Hybrid welding/joining of lightweight metals to carbon fiber reinforced polymers (CFRPs) typically relies on the adhesive bond created when the molten polymer matrix hardens in contact with the metallic surface. It is hypothesized that these bonds can be improved upon by fully displacing the polymer and infiltrating the carbon fibers with the metallic constituent to create load-bearing fibers that bridge the two materials. Friction stir welding (FSW) holds potential to melt and displace the polymer matrix, plasticize the metal constituent, and force the plasticized metal to flow around the fibers. Preliminary investigations were performed by FSW in AA 6061-T6 plates sandwiched against dry carbon fiber bundles. The FSW process plasticizes the aluminum while applying pressure, forcing the material to flow around the fibers. Cross-sectional images of the samples were used to measure the distance of infiltration of the aluminum into the carbon fiber bed. A fiber infiltration model previously developed to describe the infiltration of carbon fibers with epoxy resins during resin transfer molding was applied to this solid-state infiltration situation, thus modeling the plasticized aluminum as a fluid with an effective viscosity. Promising agreement was seen between the measured distances of infiltration and the predicted distances of infiltration when using effective viscosity values predicted by computational fluid dynamics (CFD) simulations of FSW found in literature. This work indicates that the well-established epoxy infiltration model can form the basis of a model to describe solid-state infiltration of carbon fibers with a plasticized metal.

## Introduction

Environmental and economic issues are driving an academic and industrial focus on the development and implementation of lightweight materials for transportation systems [1]. Carbon fiber reinforced polymers (CFRPs) are one class of materials that is gaining ever increasing interest for such applications due to their specific strength and specific stiffness. One challenge of integrating carbon fiber composites into transportation structures is joining them to adjacent metallic components, specifically lightweight metals such as aluminum and magnesium alloys. An overview of the current state of technology with respect to the joining of dissimilar materials is given by Martinsen et al. [2]. The conventional means of joining CFRP to metal include mechanical fasteners or adhesives. Both of these conventional means have potential drawbacks. In the case of mechanical fasteners, they add weight and create stress concentrations, and in the case of adhesives, they require extensive surface pretreatment and require long cure times. These limitations have driven an emerging area of research focusing on hybrid welding of the two dissimilar materials.

This work is a continuation of the work first presented by Franke et al. [3]. For that reason, the literature review contained in this paper is a condensed version of the aforementioned paper. For a more detailed description of previous research in hybrid welding of CFRPs and metals, refer to Ref. [2] or [3]. The main concept of hybrid welding of CFRP and metal is that a bond can be formed by melting the polymer matrix of the composite (must be thermoplastic) in contact with the metal constituent, then allowing the molten polymer to harden leading to adhesion and interlocking with the asperities on the metal surface. This method of bonding has been examined for the following processes:

• induction welding [4],

• laser welding [57],

• ultrasonic welding [811], and

• various forms of friction welding [1216].

The simplest way of joining a thermoplastic carbon fiber composite to a metal is to apply a heat source that will melt the polymer matrix allowing it to adhere to the metal surface. Mitschang et al. [4] used an induction coil to provide the heat required to melt the polymer matrix. The inductor was placed on top of the metal constituent, which in turn was placed on top of the CFRP in a lap joint configuration. Bergmann and Stambke [5], Schricker et al. [6], and Rodriguez-Vidal et al. [7] all investigated the use of a laser to melt polymers to metal surfaces. Much like with the induction welding method, the strength of the joint was directly related to the surface roughness of the metal surface. Joint strength was increased by adjusting the surface with methods such as acidic pickling and corundum blasting.

Solid-state welding processes that are typically used for joining metal to metal have also been applied to the metal constituent in a CFRP to metal joint in order to generate the heat needed to melt the polymer matrix. Investigations of ultrasonic spot welding have been performed by Balle et al. [811]. The studies examined the idea of applying a loaded ultrasonic sonotrode to the top of the metal constituent in a lap joint configuration with the CRFP. Greater shear strengths were realized when the welding parameters allowed for the polymer matrix to be displaced in the joining zone leading to direct contact between the carbon fibers and the metal constituent without breaking the carbon fibers. A similar conclusion was drawn by Goushegir et al. [12] when applying friction spot joining of metal to CFRP joints. Due to the plasticity induced within the metal constituent, the polymer matrix was displaced allowing for contact between the metal and carbon fibers, leading to increased strength. Andre et al. [13] used the same friction spot joining process except with an additional interlayer of polyphenylene sulfide between the aluminum and composite. The results showed that the interlayer increased the joining area resulting in stronger joints. Friction lap joining (FLJ) is a similar process to friction spot joining but with a translational aspect. Nagatsuka et al. [14] utilized the FLJ process to create a continuous joint between an aluminum alloy and a CFRP. Buffa et al. [15] used the same FLJ process but introduced holes into the metal constituent to allow molten polymer to flow into those holes and increase mechanical interlocking. Another variant of the friction stir (FS) process known as friction stir scribe technology was developed by Upadhyay et al. [16] to join dissimilar materials such as CRFP and metal.

The primary mechanism of load transfer between the metal alloy and fiber-reinforced polymer materials in most of the aforementioned studies is the adhesive bond formed when the polymer matrix melts and subsequently hardens in contact with the metallic surface. Studies in ultrasonic joining [811] and friction spot joining [12,13] have shown that contact between the metal constituent and the fibers has led to increased joint strength. It is hypothesized that strength can be increased by partial infiltration of the carbon fibers with the joining metal in order to form continuous load-bearing fibers that cross the metal–polymer interface, enabling fibers to contribute to the bond strength. Another way of viewing this is to apply the concept of friction stir forming [17] to achieve aluminum flow around carbon fibers. The goal of this study is to demonstrate aluminum infiltration of carbon fibers during friction stir lap welding. Models for predicting infiltration are discussed and compared to experimental results.

