases Rv. The discussion of experimental data with the help of different closure evaluation methods sorted the 2/p method as the one which gave the most efficient description of the differences among parent material and joint, loading conditions and crack location. Acknowledgements The authors wish to acknowledge Dr. L.M. Volpone, formerly at Fincantieri, Genoa (I), and Dr. J. Dos Santos, GKSS Research Centre, Geesthacht (D), for supplying the joints, Prof. R. Konecˇna, University of Zˇilina (SK), for the metallographic images, and Prof. M. Guagliano, Polytechnic of Milan, Milan (I), for residual stresses evaluation. References [1] http://www.bccresearch.com/editors/RGB-108N.html. [2] Curran G. Mater World 1998:20–1.

[3] Clyne TW, Withers PJ. An introduction to metal matrix composites. Cambridge, UK: Cambridge University Press; 1995. [4] Shang JK, Ritchie RO. On the particle size dependence of fatigue crack propagation threshold in SiC particulate reinforced aluminium-alloy composites: role of crack closure and crack trapping. Acta Metall 1989;37(8):2267–78. [5] Shang JK, Yu W, Ritchie RO. Role of SiC particles in fatigue crack growth of SiCparticulate-reinforced Al-alloy composite. Mater Sci Eng 1988;102(2): 181–92. [6] Lodgson WA, Liaw PK. Tensile, fracture toughness and fatigue crack growth rate properties of silicon carbide whisker and particulate reinforced aluminum metal matrix composites. Eng Fract Mech 1986;24(5):737–51. [7] Watt DF, Xu XQ, Lloyd DJ. Effects of particle morphology and spacing on the strain fields in a plastically deforming matrix. Acta Mater 1996;44(2):789–99. [8] Papakyriacou M, Mayer HR, Tschegg-Stanzl SE, Groschl M. Near-threshold fatigue crack growth in A12O3 particle reinforced 6061 aluminium alloy. Fatigue Fract Eng Mater Struct 1995;18(4):477–87. [9] Kumai S, King EJ, Knott JF. Short and long fatigue crack growth in a SiC reinforced aluminium alloy. Fatigue Fract Eng Mater Struct 1990;13(5): 511–24. [10] Wang Z. Fatigue of particulate ceramics reinforced metal matrix composites. Key Eng Mater 1995;104–107:765–90. [11] Fonte M, Romeiro F, Freitas M. Environment effects and surface roughness on fatigue crack growth at negative R-ratios. Int J Fatigue 2007;29:1971–7. [12] Dinda S, Kujawski D. Correlation and prediction of fatigue crack growth for different R-ratios using Kmax and DK* parameters. Eng Fract Mech 2004;71:1779–90. [13] Fonte MA, Stanzl-Tschegg SE, Holper B, Tschegg EK, Vasudevan AK. The microstructure and environment influence on fatigue crack growth in 7049 aluminum alloy at different load ratios. Int J Fatigue 2001;23:S311–7. [14] Da Fonte M, Romeiro F, de Freitas M, Stanzl-Tschegg SE, Tschegg EK, Vasudevan AK. The effect of microstructure and environment on fatigue crack growth in 7049 aluminium alloy at negative stress ratios. Int J Fatigue 2003;25:1209–16. [15] Silva FS. Crack closure inadequacy at negative stress ratios. Int J Fatigue 2004;26:241–52. [16] Silva FS. The importance of compressive stresses on fatigue crack propagation rate. Int J Fatigue 2005;27:1441–52. [17] Silva FS. Fatigue crack propagation after overloading and underloading at negative stress ratios. Int J Fatigue 2007;29:1757–71. [18] Schwartz MM. Composite materials, vol. 2. Prentice Hall; 1997. [19] Braun R, Dalle Donne C, Staniek G. Laser beam welding and friction stir welding of 6013-T6 aluminum alloy sheet. Mat.-wiss. u. Werkstofftechnik. 2000;31:1017–26. [20] Mishra RS, Ma ZY. Friction stir welding and processing. Mater Sci Eng R 2005;50:1–78. [21] Wang HM, Chen YL, Yu LG. In-situ weld-alloying/laser beam welding of SiCp/ 6061Al MMC. Mater Sci Eng A 2000;293(1-2):1–6. [22] Niu J, Pan L, Wang M, Fu C, Meng X. Research on laser welding of aluminum matrix composite SiCw/6061. Vacuum 2006;80(11-12):1396–9. [23] Bassani P, Capello E, Colombo D, Previtali B, Vedani M. Effect of process parameters on bead properties of A359/SiC MMCs welded by laser. Compos Part A: Appl Sci Manuf 2007;38(4):1089–98. [24] Lee JA, Carter RW, Ding J. Friction stir welding for aluminum metal matrix composites (MMC’s). NASA/TM1999-209876; 1999. [25] Zhang XP, Quan GF, Wei W. Preliminary investigation on joining performance o SiCp-reinforced aluminium metal–matrix composite (Al/SiCp–MMC) by vacuum brazing. Compos Part A: Appl Sci Manuf 1999;30(6):823–7. [26] http://imi.cnrc-nrc.gc.ca. [27] Thomas WM, Johnson KI, Wiesner CS. Friction stir welding – recent developments in tool and process technologies. Adv Eng Mater 2003;5(7):485–90. [28] Matrox SJ. Review of fatigue assessment procedures for welded aluminium structures. Int J Fatigue 2003;25:1359–78. [29] Dickerson TL, Przydatek J. Fatigue of friction stir welds in aluminium alloys that contain root flaws. Int J Fatigue 2003;25:1399–409. [30] Mir Zahedul H, Khandkar Kahn JA. Thermal modelling of overlap friction stir welding for Al-alloys. J Mater Process Manuf Sci 2001(10):91–105. [31] Shercliff PA, Thyoe HR. Development of the TrivexTM friction stir welding tool for making lap welds. In Proceedings of the fifth international symposium on friction stir welding, Metz, France, 14–16 September; 2004. [32] Cederqvist L, Reynolds AP. Factors affecting the properties of friction stir welded aluminum lap joints. Welding J (Research Supplement) 2001:281–7. [33] Bussu G, Irving PE. The role of residual stress and heat affected zone properties on fatigue crack propagation 2024-T351 aluminium alloys. Int J Fatigue 2003;25:77–88. [34] Ericsson M, Sandstrom S. Fatigue of FSW overlap joints in aluminium welded with different tool designs. In: Proceedings of the fifth international symposium on friction stir welding, Metz, France, 14–16 September; 2004. [35] Sutton MA, Reynolds AP, Yang B, Taylor R. Mode I fracture and microstructure for 2024-T3 friction stir welds. Mater Sci Eng A 2004;354(1-2):6–16. [36] Lomolino S, Tovo R, Dos Santos J. On the fatigue behaviour and design curves of friction stir butt-welded Al alloys. Int J Fatigue 2005;27(3):305–16. [37] Marzoli LM, Strombeck AV, Dos Santos JF, Gambaro C, Volpone LM. Friction stir welding of an AA6061/Al2O3/20p reinforced alloy. Compos Sci Technol 2006;66(2):363–71.

