Process envelope and void defect analysis in FSW of AA2024,AA5052 and AA7075 aluminium alloys

Abstract

This article investigates void defect formation in friction stir welding (FSW) of AA2024, AA5052 and AA7075. A total of 159 bead-on-plate welds were performed with real-time monitoring of temperature, force and spindle power. Travel speeds from 25 to 500 mm/min and tool rotation speeds from 300 to 1500 RPM were investigated. Process maps identified critical parameter ranges for different void defects. A new method was proposed to estimate torque from spindle power. Results show that high temperatures ( $> 0.7 T_{solidus}$ ) are necessary but not sufficient for defect-free welds. Torque thresholds can distinguish cavity from tunnel and groove defects, while power input below 1000 W (AA2024) or 1050 W(AA7075) is generally necessary and sufficient to avoid voids. These findings enable real-time defect monitoring and process optimisation in FSW.

Keywords

friction stir welding (FSW)defect map void defects torque estimation welding quality assessment process optimisation

Introduction

Friction stir welding (FSW), invented by TWI in 1991, has become widely used in industries such as aerospace, automotive, and shipbuilding. Despite its maturity, FSW still faces challenges, particularly in avoiding welding defects that compromise joint strength and reliability. Among these, void defects are especially detrimental. Their formation is influenced by a complex interplay of process parameters, material properties and tool design, yet the mechanisms remain incompletely understood.

Many studies have mapped optimal FSW process windows for various aluminium alloys, such as ADC 12,¹ AA2219,² AA2024/AA1100³ and AA5083/AA6082/AA7075.⁴ These investigations primarily focus on identifying suitable welding parameters, but they do not provide detailed microstructural characterisation or systematic classification of defects observed in the joints. Some studies have attempted to identify critical parameters for defect-free welding,^1,^3–7, however, none of these have incorporated an evaluation of all the outputs including microstructure, peak temperature and torque. Kim et al.¹ emphasised the influence of plunge downforce on the defect-free windows. Abboud et al.⁷ claimed defect-free welds in AA6082 and AA5083 with peak temperatures above 0.65 $T_{s o l i d u s}$ .⁷ Force and contact pressure responses have also been related to defect formation.^6,8 Qian et al.³ also built an analytical model to obtain optimal process window based on temperature and strain rate. Although some machine learning models have been applied for defect prediction,^5,8 these have not been able to correlate defects conclusively to a combination of physical variables such as temperature and torque. Numerical simulations have also been used to advance understanding of material flow and void formation,^9–11 but their high computational cost and limited generalisation restrict industrial application. Despite these advances, a convincing relationship between process parameters, physical variables (temperature, force, torque) and void defect formation is still lacking.

This study addresses this gap by systematically investigating void defects in FSW of several aluminium alloys across a wide range of welding parameters. Real-time data acquisition of temperature, force, and spindle power enables detailed analysis of process conditions. By correlating these variables with observed defects, the study constructs process maps that define safe operating windows for defect-free welds. The findings provide new insights into the mechanisms of void formation and offer practical guidelines for optimising FSW parameters and implementing real-time defect monitoring.

Experiments procedures

Materials and tools

Three types of aluminium alloys are chosen: AA2024-T4, AA5052-H32 and AA7075-T6. The selected alloys span a wide range of flow stress and defect susceptibility. Compared to more weld-tolerant alloys like AA6061, these materials are more prone to void formation. The chemical compositions are in Table S1 of supplemental materials. Each plate had a thickness of 5 mm and dimensions are shown in Figure 1(a). A threaded (M5 thread) and tapered tool made of H13 tool steel was employed. The tool has a right hand threaded probe under anti-clockwise rotation so that materials will be drawn downwards by the threads.¹² The tool also has a pin length of 4.3 mm, tapering from 4.5 to 5.5 mm, and a shoulder diameter of 15 mm. The tapered shape will also enhance the material stirring with a larger contact area and high hydrostatic pressure in the weld zone, as noted by Zhang et al.¹³ The geometry figure of the tool is included in Figure S2 in the supplemental materials.

Figure 1.

(a) Plate configuration (top view); (b) schematic experimental setup and tool geometry.

Welding parameters and data acquisition

The experiments were conducted using a Liburdi FSW machine (photos included in the supplemental materials Figure S1) operating in position control mode. A wide range of welding parameters was explored to determine the optimal process window and a comprehensive defect map, extending the limits until either the machine’s capacity (maximum ability: 1500 RPM, 500 mm/min) was reached or tool failure occurred. Both rotation speed (300–1500 RPM) and travel speed (25–500 mm/min) were varied to study their effects on weld quality, with a constant tool tilt angle of 2.5 degrees. Detailed welding parameters are provided in Tables S2 to S4. A total number of 159 welds were performed. With a 500 RPM, 50 mm/min baseline, plunge depth was tuned in 0.1 mm steps to achieve optimal surface conditions. The nominal plunge depths were set to 6.3 mm for AA2024, 5.1 mm for AA5052 and 6.4 mm for AA7075, based on the preliminary trials. It should be noted that these values do not represent the actual plunge depth during welding, as spindle and fixture deformation under load affects the final depth. The values selected here were selected to compensate the machine compliance and avoid surface profiles resulting in excessive surface flash. The initial dwell time was set to 5 seconds.