Friction stir welding (FSW) is chosen because it is known for its ability to join very dissimilar materials, has a relatively low joining temperature, and produces little or no thermal distortion in the welded structure [18,19]. FSW is a solid-state welding process in which (usually) metallic components are plastically deformed and mechanically intermixed under mechanical pressure at elevated temperatures. A nonconsumable rotating friction stir tool, consisting of a specifically designed probe (pin) and shoulder, is plunged into the workpiece and traversed along the defined weld path. Initially, the tool generates heat through friction, which facilitates plastic deformation of the workpiece material. Once plastic deformation occurs, heat is generated by both friction and heat dissipation due to plastic deformation. The plasticized material is mixed and extruded around the tool, where it is forged together in the wake of the tool. Significant research has shown that FSW can produce high-quality joints in aluminum alloys [1821]. Due to the significant differences in material properties between aluminum alloys and polymer composites, the intent of the present work is not to stir the two dissimilar materials together, but rather to contain the stirring action inside the aluminum alloy while in contact with the composite in a lap joint configuration. Ultimately, the intention is to melt and displace the polymer matrix and infiltrate the carbon fibers with the hot plasticized aluminum alloy. In this paper, the aluminum will be stirred into dry carbon fibers to examine the process of plasticizing aluminum and forcing it to flow around the fibers.

## Infiltration Model

The concept of using an infiltration model developed to describe the infiltration of carbon fibers with epoxy resins during resin transfer molding to model the infiltration of fiber with the plasticized solid-state aluminum was first proposed by Franke et al. in a conference paper [3]; the purpose of the current paper is to explain new measurement and analysis techniques that have been applied to the previous experimental work. These new techniques have led to a greater understanding of the interaction between FSW processed aluminum and carbon fibers. The infiltration model relies on well-studied models developed for infiltration of carbon fibers with polymer resins. The foundational relationship for modeling the infiltration of permeable media is known as Darcy's law [22]. The feasibility of using Darcy's law to describe the infiltration of carbon fibers has been examined extensively in the realm of resin transfer molding of carbon fiber/epoxy composites. Based on Darcy's law, Amico and Lekakou [23] have derived an expression for the infiltration distance of a fluid into an array of unidirectional fibers for application in resin transfer molding of composites. This equation takes the form of
$x2=2kμ(1−vf)(Pm+Pc)t$
(1)
where x is the distance of infiltration, k is the permeability constant of the fibers, μ is the dynamic viscosity of the fluid, Vf is the volume fraction of fibers of the final product, Pm is the mechanical pressure applied to the system, Pc is the capillary pressure between the fluid and the fibers, and t is the process time. The capillary pressure is dependent on the surface tension of the fluid and its wettability to the fibers. Amico and Lekakou [23] used the following equation to describe the capillary pressure experienced between a resin and fibers in transverse flow:
$Pc=2DfVf(1−Vf)σ cos θ$
(2)

where Df is the average diameter of the fibers, σ is the surface tension of the fluid, and θ is the wetting angle of the fluid on the fiber material. If a fluid is said to wet a material (Sessile drop test angle less than 90 deg), then the capillary pressure is positive and helps infiltration. If the two constituents behave in an unwetting manner (Sessile drop test angle greater than 90 deg), then the capillary pressure will be negative and oppose the infiltration of fibers.

In the present FSW investigation, solid-state aluminum alloy at approximately 85–95% of its solidus temperature and under high localized pressure is being forced to flow around carbon fibers in a manner similar to a fluid. Because the advancing material is not liquid, it is anticipated that the interaction between the carbon fiber and aluminum alloy cannot be properly captured by a capillary pressure term. In FSW, local melting can occur, but if present, will only occur at the tool interface and not at the bottom of the weld zone where the fibers reside. The wettability between a solid and a solid does not physically make sense nor can it be easily described. The closest possible value can be estimated using wettability values of molten aluminum on a carbon surface. The surface tension of a molten aluminum alloy similar in composition to AA6061 was taken to be 0.9 Pa m [24] and the wetting angle of molten aluminum on carbon was taken as 160 deg [25]. Using these values, the capillary pressure was found to oppose infiltration but is 2 orders of magnitude smaller than the applied mechanical pressure, hence insignificant to the values predicted in this study.

As shown in Eq. (1), infiltration is dependent on the permeability of the fibrous preform. The determination of the permeability of fibrous preforms is a vast and extensively examined subject matter. One of the most widely cited expressions for the permeability constant (k) of unidirectional fibers in transverse flow was derived by Gebart [26] and takes the following form:
$k=C(VmVf−1)52r2$
(3)

where r is the average radius of the fibers, C is a constant based on the packing geometry of the fibers, and Vm is the maximum volume fraction of fiber for that packing geometry. The constant C and value of Vm for quadratic and hexagonal packing geometries are shown in Table 1. This particular equation describes the flow into unidirectional fibers. The permeability value can be adjusted to describe more complex situations, such as infiltration of a carbon fiber weave (typical configuration in a CFRP).

