Influence of Tool Thread Pitch During Friction Stir Welding of High-Density Polyethylene Plate

The push for more economical and efficient materials drives increased usage of lightweight plastics. Reducing a vehicle’s weight by 10% can increase fuel economy by an estimated 5−7% in an automobile [1]. Plastic applications include but are far from limited to body panels, lamp lenses and screens, fuel and other liquid tanks, equipment housings, and wear items such as gears and bushings [1]. Integrating polymers into engineered structures requires the ability to join them to each other and dissimilar materials. Joining methods using mechanical fasteners, adhesives, and welding have different benefits, limitations, and applications.

Welding of polymers, put simply, utilizes the melting of thermoplastics to promote bonds that solidify after cooling. The bonds form through adhesion, which includes variants such as van der Waals forces, covalent bonds, capillary bonds, electrostatic bonds, magnetic bonds, mechanical interlocking, and interdiffusion [2]. Typically, no welding joint relies solely on only one of these adhesion mechanisms. Miscible polymers can achieve interdiffusion, where the molecules become intertwined across a joint, due to their homogeneous properties (i.e., similar glass transition temperature). Polymer interdiffusion can be described by the reptation model, where two similar amorphous polymers held together above their glass transition temperature will diffuse together [3]. The reptation model describes a polymer chain restricted inside of a hypothetical tube. The reptation time, τ, is the time for the polymer chain to move completely, like a snake, through the hypothetical tube. τ is inversely proportional to the polymer’s temperature, so higher temperatures will have faster interdiffusion, or movement of polymer chains across each other [3].

Of the many methods to join polymeric materials, friction stir welding (FSW) is an attractive process utilizing a rotating non-consumable tool to melt and blend polymers at their joining surface. FSW necessitates melting of the material only during joining of plastics. Seemingly unknown to many, FSW, patented by Thomas et al. in 1991 at The Welding Institute (Cambridge, UK) was originally intended for solid state joining of aluminum alloys and dissimilar metals [4]. It is considered an environmentally friendly manufacturing process since the low processing temperatures are energy efficient, no filler material is needed, and no gasses or fumes are produced. FSW polymers is not a solid state process since polymers consist of entangled chained molecules with different lengths that must melt for them to fuse across a joint [5]. Melting is necessary for welding plastics since a polymer chain cannot cross the path of another chain. Carbon, oxygen, and nitrogen typically construct these chains, and their bonds are arranged at various angles in solid form. When polymers melt, these molecular bonds can rotate freely allowing for entangled polymer chains to slip past each other [6]. Polymer structure is analogous to ramen noodles in a square pack. The dry noodles are entangled in a brick and no single noodle can be removed without breaking. After cooking and softening the noodles, it is possible to remove a now compliant noodle without rupturing it. An advantage with FSW is that it generates heat in both joining parts, allowing for welding of dissimilar polymers such as acrylonitrile butadine styrene (ABS) and polycarbonate (PC) [7]. Ultrasonic welding is an example where welding dissimilar polymers with drastically different melting temperatures is challenging since the vibrations during welding may melt one polymer and not the other [8].

Extensive studies have shown that parameter selection and tool geometry greatly affect the quality of friction stir welded joints in both metals and polymers [9,10]. Polymers present additional challenges over metals to FSW due to their low thermal conductivity and limited range between melting temperature, T_m, and thermal degradation temperature, T_d. To weld polymers, the temperature in the weld zone must be high enough to allow slippage of polymer chains but cannot exceed temperatures that lead to degradation. Polymer degradation is a change in chemical or physical structure induced by factors such as excess heat or light that leads to reduction of strength or changes in other properties such as color and shape [11].