A data acquisition system was implemented to monitor temperature, force and spindle power during FSW (see Figure 1(b)). K-type thermocouples (TC-1, TC-3) were embedded at mid-thickness from the plate back, while TC-2 and TC-4 measured temperatures near the shoulder edge. The trajectory of the FSW tool centre was aligned with the position of thermocouples TC-1 and TC-3 by using the alignment slots in Figure 1(a). Temperature data was sampled at 10 Hz using NI hardware. Downward force and spindle power were recorded by the FSW machine system at 10 Hz. For AA5052, additional thermocouples were placed inside the pin. A thermal camera was applied to monitor tool temperature during welding.

Experimental results

This section examines key aspects of the FSW process, including defect analysis, the determination of the process window, and the assessment of temperature and spindle power data. Overall, 159 welds were conducted for those three types of materials as shown in the table in the appendix.

Defect examination and analysis

Each weld was sectioned at mid-length, prepared metallographically, and examined with a Keyence VHX microscope. Defects were classified as cavities, tunnel defects, or surface grooves based on cross-sectional morphology. Tunnels and grooves extend along the weld, while cavities are small and irregular. Only void-type defects were considered, as bead-on-plate welds lack joint interfaces.

For AA2024, several types of defects are observed, including cavities, tunnel defects (including incipient tunnel), and surface grooves, as shown in Figure 2(a). These defects are distributed in different regions of the defect map. Cavities are primarily found in the red area at RPM values higher than 650. The morphologies of these cavities vary, as illustrated in Figure 3(a), suggesting different formation mechanisms, which will be discussed in later sections. At high RPM and increased travel speed, both cavities and tunnel defects can appear in the same sample (purple area). When the tool rotation speed is 650 RPM or lower, only tunnel defects and surface grooves are observed (grey and blue areas). The incipient tunnel defect usually appears as several voids on the AS, indicating the early stages of tunnel formation. With increasing travel speeds, those voids become a continuous tunnel. This distribution indicates that each defect type has a preferred region on the map, with some overlap, suggesting distinct formation processes and conditions for each defect type.

Figure 2.

Defect maps for (a) AA2024, (b) AA5052 and (c) AA7075.

Figure 3.

(a) Cavities of AA2024. (b) Cavity in AA5052. (c) Cavities of AA7075.

For AA5052, the defects observed include surface grooves, tunnel defects, and singular cavity defects as shown in the defect map in Figure 2(b). Similar to AA2024, tunnel defects and cavity are distributed in distinct regions of the map, indicated by red and blue areas, respectively. However, in AA5052, defects are not present in the upper left corner of the process map. Moreover, cavities found in AA5052 are also different from those in AA2024. Only a singular cavity is found in AA5052 as shown in Figure 3(b). The cavity is a void with a round shape near the top edge in the advancing side. It formed at the junction of material flow between the shoulder and the stir zone of the pin.

For AA7075, cavities, surface grooves, and tunnel defects are observed. As shown in Figure 2(c), the optimal process window appears as a triangular region at the lower left of the map. Different defects are distributed across the parameter space. In the blue region, both tunnel defects and incipient tunnel formation are indicated with the same colour, as incipient tunnel represents the early stage of tunnel defect formation near the bottom on the advancing side (see Figure 3(c)). With increasing travel speed, these incipient tunnel defects transit into fully developed tunnel defects. In the red region, cavities similar to those found in AA2024 are observed near the centre and top of the stir zone for AA7075, as shown in Figure 3(c).

In summary, defect maps and process windows have been established for all three alloys based on macrostructural examinations. The size and location of the defect-free zones vary among the materials: AA2024 exhibits a small defect-free region at the lower left of the map, AA5052 shows a broader defect-free area at the upper left, and AA7075 presents a wider defect-free zone at the lower left. Considering the similar material flow in bead-on-plate and butt joints, the proposed process window should also minimise void defects for butt joints, while interface-related defects such as kissing bonds lie beyond the scope of this study. To further interpret these defect maps, analysis of the collected thermal and mechanical data is required. The influence of these physical variables on defect formation is discussed in the following sections.

Temperature analysis

Temperature is a key factor in defect formation, reflecting material hot deformation and phase changes. Maximum temperatures, taken as the higher of TC-1 or TC-3, are summarised in Figures S3 and S4 with outliers removed. Generally, higher temperatures are found at higher rotation speeds and lower travel speeds (upper left of the process map), though some deviations occur due to thermocouple displacement, especially when defects form, making precise measurement challenging.