A. Pirondi, L. Collini / International Journal of Fatigue 31 (2009) 111–121 [38] Cavaliere P, Cerri E, Marzoli L, Dos Santos J. Friction stir welding of ceramic particle reinforced aluminium based metal matrix composites. Appl Compos Mater 2004;11(6):247–58. [39] Ceschini L, Boromei I, Minak G, Morri A, Tarterini F. Effect of friction stir welding on microstructure, tensile and fatigue properties of the AA7005/ 10 vol.%Al2O3p composite. Compos Sci Technol 2007;67(3-4):605–15. [40] Ceschini L, Boromei I, Minak G, Morri A, Tarterini F. Microstructure, tensile and fatigue properties of AA6061/20 vol.%Al2O3p friction stir welded joints. Compos Part A: Appl Sci Manuf 2007;38(4):1200–10. [41] Uzun H. Friction stir welding of SiC particulate reinforced AA2124 aluminium alloy matrix composite. Mater Design 2006;28(5):1440–6. [42] Wert JA. Microstructures of friction stir weld joints between an aluminiumbase metal matrix composite and a monolithic aluminium alloy. Scripta Mater 2003;49(6):607–12. [43] ASTM Standards. ASTM E 647, USA; 2008.

121

[44] Donald JK. Introducing the compliance ratio concept for determining effective stress intensity. Int J Fatigue 1997;19(1):S191–5. [45] Paris PC, Tada H, Donald JK. Service load fatigue damage – a historical perspective. Int J Fatigue 1999;21:S35–46. [46] Shaw WJD, Zhao W. J Test Eval 1994;22:512–6. [47] Wojnar L. 10 Lat rozwoju fraktografii ilosciowej. Inzˇyneria materialowa 1993;4:89–99. [48] Flores OV, Kennedy C, Murr LE, Brown D, Pappu S, Nowak BM, et al. Microstructural issues in a friction-stir-welded aluminum alloy. Scripta Mater 1998;38(5):703–8. [49] Liu G, Murr LE, Niou CS, McClure JC, Vega FR. Microstructural aspects of the friction-stir welding of 6061-T6 aluminum. Scripta Mater 1997;37(3):355–61. [50] Nicoletto G, Konecˇná R, Pirondi A. Fatigue crack paths in coarse-grained magnesium. Fatigue Fract Eng Mater Struct 2000;28(1-2):237–44.