Tool geometries have an effect on heat generation, material flow and stress, welding forces, and ultimately weld quality. For example, it has been shown that the number of edges on a tool pin for various geometries influences the material flow, temperature, hardness, strain, and welding forces. Generally, increasing the number of edges on a tool increases the temperature and traverse forces [12]. Tool pin geometries explored for FSW polymers include square [13–16], triangular [15–17], cylindrical [13–15,18], tapered [13,14,16–18], threaded cylindrical [13,15,17–19], and threaded tapered [17,18], each of which exhibit different behavior. Triangular and square pins tend to have a pulsating effect which can provide good material mixing [14] but can also induce blow holes through the bottom of welded sheet [20]. Smooth cylindrical pins tend to poorly mix polymers resulting in very weak joints [14,21]. Smooth tapered pins though tend to produce better quality joints than straight cylindrical geometries [21]. Several studies test threaded tool pins against non-threaded profiles. These studies, however, do not account for the thread pitch as a variable parameter. The thread pitch is the distance, typically in millimeters, between thread crests as illustrated in Fig. 1. Another metric is the thread count per unit length of the shaft, and is often described using threads per inch (TPI). For the threaded tools used, these studies conclude that threaded tool pins can form joints with high joint efficiency for different polymers [22] with lower linear force due to greater heat generation [20]. The increased heat generation from threaded pins is a result of increased surface heat flux from a larger interfacial surface area [23]. The orientation of thread flutes with respect to tool rotation direction also affects material flow and joint quality. Thread flute movement from the bottom of the tool to the material surface tends to pull the welded material out of the joint, whereas better joint quality is achieved by downward moving flutes that force material into the joint [19].

Researchers performed extensive studies pertaining to the effects of threads on FSW tools in polymers, but very little has been done to include thread pitch as an experimental parameter in FSW polymer optimization. However, varying thread pitch has been studied for various aluminum alloys. A study utilizing bead-on-plate welds for various aluminum alloys found that intermediate thread pitches were optimal for weld quality and reduction of defects [24]. The thread pitches ranged from 1.02 mm (25 TPI) to 3.18 mm (8 TPI) on large 15.9 mm diameter cylindrical pin tools. Welding forces and their relationship to thread pitch varied between aluminum alloys. Another study joining butted AA1080 plates found that larger thread pitches for a particular set of welding parameters has a drilling effect, where metal chips move up the tool and are expelled from the weld zone [25].

A thread pitch study was also conducted during friction stir spot welding (FSSW) of polypropylene (PP) plates where thread pitch ranged from 0.8 mm to 2.0 mm [13]. This study produced similar conclusions to that of Boz et al. [25], where a smaller thread pitch produced stronger welds and too large of a thread pitch expelled material from the weld zone.

The FSW community lacks studies pertaining to the effects of thread pitch during traversing polymer welds. Due to fundamental differences in material flow during FSW of alloys and polymers, thread pitch effects found in Rabby et al. and Boz et al. cannot be assumed to directly correlate to polymers. It has been shown that material flow during FSW polymers differs from that of metals, where polymers exhibit fewer material flow zones and a much smaller thermomechanically affected zone (TMAZ) [26]. Lateral weld fractures also tend to form along the retreating side due to lack of consolidation, whereas in metals defects tend to occur along the advancing side.

Former studies indicate that sound polymer joints with FSW requires finely tuned interacting welding parameters. The tool pin geometry plays a significant role in heat generation, material flow, and weld consolidation. Threaded pins can improve these functions over other pin types, but the field lacks studies pertaining to the influence of thread pitch during polymer welds. This study considers the influence of thread pitch on welded high-density polyethylene (HDPE) tensile strength, tool temperature, and tool forces. Because welding parameters can have interaction with each other, tool rotation and traverse speeds are also considered. The effects of these parameters are tested using design of experiments (DOE).

High-density polyethylene consists of long hydrocarbon chains with very little branching [27]. Branching refers to architectures where short polymer chains protrude off of a long main chain [28]. These chains are formed by free radical addition polymerization of C₂H₄ repeated units. Polyethylene, a polyolefin resin, is by far the most commercially used polymer and can be processed into different forms [1]. HDPE is the most rigid of the polyethylene forms due to less than 7 branches per 1000 carbon atoms that allows for tight packing of polymer chains, making it a highly crystalline material [3]. It is also studied most often in the field of FSW compared to other thermoplastics with a presence in approximately 36% of publications as of 2018 [14]. Thus, its popularity in FSW and industry make it a good candidate to study the influence of thread pitch.