Table 1

Constants in Gebart permeability equation dependent on packing geometry

PackingCVm
Quadratic$169π2$$π4$
Hexagonal$169π6$$π23$
PackingCVm
Quadratic$169π2$$π4$
Hexagonal$169π6$$π23$

Infiltration is also dependent on the viscosity of the fluid infiltrating the fibers (Eq. (1)). Minimal experimental work has been performed in terms of directly measuring the effective viscosity of metal during friction stir welding. However, many studies performed by the FSW community attempt to model the flow of metal during FSW using numerical fluid flow simulations. In these simulations, the effective viscosity is often estimated as a function of position around the tool based off of other known inputs such as the torque applied to the tool. Determining the effective viscosity is not the focus of such simulations, but is an intermediate step toward determining more pertinent information for FSW, such as the heating due to viscous dissipation. Although it is not the end goal, several numerical studies report ranges of estimated effective viscosities for plasticized aluminum alloy 6061 during FSW. Nandan et al. [27] numerically estimated the viscosity of plasticized AA6061-T6 during FSW based on a formulation of flow stress. The study estimated that the viscosity of the aluminum increases with both radial and axial distance away from the tool. The viscosity ranged from 105 Pa·s to 5 × 106 Pa·s for welding parameters of 637 rpm rotational speed and 95 mm/min traverse rate. Another numerical study performed by Crawford et al. [28] compared the results of simulations using a viscoplastic model (similar to Nandan et al.) and a Couette flow model. They concluded that the viscoplastic model correlated better with experimental results. The study examined a range of welding parameters (1000–5000 rpm and 290–1600 mm/min). Since determining the effective viscosity was not the focus of the paper, only the range of 104 Pa·s to 107 Pa·s was reported. In both of these studies, the term isoviscosity surface was used. This term describes the upper limit of viscosity at which significant stirring stops (edge of the stir zone). The isoviscosity limits of the two studies mentioned are 5 × 106 Pa·s and 107 Pa·s, respectively.

## Method

### Apparatus and Materials.

Friction stir welding was performed using a commercial three-axis computer numerically controlled mill (HAAS, Oxnard, CA, model TM-1). A constant travel angle of 3 deg was imposed by an angled clamping fixture. This fixture was mounted atop a three-axis piezoelectric force dynamometer (Kistler, Winterthur, Switzerland, model 9285), which measured the real-time welding forces. The dynamometer fed electrical charges to charge amplifiers, which in turn fed voltage signals that are linearized with respect to force to the data-acquisition system (labview by National Instruments, Austin, TX).

Aluminum plates of two distinct thicknesses were utilized. The first being AA6061-T6 plates that were 203 mm (8 in) long, 102 mm (4 in) wide, and 4.76 mm (3/16 in) thick. The second being AA6061-T6 plates of the same length and width but a thickness of 6.35 mm (1/4 in). Two FS tools made of H13 tool steel were used (one for each thickness of plate), both with a concave shoulder and conical probe with three flats [29]. The first tool, used in conjunction with the 4.76 mm thick workpieces, had a shoulder diameter of 15 mm, a probe that tapered from 7 mm in diameter to 5 mm in diameter at the tip, and a probe length of 4.1 mm. The second tool used in conjunction with the 6.35 mm thick workpieces had the same dimensions as the first tool except for a longer probe (4.7 mm). The carbon fiber used in this study consisted of continuous long fiber tows with 50,000 filaments per tow. The carbon fiber tows were provided by the SGL Group (Wiesbaden, Germany) and were manufactured with a polyamide compatible sizing.

### Experimental Procedure.

The fundamental process of forcing plasticized aluminum to flow around carbon fibers was studied by sandwiching a carbon fiber tow between the aluminum workpiece and the steel backing plate to create a bed of unidirectional fibers at the bottom surface of the aluminum. The 50,000 filament tow was spread to a consistent width of 20 mm at the backing plate with the fibers oriented in the welding travel direction. When the welding process is performed within the aluminum workpiece, the plasticized material is forced to flow down into the fiber bed. A depiction of the experimental setup is shown in Fig. 1. Prior to welding, the mill was zeroed in the axial direction by first firmly clamping the workpiece in place, placing a precision-ground gage block on top of the workpiece near the weld start position, bringing the trailing edge of the FS tool shoulder into contact with the gage block surface, and loading to a defined preload of 20 N. Shoulder plunge depths at the center of the tool were specified in reference to this zero. All the welds were performed at a weld length of 50 mm.

Fig. 1
Fig. 1
Close modal

The first set of experiments utilized aluminum workpieces of 4.76 mm thickness and the FS tool with a 4.1 mm long probe. The spindle speed, traverse rate, and commanded plunge depth varied according to the values shown in Table 2. The second, third, and fourth sets of trials utilized the 6.35 mm thick aluminum workpieces to allow for a larger working distance between the bottom of the probe and fiber contact surface during welding. The tool with a 4.7 mm probe length was used for all the samples of this thickness. The second set of trials utilized the welding parameters shown in Table 2. Welding in the third set of samples utilized a constant rotational speed of 1250 rpm and a constant traverse rate of 25 mm/min. Four samples were processed with increasing amounts of welding passes in an attempt to increase infiltration time. The four samples were processed with one, two, three, and four welding passes, respectively. The fourth set of samples utilized a constant rotational speed of 2500 rpm, a constant traverse rate of 50 mm/min (same advance per revolution (APR) as the third set), and also varied in the number of welding passes. The relatively high spindle speeds (greater heat generation) and low traverse rates (greater process time) were utilized in order to promote greater infiltration of the fibers. All the welding parameters are summarized in Table 2.