FSW experiments occurred in the Vanderbilt University Welding Automation Lab (VUWAL) on a modified Kearney and Trecker Milwaukee Model K milling machine. External motors mounted to the machine controlled using Simulink provided precise automated operation. Three thread pitches were tested using varied tool rotation and traverse speeds in HDPE. The physical properties of HDPE used in this study are available from the data sheet by Polymer Industries [29].

The tested rotation speeds were 1000 rpm, 1250 rpm, and 1500 rpm, and the traverse speeds were 25.4 mm/min, 63.5 mm/min, and 101.6 mm/min. The shoulderless tool rotated within a stationary shoe heated to 65^◦C at the shoe’s tool opening with zero tilt. Stationary shoes prevent ejection of molten polymer during the welding process [26]. Heating the stationary shoe aids in controlling the welded polymer’s cooling rate, reducing voids induced from fast shrinkage and ultimately improving the weld quality [33,34]. A pillow block bearing was fixed to the tool shank. Floating spring loaded slides guided the stationary shoe to follow the tool and apply pressure to the work piece as it traversed. Guide rails clamped to the work piece prevented the shoe from rotating with the tool.

A thermocouple inside the main tool shank measured the tool temperature during the entire welding process, allowing for the authors to infer heat generated from the tool temperature change since the tool ultimately conducts a portion of the heat generated [35]. Measuring the temperature inside the tool isolated the measured temperature changes from the shoe’s heat input so that only the tool parameter effects were observed. The thermocouple is located at the top of the tool pin 34 mm from the weld centerline and connects to a Teensy LC microcontroller mounted on the tool spindle that stores the temperature data on a micro SD card. Temperature data collection is activated by an accelerometer that senses tool rotation. Tool temperature change was calculated using the difference between the minimum and peak value measured during each weld. The tool was cooled to room temperature before the start of each weld and it reached peak temperature before tool retraction. Therefore, the minimum and peak tool temperatures correspond to the temperature prior to tool plunge and at tool retraction. The values were normalized against the time duration since the traverse speed affects the welding time.

A Kistler type 9123C dynamometer measured forces experienced on the tool in the axial and planar directions. Welding forces were analyzed by first applying filters to the raw data from the dynamometer. The largest contributor to signal noise indicated by the frequency spectra of the data sets is the tool rotation frequency. Thus, low bandpass filters with cut-off frequencies of 10 Hz, 15 Hz, and 20 Hz were applied to 750 rpm, 1125 rpm, and 1500 rpm data sets, respectively.

Mechanical tests followed the ASTM D638-14 guidelines to characterize tensile strength. Specimens used in tensile testing were precisely cut using a Shapeoko XXL CNC router. Figure 3 provides a schematic and shows the real setup of the FSW tooling used in this experiment, outlining the dynamometer, temperature measurement system, and stationary shoe.

A series of bead-on-plate welds followed a face-centered central composite (FCC) design of experiments. The bead-on-plate configuration was chosen to reduce variation in experimental setup from variables such as surface roughness between butted plates and possible tool offset from machine backlash. Joint configurations also play a role in the strength and joint characteristics [36], so the bead-on-plate configuration was used to provide a simple foundation. An FCC design is a type of response surface methodology (RSM), which are useful for analyzing problems where multiple variables influence the response. In essence, a fitted model with two parameters produces a three-dimensional surface, hence the name RSM [37]. FCC designs are beneficial for three factor experiments because they only require three levels for each parameter, requiring fewer runs, and consequently resources, than uniform-precision designs. Because tool parameters tend to have quadratic effects on material strength [38], a second order model is necessary to describe the parameter response and requires a minimum of three parameter levels to do so [37]. A cube can signify a three-level design, with each cube axis representing a range of values for a respective parameter. Each corner of the cube lies at a minimum or maximum. The FCC design uses data points from each corner, face center, and the cube center. A disadvantage of the FCC design is that it is non-rotatable, meaning some biases can occur since the corner and face points are not equidistant from the center [37]. However, FCC designs utilize center points that allow for a good estimate of experimental error.

The design utilized three parameters each with three levels, requiring a total of 20 runs with six repeated center point runs. Replicates for selected non-repeated runs were added to the experiment to refine the models from experimental error and investigate unusual observations. Minitab statistical software generated a random order to complete the runs. Table 2 outlines the values for each parameter level. The parameter levels were determined using observations and data from preliminary experiments and scientific literature review. A summary of optimum FSW joining parameters for various polymers is shown in a literature review by Zafar et al. [22]. The thread pitch levels are determined by equally spaced thread counts, where the center point is the average thread count of the high and low levels. Thus, the low level has smallest thread count, but the largest thread pitch. The high level has the greatest thread count, but the smallest thread pitch. Thread pitch, rotation speed, and traverse speed will be referred to as A, B, and C respectively moving forward. Responses included the tool axial and planar forces, tool temperature change from tool plunge to tool retraction, and average tensile data between five specimens from each run. Table 3 shows the experimental design with the response values. “Run” is the group number of each combination of parameter levels. “Run Order” refers to Minitab’s randomized order in which each run was completed. The responses were analyzed using Minitab’s response surface analysis tools.

Figure 4 shows the surface of completed welds for each parameter combination. The highest experimental maximum tensile strength was 23.18 MPa using A = 1.27 mm (20 TPI), B = 1500 rpm, and C = 25.4 mm/min. This results in an 89% joint efficiency from the 26.0 MPa measured in virgin material.

Minitab by default produces a response surface model consisting of all parameters, square of these parameters, and 2-way interactions. Each term in the model has a combination of parameter values multiplied by a coefficient. For example, an interaction term between thread pitch and rotation speed will look like β_ijA × B, where β_ij is the coefficient corresponding to their contribution to the model, and A and B are the respective parameter values. Five terms are statistically significant at a 95% (α = 0.05) confidence interval with their F and P values displayed in Table 4. The critical F and T values (F_crit and T_crit) were calculated in Minitab, where F_crit = 3.885 and T_crit = 2.093. A statistically significant parameter term has P < α and F > F_crit when using an F test and |T| > |T_crit| when using a T test. Removing all of the terms incorporating thread pitch improves the coefficient of determination (R²) for the model. However, the lack of fit (LOF) parameter has a P-value of p = 0.000. When p ≤ 0.050, the lack of fit has significantly more error than the pure error, and the regression model does not adequately describe the data, even if the data have a good fit to it [37]. Upon further investigation, it was found that the model residuals were skewed to the left, as shown in Fig. 5. The default models assume that the residuals are normally distributed which can skew model predictions due to the non-constant variance of the observations. Non-normal residual distributions can lead to inaccurate parameter significance levels [37]. Applying a variance stabilizing transformation can normalize an error distribution.

The reduced transformed model has linear, square, two-way interaction, and error contributions of 43.22%, 37.48%, 4.15%, and 15.15%, respectively. The high contribution of square terms confirms curvature in the model response and that three levels were indeed necessary for this study. C, C², B², A², and A are statistically significant terms in the reduced model with 34.67%, 13.16%, 20.86%, 3.46%, and 6.08% contribution. Parameter contribution is calculated by the percentage of the respective term’s sequential sum of squares to the data’s total sequential sum of squares. The sequential sum of squares measures variation in the response from each parameter. Figure 6 shows a Pareto chart of the terms included in the reduced transformed model for maximum tensile strength. The magnitude of each parameter’s standardized effect is related to their contribution to the model response. B is left in the model to satisfy hierarchical effects, and B*C interaction improves the overall fit. The coefficient of determination for the reduced transformed model is R² = 84.85%, R²(adj) = 76.01%, and R²(pred) = 48.10%. The lack of fit is well beyond the threshold of statistical significance with p = 0.098. Though the regression has room for improvement, the lack of fit suggests that the model is a good representation of the parameter effects. The low R² values are similar to the outcome reported in Rezgui et al.’s HDPE FSW parameter study [38]. Their study used a slightly wider range of rotation speeds and a lower range of traverse speeds with a smaller center point.