Table 2

Summary of test conditions for each set of welding trials

Trial setPlate thickness (mm)# of samplesSpindle speed (rpm)Traverse rate (mm/min)# of passesPlunge depth (mm)
14.7641500–180010010.2–0.6
26.35101200–30005010.6
36.3541250251,2,3,40.3
46.3532500501,2,30.3
Trial setPlate thickness (mm)# of samplesSpindle speed (rpm)Traverse rate (mm/min)# of passesPlunge depth (mm)
14.7641500–180010010.2–0.6
26.35101200–30005010.6
36.3541250251,2,3,40.3
46.3532500501,2,30.3

After welding, all the dry fibers that were not mechanically attached to the aluminum plate or infiltrated by the aluminum were removed by gently scrubbing the bottom surface with a cotton cloth. This action removed the majority of the 50,000 filaments of the tow leaving only the few layers of fibers that were embedded in the surface of the aluminum. Samples from sets 1 and 2 were examined optically on the bottom fiber contact surface. Samples from sets 3 and 4 were cross-sectioned, mounted, and polished for cross-sectional examination. All the cross sections were taken at a length of 40 mm into the weld length to ensure that the FSW process had reached an approximate steady-state. The approximate steady-state can be estimated from the measured process forces (flat trends in force). It was assumed that the infiltration process was consistent along the length of the weld as long as the process was at its approximate steady-state condition. Cross-sectional images from sets 3 and 4 that show infiltrated fibers at the interface were processed using an image analysis method developed in matlab (The MathWorks, Inc., Natick, MA) in order to measure the distance of infiltration from the images. Due to the uneven nature of the interface produced in the majority of samples, an image analysis method is superior to manually drawing lines to represent the distance of infiltration. The method consists of first converting the image to grayscale then segmenting the image into three bins: aluminum, carbon fiber, and mounting compound (see Fig. 3(a)). The threshold values were determined from the histogram of each image. A histogram of a typical image is shown in Fig. 2. The carbon fibers create a distinct left-skewed distribution in the center of the histogram. The threshold between the carbon fibers and the mounting compound was set as the lowest valley to the left of the carbon fiber distribution. The threshold between the carbon fibers and the aluminum was set as the right corner of the carbon fiber distribution.

Fig. 2
Fig. 2
Close modal
Fig. 3
Fig. 3
Close modal

Once the image was binned into the three distinct categories (aluminum, fiber, and mounting compound) based on the pixel intensity, the image was further cleaned setting a minimum pixel group size of 5 pixels (0.5 μm) in radius. Any pixel groups smaller than the cutoff grouping size are changed to the same category as the surrounding area. This minimum group size was intentionally chosen to be much smaller than the nominal fiber radius (7 μm). The program then finds the first nonaluminum pixel value from the top of each column and the first nonmounting-compound value from the bottom of each column and defines the distance between these two points as the infiltration distance for that column based on the number of column entries (number of pixels) and pixel size in micrometer for the image. The distances for each column were then averaged across the entire width of the image. Figure 3(b) shows a binned image with two examples of infiltration distances measured within the image. In the areas where there are no fibers between the aluminum and mounting compound (right set of arrows in Fig. 3(b)), the program measures the local infiltration distance as close to zero and averages those values into the final distance. This may artificially reduce the final distance measurement depending on one's definition of infiltration. Conversely, the full thickness of a fiber that is not fully enclosed in aluminum is counted in the measurement. This artificially increases the distance measurement. These two imperfect cases are present but counteract each other.

## Results and Discussion

### Effect of Probe Tip to Fiber Interface Distance.

The objective of this research is to determine if a correlation existed between the infiltration models used to describe the infiltration of carbon fibers in resin transfer molding and the infiltration of fibers with the plasticized aluminum produced during friction stir welding. In order to examine this correlation, an even and measurable distance of infiltration must be produced. It was determined that the fibers react differently to the presence of plasticized aluminum depending on the distance of the fibers from the bottom surface of the rotating probe of the FS tool (working distance). When the bottom of the probe (on the FS tool) was in too-close proximity to the fibers, the stirring forces of the plasticized aluminum break the carbon fibers into segments (Fig. 4). This phenomenon was seen in all the samples made with 4.76-mm thick aluminum plate, processed with the FS tool with a 4.1 mm long probe, and at plunge depths of 0.2–0.6 mm, which corresponded to working distances of 0.4–0.1 mm. At these distances, the fibers encounter the stir zone of the weld, where the forces destroy the full integrity of the fibers and stir the broken fibers up into the aluminum.

Fig. 4
Fig. 4
Close modal

Segmented fibers may possess the potential to produce load-bearing capabilities across the aluminum–polymer interface when processing a polymer carbon fiber composite. However, Balle et al. [8] encountered fragmented fibers when dealing with ultrasonic welding and concluded that conditions resulting in fractured fibers resulted in samples with lower tensile strengths than those where fibers were kept intact. Therefore, conditions were chosen such that the fibers remained intact during processing. This was achieved by creating a larger working distance between the FS tool's probe and the carbon fibers. Using a thicker aluminum plate (6.35 mm thick), it was experimentally determined that at a spindle speed of 2000 rpm and a traverse rate of 50 mm/min, the probe should be at least 1 mm away from the fibers to prevent fiber segmentation. This minimum distance was found to depend on the FS tool rotational speed: as the rotational speed increases, the volume of the stir zone increases due to increased energy input, thus the working distance (probe–fiber) must be increased to prevent segmentation. This trend held true up to approximately 2800 rpm where the volume of stirred material appeared to reach a maximum. This effect has been explained by Threadgill et al. [19] as a self-limiting aspect of heat generation during FSW. Just outside the stir zone, there lies the thermomechanically affected zone: a region where plasticized aluminum will infiltrate the carbon fibers, yet leave them intact. Carbon fibers processed in this region remained in the direction they were prepositioned (Fig. 5). Due to the absence of excessive stirring in this region, an even distance of infiltration can be observed in the cross section of such samples. This stirring limitation frames the effective viscosity term used in the model to predict the infiltration. The fibers are placed at the edge of the stir zone, which should result in effective viscosities similar to the isoviscosity surfaces estimated in the numerical fluid flow simulations (5 × 106 Pa·s to 107 Pa·s [26,27]).