Figure 7 shows the individual parameter contribution to tensile strength. The significance of the squared model terms is visually apparent due to the parabolic response when sweeping each parameter across its range. The minimum/maximum parameter responses in Fig. 7 indicate that the chosen parameter ranges were sufficient to capture the general behavior of parameter values. Figure 8 also visualizes this behavior through three-dimensional (3D) surface plots. Increasing traverse speeds sharply decreases the ultimate tensile strength when sweeping through the center point, while sweeping between the center point and high value has a smaller contribution to tensile strength. Rotation speed has a predicted minimum effect on tensile strength just below its center point, though overall the higher rotation speeds tend to promote greater tensile strengths as found in other studies [9]. The linear effects of rotation speed do not have statistical significance, nor do it in the study by Rezgui et al. [38]. Studies finding significant contribution from the rotation speed implemented higher rotation speeds than this study [9], but exceedingly large rotation speeds hinder weld strength from material expulsion and degradation [39]. The local minima near the center point for both rotation speed and traverse speed may suggest some interaction between the two parameters. Though no parameter interaction provided a statistically significant contribution to the ultimate tensile strength, the B × C interaction does improve the model fit. In the case of this study, every run incorporating the center point parameter values for both B and C resulted in tensile strength values below 7 MPa regardless of the thread pitch. In some cases where only B or C had a center point parameter value, tensile strength greatly improved. Thread pitch shows significant quadratic effects in the model with local maxima between the low and center point parameter values. The linear thread pitch term suggests that generally the coarser threads in this study produce welds with higher tensile strength. On average, the 1.27-mm pitch threads provide 2.83 MPa over the 0.58-mm pitch threads. The model predicts a drop-off in tensile strength at the coarse thread pitch in this study, though the response of the −1 thread pitch is still significantly higher than the +1 thread pitch. This may be a slight indication of the milling effect described by Boz et al. [25], though significant milling was not observed. As threads become finer and finer, the pin geometry approaches that of a smooth cylindrical pin, which as noted before performs poorly compared to other pin geometries. Thus, the downward curvature nature of the thread pitch response on tensile strength is expected.

Run 12 is labeled as an unusual observation with a disproportionally large residual, potentially making it an outlier in the data. The weld logs were checked for run 12 with no obvious issues. Further, replicate trials of run 12 produced similar results, so the original data were left in the model. Removing run 12 from the model altogether did not improve the regression or fit either. The poor regression model may be a product of non-rotatability in the FCC design. In the cube model described before, runs 1–8 are all radially equidistant from the center points due to their extremums along all three axes. Runs 9–14 are all radially closer to the center points since their parameters put them at an extremum along only one axis. Thus, the variance of the predicted response is not equal in all directions from the cube center [37]. Further evidence of the effects of non-rotatability is the extreme model predictions shown in plots “A versus B” and “B versus C” in Fig. 8. Both plots predict unrealistic maximum tensile strengths when A, B, and C are all at their optimum values.

The fractured tensile specimens displayed clear patterns of break morphology when categorized by thread pitch. All specimens ruptured on the retreating side of the weld zone. Polymers tend to rupture at the retreating side distinctly along the tool pin’s path since the TMAZ is limited to the stir zone during the FSW of polymers [26]. The smaller TMAZ size in polymers compared to metals is a result of polymers’ low thermal conductivity. Higher stresses on the tool advancing side tend to fully consolidate the stirred polymer with the base material. The higher stresses arise from molten polymer being pulled around from the back side of the pin, which is then extruded into the advancing side. The retreating side experiences a smaller forging pressure, which results in less consolidation at the stir zone wall [26].

The first eight runs with the +1 and −1 thread pitch levels follow a 2³ full factorial design, consisting of two levels for each parameter. Runs 9 and 10 use +1 and 1 level thread pitch with center points for rotation speed and traverse speed. Thus, the fracture surface for each thread pitch can be compared between rotation/traverse speed pairs. Figure 9 shows micrographs of weld cross sections at the weld retreating side for coarse and fine thread welds at high and low rotation speeds. For both rotation speeds, the coarse threads show enhanced polymer interdiffusion between the weld zone and base material over the fine thread tools as indicated by a fainter defect line visible at the stir zone interface. Figure 9 also lacks a heat-affected zone (HAZ) far beyond the stir zone, which is common for polymer FSW [22]. Figure 10(a) indicates a fracture between the stir zone and base material along the retreating side of the weld, which was observed for all runs. Figure 10(b) shows fracture surfaces of coarse and fine thread tensile specimens. Lightly applied brushed gold paint adhered to the high spots on the fracture surface to provide a distinction between the fracture surfaces. In general, the fine thread pitch specimens tend to have a consistent vertical break at the stir zone/retreating side interface, with a surface reflective of the tool threads. The tool thread shape at the interface reveals a kissing bond, where the polymer flows to fill the cavity left by the tool but adheres mostly between the joint surfaces. The coarse thread pitch specimens have less distinct fracture patterns that indicate fewer kissing bonds than seen in the finer thread pitch specimens. The less distinct fracture pattern in the coarse thread specimens may be a result of greater interdiffusion across the joints induced by greater mixing. The coarse thread tool provides a greater overlap between the stir zone and base material with a thread depth of over double that of the fine thread tool, shown in Table 1. Furthermore, the coarser threads have a higher velocity of material in the thread grooves, which facilitates greater volumetric heat generation in the shear layer around the tool [23] and aids in polymer interdiffusion.

The reader is encouraged to refer to Sun and Wu [23] for more information regarding Eq. (3). The linear parameter effects on tool temperature in this study agree with the relationships found in Sun and Wu’s model. Since κ relates to increases in the contact area, an increase in the thread surface area from a lower thread pitch will increase the heat generation. Sun and Wu demonstrate in their model that the thread pitch affects surface heat generation and volumetric heat generation inversely to each other. Heat is generated in the FSW process by both frictions at contact surfaces and through plastic deformation in the material shear layer. Smaller thread pitches have a greater surface area, which increases the heat generated at the interfaces along the threads. Conversely, less heat is generated in the shear layer between the threads with a small thread pitch, partially because the increased contact temperatures reduce the material flow stress. Therefore, the tool surface area dominates the heat generation for the thread pitch contribution. Temperature profiles that show increased total heat generation from finer threads are shown by Sun and Wu [23].

The thread pitch does not provide a statistically significant contribution to the tool axial forces during traversing welding periods. Rather, tool rotation speed has the greatest contribution (50.06%), followed by traverse speed (21.80%), and their squared terms which combined provide a 15.61% contribution in the reduced model. The thread pitch, however, has a large contribution to the plunge force. The rotation speed once again has the greatest contribution at 56.93%, followed by thread pitch at 17.14%, and square of rotation speed at 5.16% in the reduced model. There is no significant interaction between the thread pitch and rotation speed. Traverse speed provides no contribution during the plunging period since no lateral movement occurs.

As stated in Sec. 3.3, the rotation speed and thread pitch provide the greatest contribution to tool temperature and therefore heat generation. Heat generation is inversely proportional to axial force under a constant plunge rate since heat softens the material [40]. Since rotation speed has the greatest contribution to heat generation, it is no surprise that the highest rotation speeds result in the lowest plunge forces. The thread pitch has an inverse effect on the maximum plunge force than the heat generated, where a lower thread pitch actually results in a higher maximum plunge force, as shown in Fig. 13. The maximum plunge force occurs during initial contact between the tool and the material. The tool axial force slowly reduces throughout the plunge as the material softens [40]. Thus, vertical side tool geometry provides no contribution to the plunge force at initial contact, which means that the thread pitch’s contribution to the plunge force is a result of its effect on the tool’s bottom surface. Most notably, the surface area decreases for finer threads, and the center of the flat region becomes more offset from the tool’s center axis, as shown in Fig. 14. A 10% reduction in the surface area occurs between the highest and lowest thread pitch tools used in this study. The bottom tool surface areas for coarse to fine thread pins are 24.5 mm², 25.16 mm², and 27.19 mm², respectively. The bottom surface offset carves and expels material through pulsation, and the small surface area inherently requires a smaller force to plunge. For runs 1–10, the coarse threads on average have a 25% reduction in plunge force over the fine threads. Because the greatest forces during FSW occur during the plunge, smaller machine requirements are necessary for coarser thread tools.