Fig. 5
Fig. 5
Close modal

### Experimental Infiltration of Carbon Fibers.

Data were collected by conducting experiments consisting of two different sets of welding parameters at various effective infiltration times (trial sets 3 and 4 described in the “Experimental Procedure,” section Table 2). Trial set 3 samples were welded at 1250 rpm and 25 mm/min with one, two, three, and four passes, which resulted in infiltrations times of 12, 24, 36, and 48 s, respectively (Table 3). The approximate time of infiltration was estimated as the time the tool probe was present over a particular point in the sample calculated by dividing the probe diameter (5 mm) by the travel speed. For example, the effective infiltration time over a given location at a traverse rate of 25 mm/min (0.42 mm/s) is approximately 12 s for one pass at the center of the weld. When multiple welding passes are made, the infiltration time was multiplied by the number of passes. Trial set 4 was run at a different set of parameters with the same APR (2500 rpm and 50 mm/min) for one, two, and three total passes per sample (Table 3). Representative images showing the degree of infiltration after welding with the parameters listed in Tables 2 and 3 can be seen in Figs. 6 and 7, showing increasing infiltration distance with increasing process time.

Fig. 6
Fig. 6
Close modal
Fig. 7
Fig. 7
Close modal
Table 3

Estimated infiltration process times for each sample in trial sets 3 and 4

Passes25 mm/min (s)50 mm/min (s)
1126
22412
33618
448
Passes25 mm/min (s)50 mm/min (s)
1126
22412
33618
448

Figures 6 and 7 show that the infiltration distance increases with increasing number of welding passes (residence time). Fifteen images were taken at the fiber interface directly below the tool probe for each sample. Only images from directly below the probe (5 mm width) were used since it was observed that the infiltration distance begins to decrease rapidly outside of this zone. All the images were processed using the image analysis method described in the Experimental Procedure section to produce an average distance for each image and the 15 distances for each sample were then averaged to produce a single value. These mean values are reported in Table 4 along with the calculated standard deviation of the 15 measured infiltration distances. It can be seen from the table that both the mean and the standard deviation of the infiltration distance increase with the residence time.

Table 4

Measured distances of infiltration

Passes25 mm/min (μm)50 mm/min (μm)
14.3 ± 1.24.7 ± 1.4
27.4 ± 1.88.7 ± 2.2
38.1 ± 2.011.9 ± 2.8
412.3 ± 3.3
Passes25 mm/min (μm)50 mm/min (μm)
14.3 ± 1.24.7 ± 1.4
27.4 ± 1.88.7 ± 2.2
38.1 ± 2.011.9 ± 2.8
412.3 ± 3.3

### Discussion of Model Assumptions.

During the resin transfer molding process, the applied vacuum pressure constrains the fibers against each other prior to infiltration. Then during the infiltration process that pressure pulls the epoxy through the fibers separating them so that the gaps between the fibers are filled with the epoxy. It is assumed that essentially the same process will occur during the FSW solid-state infiltration. The pressure of the FSW process constrains the fibers against each other between the aluminum workpiece and backing plate, concurrently the immense pressure forces the plasticized material to flow into the gaps between the fibers separating them and enclosing them in aluminum as shown in Figs. 6 and 7. To the best of our knowledge, the process conditions during the FSW solid-state infiltration mirror the process conditions under which the resin transfer models were developed. In both cases, the fibers are assumed to be rigid and that the material infiltrating them is a fluid with an effective viscosity.

When calculating the distances of infiltration using the model, various assumptions need to be made. In terms of the applied mechanical pressure, there was no method readily available to directly measure the pressure at the fiber interface. Therefore, the pressure was estimated as the measured axial force during welding applied over the axial surface area of the workpiece in contact with the tool, i.e., the area of a circle defined by the diameter of the FS tool shoulder. While processing the samples, the measured axial force fluctuated around 4500 N during the steady-state portion of all the welds measured and reported in Table 4. The shoulder diameter of the FS tool was 15 mm corresponding to an area of 1.767 × 10−4 m2 over which this axial force is applied, resulting in approximately 25 MPa of pressure. In the future, this approximation could be improved upon by numerically simulating the local pressure under the FS tool probe at the backing plate.

The measurement of fiber volume fraction (Vf) of the final composite is important because the predicted infiltration distance (x) is dependent on this volume fraction directly as well as being dependent on a permeability constant (k) that is also dependent on volume fraction (Eq. (3)). The volume fraction was determined by assuming that the infiltrated fibers are continuous, and therefore, the cross-sectional area fraction of fibers is equal to the volume fraction. The cross-sectional area fraction of infiltrated fibers was approximated using a variation of the image analysis program developed for the measurement of infiltration. Small sections of images showing the infiltrated fibers were cropped as shown in Fig. 8(a). The images were converted to a binary image (fiber/nonfiber) as shown in Fig. 8(b), based on the histogram method previously described. From this image, the fraction of pixels that represent the fiber can be determined. Fifteen different sections of infiltrated areas at the two sets of welding parameters were processed and averaged resulting in a volume fraction of 0.51 with a standard deviation of 0.07. An effort was made to standardize the placement of the infiltrated media (bed of dry fibers) for each friction stir welding test so that it is valid to use the same value of volume fraction (0.51) in the model prediction of each condition.

Fig. 8
Fig. 8
Close modal

The permeability constant (k) used in Eq. (1) was determined using hexagonal fiber packing (Eq. (3)), as it is believed that the fibers will arrange in a close packing manner when constrained by the aluminum and steel plates and put under high load. The fibers are on average 7 μm in diameter, so a fiber radius (r) of 3.5 μm was used in Eq. (3). This value was measured directly from the cross-sectional images.