The measured planar forces did not produce any meaningful models for parameter influence solely in the traverse or lateral directions. However, the lateral axis consistently experienced considerably higher tool forces toward the advancing side relative to the traverse axis. Planar force measurement failed during run 11, as indicated by “*” in Table 3. Since no correlation was present in the planar forces without run 11, no further replicates were performed to include it in the data. The traverse forces F_t ranged from approximately 3 N to 120 N, whereas the lateral forces F_l ranged from 17 N to 170 N. The planar forces, where $Fp=(FT2+FL2)$⁠, show a general relationship with the greatest contribution from traverse and rotation speed. Slow traverse speeds and high rotation speeds tend to reduce overall planar forces. Though these trends are apparent, the lack of fit for all produced models is well beyond the threshold of statistical significance. Lack of fit for planar forces is likely due to the implementation of the stationary shoe, whose sliding friction and interaction with the polymer melt have overwhelming effects on the tool parameters. In studies by Mendes et al. [41] and Eslami et al. [42], a contrary relationship was found where the traverse forces were significantly higher than the lateral forces. The axial forces were also significantly higher than those reported in this study. The difference in forces between the two studies lies in stationary shoe’s degrees-of-freedom. In the current study, the shoe is allowed to slide along the vertical tool axis independent of the tool, whereas the shoe in the study by Mendes et al. is fixed to the axis. Thus, the shoe has a greater axial force in Mendes et al., contributing to higher frictional forces in the traverse direction. Since the shoe’s smaller axial pressure in this study allows it more freedom to float across the polymer, a higher net lateral force can be observed.

Thread pitch clearly has noticeable effects on the ultimate tensile strength, heat generated, and contact plunge force during FSW of HDPE. Understanding the effects on machine forces is necessary for a reliable tooling setup during manufacturing. The thread pitch has noticeable effects on the axial forces, but the required forces are related to the tool’s bottom contact area and the degree to which it carves away the base material. The coarse threads have a larger “cutting area” at the tool’s bottom leading edge, allowing it to perform with lower plunge forces. With a stationary shoe setup necessary to FSW many polymers, strong correlations are not observed with the parameters studied, and it is likely that the thread pitch has negligible effects.

The tool temperature data produced a much better fitting model than the tensile data. This model shows that the heat generated is clearly more predictable than joint strength due to several reasons. The linear relationship between all main effects inherently makes for a simpler model to fit. Heat generation is a lower level effect than joint strength since joint strength is dependent on heat generation in FSW. The joint strength, therefore, has more variables to consider, such as the stationary shoe temperature and contact pressure. It is also clear that an improved model to predict weld strength will require a wider range in rotation speed.

Generally, a hotter melt allows a polymer to flow more easily so that molecules can diffuse across a joint and bond to other molecules [27]. Interestingly, the higher tool temperatures in this study do not necessarily correlate to higher weld strength. Figure 15 shows a comparison between the maximum tensile strength and tool temperature for all 20 runs and does not show a relationship between the two.

The tool temperature does not adequately predict tensile strength for FSW polymers due to the low polymer thermal conductivity. As mentioned in Sec. 3.3, the surface interaction between the tool and polymer dominates the heat generation over material layer shear, so most heat is generated at the tool’s surface. The tool, made of tool steel, has a significantly higher thermal conductivity than the HDPE, permitting more heat flow to the tool than into the polymer. This contradicts the results found by Chao et al., but they welded AA 2195-T8 which has an estimated 3 × the thermal conductivity over their tools [35], whereas the tools used in this study have an estimated thermal conductivity of 50 × that of the HDPE. Therefore, changing the parameters will not drastically alter the heat flux through the polymer, keeping the melt local to the tool. Another property difference between aluminum and polymers is the change in thermal conductivity with temperature. Aluminum, like many metals, has a maximum thermal conductivity on a downward curve [43], and this maximum occurs within temperatures experienced during FSW. Polymers have an anisotropic thermal conductivity because the heat transport favors the chain alignment direction. Semi-crystalline polymers such as HDPE tend to have higher thermal conductivity than amorphous polymers because there are more orderly packed chains to quickly transfer molecule vibrations [44]. As the polymer heats up, the spacing between the adjacent polymer chains increases, and order reduces through rotation in the chains, overall decreasing the thermal conductivity with an increase in temperature and further slowing the spread of heat.