### Model Verification.

Using the values previously discussed, the distance of infiltration was modeled in matlab (The MathWorks, Inc.) using the infiltration model presented in Eq. (1). The three lines in Fig. 9 show the predicted infiltration distance as a function of process time for three values of effective viscosity that correspond to the isoviscosity surface (i.e., the edge of the stir zone) proposed in numerical simulations found in the literature [26,27]. Discrete experimentally measured values are plotted alongside the predicted curves in Fig. 9 and are further summarized in Tables 2 and 3. The distances of infiltration predicted by the model are on the same order of magnitude as distances seen in the experimentally processed samples. Furthermore, the experimental data appears to follow a trend similar to the model as time is increased. Samples processed at 1250 rpm and 25 mm/min appear to trend well with an effective viscosity closer to 107. This would suggest that the fibers lie just outside the stir zone. However, the samples processed at 2500 rpm and 50 mm/min trend with the model with an effective viscosity between 106 and 5 × 106. Even with the same APR, doubling the spindle speed leads to a greater energy input, which increases the distance the stir zone extends away from the FS tool. Since both sets of samples were performed with the same probe to fiber working distance, the increase in the stir zone size will cause the fiber interface to lie slightly inside the stir zone at the 2500 rpm condition. The numerical fluid flow simulations cited [26,27] would predict that the effective viscosity at the fiber interface would significantly decrease as the fibers are moved inside the shear zone due to the higher strain rates closer to the FS tool. One important limit to the increased infiltration observed by moving the fiber interface further inside the stir zone is that the fibers will begin breaking under the high local shear stress. Therefore, achieving the greatest amount of infiltration involves accurately determining the working distance and welding parameters that produce the lowest viscosity yet also leave the fibers intact. Ideally, this will be determined by developing a method of experimentally measuring the effective viscosity of the plasticized aluminum at different interface distances.

Fig. 9
Fig. 9
Close modal

Qualitatively, the model developed to describe the infiltration of fibers with polymer resins can be used to describe the infiltration of fibers with plasticized aluminum produced during FSW. In general, the value of this model is that it can be used as a means of predicting how changes in certain factors (pressure, viscosity, fiber radius, etc.) affect infiltration. Several things can be learned from this model: First, the capillary pressure (whether applicable or not) appears to have little effect on this system due to the overwhelming applied pressure. This would suggest that attempting to alter the wetting characteristics of the aluminum on the carbon fibers by coating the carbon fibers (which is effective in molten aluminum infiltration) may have a minimal effect on the infiltration by plasticized aluminum. Second, this model shows the substantial effect of the volume fraction of fibers in conjunction with time. Figure 10 shows the estimated infiltration at 0.4, 0.5, 0.6, and 0.7 volume fractions of fibers while keeping all other parameters constant. It is clear that the higher volume fraction quickly plateaus (increasing process time leads to diminishing returns), while the lower volume fraction experiences increasing infiltration with time. This could potentially be useful in predicting infiltrations into polymer composite systems where the volume fraction of the fibers of the polymer composite is known. However, it is possible that within a molten polymer matrix, the fibers will be compacted against each other resulting in similar packing structures regardless of the original volume fraction of the composite. Third, the model can be used to examine the effect of the size of the fibers on the infiltration process. The diameter of carbon fiber is normally 5–10 μm. However, other types of reinforcing fibers could vary in size. Figure 11 shows the effect of fiber radius when the radius varies between 5 and 15 μm. As can be seen, an increase in radius leads to an increase in infiltration by that same factor. It must be noted that increasing the size of the fibers means the distance the plasticized material must flow to incorporate a single fiber increases that same amount. This suggests that the number of fibers that will (in theory) add to the strength of the joint will remain the same, but will be larger in diameter. Assuming a larger diameter fiber can carry more load, increasing the size of the fiber has potential to increase the strength. Finally, the apparent correlation of this model suggests that the plasticized aluminum produced during FSW can be modeled as a fluid with an effective viscosity in simple cases of the plasticized material flowing into geometries within a secondary material. This corroborates the concept of using computational fluid dynamics simulations to model the material flow during FSW.

Fig. 10
Fig. 10
Close modal
Fig. 11
Fig. 11
Close modal

It is important to note that the experiments contained in this study were performed with dry fibers (not contained in a polymer matrix). In a polymer composite system, the process may behave differently due to the effects of the polymer matrix. Additional pressure may be needed to displace the molten polymer matrix from the fibers, making infiltration more difficult. However, the polymer matrix will create space between the fibers and it may be easier for the plasticized metal to flow into that space instead of the dry fibers compacted tightly against themselves as they were in this study. Additional studies are therefore needed to explore and validate the proposed model for predicting infiltration into more complex woven fiber geometries in the dry state and for prediction of the hybrid joining of aluminum to finished polymer matrix composites.

## Conclusions

An understanding of the interaction between the plasticized aluminum and dry carbon fibers was developed. Two modes exist, one in which fibers are segmented and stirred up into the aluminum, and another in which the fibers are kept intact and infiltrated by the aluminum. Distances of infiltration of up to approximately 15 μm (three layers of fibers) were achieved. The experimentally measured distances of infiltration appear to be consistent with the distances of infiltration predicted by infiltration models developed for polymer resins, showing that the plasticized aluminum can be modeled as a highly viscous fluid for this purpose. The viscosity values used for this prediction are also within the ranges of viscosities previously proposed in numerical FSW studies, further supporting this modeling approach and also providing a partial validation of such simulations.