Though the finer threads have greater total heat generation due to more friction from a higher surface area, the coarse threads have greater volumetric heat generation from more severe plastic deformation [23]. Thus, it appears that the total heat generation alone is not the main contributing factor to friction stir welded polymer weld strength.

Chain entanglement at a polymer interface is the most important factor for polymer–polymer adhesion. The polymer adhesion is proportional to the fourth root of the annealing time under contact [2]. The slow tool traverse speeds during FSW cater to this critical annealing time for full interdiffusion. The pressure experienced by polymers around the tool is also non-uniform. According to Derazkola et al., polymers experience a low-pressure zone immediately following the traversing tool pin for a smooth tapered tool [16]. This low-pressure zone prevents full polymer adhesion if applied to the adjoining surfaces too early. The asymmetry in pressure around the tool is a result of higher strain rates at the leading edge and advancing side of the tool, where higher strain rates induce higher pressures [16]. Thus, coarser threads and slower traverse speeds may increase pressure at the tool’s trailing edge and limit tension pulling on solidifying polymer at the stir zone walls. The slow traverse speeds also maintain a molten weld pool behind the tool, similar to a metal fusion welding process. Strong polymer joints rely on this weld pool to heat the polymer beyond the stir zone interface. The longer the molten weld pool has contact with the stir zone interface, the more interdiffusion occurs. The enhanced polymer interdiffusion is analogous to metal FSW where slow traverse speeds facilitate greater inter-material mixing [45]. The reptation model described by Cowie states that polymers in contact will diffuse when they are above T_g, but this only applies to polymers in an amorphous state [3]. HDPE has a glass transition temperature, T_g, of −125 °C [6], but it returns to a semi-crystalline state as it cools, so interdiffusion will not continue after the melt solidifies. Therefore, a strong semi-crystalline polymer weld requires prolonged contact between the molten pool and base material, which is most effectively achieved by a low traverse speed.

A tool parameter study following a face-centered central composite response surface design was used to determine the effects of thread pitch during friction stir welding of polymeric materials. The three tool pins had thread pitches of 1.27 mm (20 TPI), 0.79 mm (32 TPI), and 0.58 mm (44 TPI). The studied effects of thread pitch were the ultimate tensile strength, fracture surface, tool temperature, and axial and planar forces. The study’s results indicate the following:

• Thread pitch has statistically significant effects on the ultimate tensile strength, heat generation, and axial plunge force. Thread pitch does not have a significant contribution to traversing axial or planar forces.

• The model for maximum tensile strength indicates both linear and quadratic contributions of thread pitch, but the strongest welds were skewed to coarser threads.

• The fine threads generate the greatest tool temperatures, but the tool temperature does not correlate to the maximum tensile strength.

• The coarser threads provide greater volumetric interaction with the polymer, improving polymer entanglement between the stir zone and base material.

• Tool threads have little effect on the welding forces. Significant thread pitch contribution to welding forces occurs only during the plunge. Coarse threads require smaller plunge forces due to less surface area on the pin bottom and increased drilling.

• The strongest weld joints are a result of prolonged contact between a molten weld pool behind the tool and base material outside the stir zone that allows for sufficient interdiffusion across the joint. Tool traverse speed provides the greatest contribution to polymer interdiffusion, as slow traverse speeds maintain a weld pool and decrease the cooling rate.

It is clear from other studies that parameter values outside the range of those performed in this study exhibit other pronounced behaviors. For example, the tool rotation speed has a greater contribution to joint strength when values range 2000–3000 rpm. Exceedingly coarse threads may have a more pronounced milling effect on the polymer that will have a negative effect on joint strength. Further, polymer types also play a role in parameter influence based on their degree of crystallinity and molecular weight. To fully capture the influence of tool thread pitch across these wider scenarios, it will be useful to employ a uniform-precision response surface design, which will utilize more levels for each parameter that can better capture the welding behaviors across a wider range at the expense of performing more runs. The uniform-precision design will also not apply biases to the predicted responses in specific parameter regions.

This work was supported by the NASA Space Grant Program through the Tennessee Space Grant Consortium.

There are no conflicts of interest.

The data sets generated and supporting the findings of this article are obtainable from the corresponding author upon reasonable request.