The present work shows that the models applied to resin transfer molding can serve as the basis for a model to describe the infiltration of fibers with plasticized metals. This is important in terms of developing a fundamental understanding of this hybrid joining process as well as forming the basis of a predictive model that can be used to identify approaches for increasing the amount of infiltration and therefore bond strength. An experimental methodology is also under development for experimentally determining the effective viscosity near the edge of the stirred zone during FSW. In the present work, this value is taken from previously published numerical simulations, so this additional experimental advance will further help to bolster the present work and provide a standard method for determining this viscosity across various alloys and welding setups.

It has yet to be seen if the levels of infiltration seen in this study will have a significant effect on the joint strength when welding to a polymer composite. Additional work is therefore also needed to explore the effects of infiltration on the strength of joints. This will lead to the ability to relate the currently proposed infiltration prediction model to joint strength for use in determining the optimum welding parameters for hybrid joining.

## Acknowledgment

The authors gratefully acknowledge the Machine Tool Technology Research Foundation, Professor Natalie Rudolph, and colleagues in the Advanced Manufacturing Lab.

## Funding Data

• Division of Civil, Mechanical and Manufacturing Innovation, National Science Foundation (NSF) (Grant No. CMMI-1332738).

• The Department of Mechanical Engineering at the University of Wisconsin-Madison.

## Nomenclature

• APR =

•
• C =

constant dependent on packing geometry of fibers

•
• CFRP =

carbon fiber reinforced plastic

•
• Df =

diameter of fibers (m)

•
• FLJ =

friction lap joining

•
• FSW =

friction stir welding

•
• k =

permeability constant

•
• Pc =

capillary pressure (Pa)

•
• r =

•
• t =

time of process (s)

•
• Vf =

volume fraction of fibers

•
• Vm =

maximum volume fraction of fibers for given geometry

•
• x =

distance of infiltration (m)

•
• θ =

sessile drop test wetting angle (deg)

•
• μ =

viscosity (Pa·s)

•
• σ =

surface tension of fluid (Pa m)

## References

1.
U.S. DOE
,
2015
, “
,” U.S. Department of Energy, Washington, DC.
2.
Martinsen
,
K.
,
Hu
,
S. J.
, and
Carlson
,
B. E.
,
2015
, “
Joining of Dissimilar Materials
,”
CIRP Ann. Manuf. Technol.
,
64
(
2
), pp.
679
699
.
3.
Franke
,
D. J.
,
Morrow
,
J. D.
,
Duffie
,
N. A.
,
Zinn
,
M. R.
, and
Pfefferkorn
,
F. E.
,
2016
, “
Towards Improved Hybrid Joining of Aluminum Alloys to Carbon Fiber Composites With Friction Stir Welding
,”
ASME
Paper No. MSEC2016-8747.
4.
Mitschang
,
P.
,
Velthuis
,
R.
, and
Didi
,
M.
,
2013
, “
Induction Spot Welding of Metal/CFRPC Hybrid Joints
,”
,
15
(
9
), pp.
804
813
.
5.
Bergmann
,
J. P.
, and
Stambke
,
M.
,
2012
, “
Potential of Laser-Manufactured Polymer–Metal Hybrid Joints
,”
Phys. Proc.
,
39
, pp.
84
91
.
6.
Schricker
,
K.
,
Stambke
,
M.
,
Bergmann
,
J. P.
, and
Brautigam
,
P. H.
,
2014
, “
Macroscopic Surface Structures for Polymer-Metal Hybrid Joints Manufactured by Laser Based Thermal Joining
,”
Phys. Proc.
,
56
, pp.
782
790
.
7.
Rodriguez-Vital
,
E.
,
Lambarri
,
J.
,
Soriano
,
C.
,
Sanz
,
C.
, and
Verhaeghe
,
G.
,
2014
, “
A Combined Experimental and Numerical Approach to the Laser Joining of Hybrid Polymer-Metal Parts
,”
Phys. Proc.
,
56
, pp.
835
844
.
8.
Balle
,
F.
,
Wagner
,
G.
, and
Eifler
,
D.
,
2007
, “
Ultrasonic Spot Welding of Aluminum Sheet/Carbon Fiber Reinforced Polymer—Joints
,”
Materialwiss. Werkstofftech.
,
38
(
11
), pp.
934
938
.
9.
Balle
,
F.
,
Wagner
,
G.
, and
Eifler
,
D.
,
2009
, “
Ultrasonic Metal Welding of Aluminum Sheets to Carbon Reinforced Thermoplastic Composites
,”
,
11
(1–2), pp.
35
39
.
10.
Balle
,
F.
,
Wagner
,
G.
, and
Eifler
,
D.
,
2010
, “
Combined Process of Ultrasonic Welding and Precipitation Hardening of Aluminum Alloy 2024/Carbon Fibre Reinforced Composites Structures
,”
12th International Conference on Aluminum Alloys
(
ICAA
), Yokohama, Japan, Sept. 5–9, pp.
856
861
.http://www.icaa-conference.net/ICAA12/pdf/2E-16.pdf
11.
Balle
,
F.
, and
Eifler
,
D.
,
2012
, “
Statistical Test Planning for Ultrasonic Welding of Dissimilar Materials Using the Example of Aluminum–Carbon Fiber Reinforced Polymers (CFRP) Joints
,”
Materialwiss. Werkstofftech.
,
43
(4), pp.
286
292
.
12.
Goushegir
,
S. M.
,
dos Santos
,
J. F.
, and
Amancio-Filho
,
S. T.
,
2014
, “
Friction Spot Joining of Aluminum AA2024/Carbon-Fiber Reinforced Poly(phenylene sulfide) Composite Single Lap Joints: Microstructure and Mechanical Performance
,”
Mater. Des.
,
54
, pp.
196
206
.
13.
Andre
,
N. M.
,
Goushegir
,
S. M.
,
dos Santos
,
J. F.
,
Canto
,
L. B.
, and
Amancio-Filho
,
S. T.
,
2016
, “
Friction Spot Joining of Aluminum Alloy 2024-T3 and Carbon-Fiber-Reinforced Poly(phenylene sulfide) Laminate With Additional PPS Film Interlayer: Microstructure, Mechanical Strength and Failure Mechanisms
,”
Composites, Part B
,
94
, pp.
197
208
.
14.
Nagatsuka
,
K.
,
Yoshida
,
S.
,
Tsuchiya
,
A.
, and
Nakata
,
K.
,
2015
, “
Direct Joining of Carbon-Fiber-Reinforced Plastic to an Aluminum Alloy Using Friction Lap Joining
,”
Composites, Part B
,
73
, pp.
82
88
.
15.
Buffa
,
G.
,
Baffari
,
D.
,
Campenella
,
D.
, and
Fratini
,
L.
,
2016
, “
An Innovative Friction Stir Welding Based Technique to Produce Dissimilar Light Alloys to Thermoplastic Matrix Composite Joints
,”
Proc. Manuf.
,
5
, pp.
319
331
.
16.
,
P.
,
Hovanski
,
Y.
,
Fifield
,
L. S.
, and
Simmons
,
K. L.
,
2015
, “
Friction Stir Lap Welding of Aluminum-Polymer Using Scribe Technology
,”
Friction Stir Welding and Processing VIII
, Springer, Cham, Switzerland, pp.
153
161
.
17.
Lazarevic
,
S.
,
Ogata
,
K. A.
,
Miller
,
S. F.
,
Kruger
,
G. H.
, and
Carlson
,
B. E.
,
2015
, “
Formation and Structure of Work Material in the Friction Stir Forming Process
,”
ASME J. Manuf. Sci. Eng.
,
137
(
5
), p.
051018
.
18.
Mishra
,
M. S.
, and
Ma
,
Z. Y.
,
2005
, “
Friction Stir Welding and Processing
,”
Mater. Sci. Eng. R
,
50
(1–2), pp.
1
78
.
19.
,
P. L.
,
Leonard
,
A. J.
,
Shercliff
,
H. R.
, and
Withers
,
P. J.
,
2009
, “
Friction Stir Welding of Aluminum Alloys
,”
Int. Mater. Rev.
,
54
(
2
), pp.
49
93
.
20.
Fehrenbacher
,
A.
,
Smith
,
C. B.
,
Duffie
,
N. A.
,
Ferrier
,
N. J.
,
Pfefferkorn
,
F. E.
, and
Zinn
,
M. R.
,
2014
, “
Combined Temperature and Force Control for Robotic Friction Stir Welding
,”
ASME J. Manuf. Sci. Eng.
,
136
(
2
), p.
021007
.
21.
Shrivastava
,
A.
,
Pfefferkorn
,
F. E.
,
Duffie
,
N. A.
,
Ferrier
,
N. J.
,
Smith
,
C. B.
,
Malukhin
,
K.
, and
Zinn
,
M.
,
2015
, “
Physics-Based Process Model Approach for Detecting Discontinuity During Friction Stir Welding
,”
,
79
(1–4), pp.
605
614
.
22.
Darcy
,
H. P. G.
,
1856
,
Les fontaines publiques de la Ville de Dijon
,
Dalmont
,
Paris, France
.
23.
Amico
,
S.
, and
Lekakou
,
C.
,
2001
, “
An Experimental Study of the Permeability and Capillary Pressure in Resin-Transfer Molding
,”
Compos. Sci. Technol.
,
61
(
13
), pp.
1945
1959
.
24.
Bainbridge
,
I. F.
, and
Taylor
,
J. A.
,
2013
, “
The Surface Tension of Pure Aluminum and Aluminum Alloys
,”
Metall. Met. Trans. A
,
44
(
8
), pp.
3901
3909
.
25.
Bahraini
,
M.
,
Molina
,
J. M.
,
Kida
,
M.
,
Weber
,
L.
,
Narciso
,
J.
, and
Mortensen
,
A.
,
2005
, “
Measuring and Tailoring Capillary Forces During Liquid Metal Infiltration
,”
Curr. Opin. Solid State Mater. Sci.
,
9
(4–5), pp.
196
201
.
26.
Gebart
,
B. R.
,
1992
, “
Permeability of Unidirectional Reinforcements for RTM
,”
J. Compos. Mater.
,
26
(
8
), pp.
1100
1133
.
27.
Nandan
,
R.
,
Roy
,
G. G.
, and
Debroy
,
T.
,
2006
, “
Numerical Simulation of Three Dimensional Heat Transfer and Plastic Flow During Friction Stir Welding
,”
Metall. Mater. Trans.
,
37
(
4
), pp.
1247
1259
.
28.
Crawford
,
R.
,
Cook
,
G. E.
,
Strauss
,
A. M.
,
Hartman
,
D. A.
, and
Stremler
,
M. A.
,
2006
, “
Experimental Defect Analysis and Force Prediction Simulation of High Weld Pitch Friction Stir Welding
,”
Sci. Technol. Weld. Joining
,
11
(
6
), pp.
657
665
.
29.
Shultz
,
E. F.
,
Fehrenbacher
,
A.
,
Pfefferkorn
,
F. E.
,
Zinn
,
M. R.
, and
Ferrier
,
N. J.
,
2013
, “
Shared Control of Robotic Friction Stir Welding in the Presence of Imperfect Joint Fit Up
,”
J. Manuf. Processes
,
15
(
1
), pp.
25
33
.