Temperatures are further categorised by defect type, as illustrated in Figure 4. For reference, the solidus and eutectic melting temperatures are also indicated on the graphs. Equilibrium solidus and non-equilibrium eutectic temperatures were obtained using Jmatpro and are summarised in Table 1. The eutectic temperature is based on the calculation from Scheil’s model for multi-component alloys. These values have been cross-checked with published data and found to be consistent with those reported previously.^14,15

Figure 4.

Peak temperature distribution: (a) AA2024, (b) AA5052 and (c) AA7075.

Table 1.

Solidus and eutectic temperatures for different aluminium alloys.

Material	Solidus temperature $\circ$ C	Eutectic temperature $\circ$ C
AA2024	519	500
AA5052	612	N/A
AA7075	528	476

Across all three alloys, defect-free welds consistently exhibit a minimum temperature threshold of approximately 440 $\circ$ C, welds below this temperature are prone to defects. The homologous temperature ( $T / T_{solidus}$ ) is always above 0.7 for defect-free welds. This shows that adequate temperature is necessary to prevent defect formation. The maximum temperature ( $> 0.7 T_{solidus}$ ) for defect-free welds can approach and even exceed the solidus temperature. However, temperature larger than $0.7 T_{s o l i d u s}$ alone does not guarantee defect-free welds, as defects may still occur above an upper limit or critical temperature (e.g. $0.7 T_{s o l i d u s}$ ). Certain defect types, such as tunnel defects and surface grooves, tend to form within lower temperature ranges. Overall, the temperature distributions for different defect types and defect-free welds overlap substantially, making it difficult to establish a quantitative relationship between temperature and defect formation.

Torque–power model

While prior studies have studied spindle power and torque (from FSW tool) relationships,^16–18 direct calibration is still lacking and necessary due to significant energy losses in FSW spindle assemblies. A series of friction stir spot welds (FSSW) were performed with plates mounted on a 6-axis load cell (JR3) fixed to the machine table. The load cell directly measured the torque applied by the FSW tool. The tool was aligned with the load cell centre using an aluminium block with a tool-sized hole, ensuring accurate torque measurement. Photos are provided in Figure S5.

During spot welding, the tool is accelerated to the target RPM and plunged into the plate, holding for a 5-second dwell. Tests were performed at 500 to 1500 RPM and plunge speeds of 20, 60 and 100 mm/min to cover a wide torque range. Spindle power (from spindle current) and torque (from the load cell) were recorded from initial tool contact through dwell, both sampled at 10 Hz. Axial force was also recorded, and the initial Z-force was used to synchronise the two data acquisition systems.

In order to understand the relationship between the spindle motor power and the output torque, analytical analysis is shown below. The mechanical power output from the torque and the electric power from the motor can be calculated:

P_{torque} = ω M_{tool}

(1)

P_{electric} = f P_{nominal}

(2)

where

M_{tool}

\cdot

m) is the output torque from the FSW tool including the pin and the shoulder and

ω

(

s^{- 1}

) is the rotation speed. The spindle motor power percentage

f

is the output electric power input (

P_{electric}

) divided by the nominal power of the spindle motor (

P_{nominal}

). The total mechanical power output of the spindle motor is the sum of the mechanical power from the torque applied to the plate and the mechanical power loss from the bearing system. Inside the spindle motor (three-phase AC motor), power loss also exist such as copper loss, iron loss, etc.¹⁹ Both the power loss from the bearing system and the internal loss will be variable with different rotation speeds and loads. They are also highly related to the design of the whole system. In order to simplify the model process, the efficiency of the motor based on the output torque on the weld plate can be defined:

η_{torque} = \frac{P_{torque}}{P_{electric}}

(3)

It describes the energy conversion efficiency from the electric power to the actual mechanical output energy by the FSW tool applied to the plate. This efficiency can be considered as a function of rotation speeds and output torque

η = G (ω, M_{tool})

. The relation between the efficiency, RPM and output torque are shown in Figure 5(a). Based on the chart, the efficiency has a strong relationship with torque and rotation speeds. Hence, a quadratic equation is applied for different RPMs to fit the data:

η_{torque} = a M_{tool}^{2} + b M_{tool}

(4)

η_{torque} = 0

when the output torque from the tool is zero. The fitted curves are shown in Figure 5(a) and corresponding coefficients are shown in Table S5. The efficiency will increase with larger torque and higher RPM. This results show the validity of the quadratic model.

Figure 5.

Results of efficiency and torque fitting based on FSSW of AA7075. (a) Fitting results of efficiency ( $η_{torque}$ ) with measured torque for different rotation speeds. (b) Fitting results of measured torque with spindle power for different rotation speeds.

Based on equations (1), (2), (3) and (4), the output torque from the tool can be calculated based on the spindle power percentage, the equation of the torque is:

M_{tool} = \frac{(\frac{ω}{f \cdot P_{nominal}} - b)}{a}

(5)

M_{tool} = \frac{\frac{2 π}{60} \cdot RPM}{a_{1} \cdot f \cdot P_{nominal}} + b_{1}

(6)

In order to improve the prediction of torque values, equation (6) is used as the fitting equation between the power percentage and the output torque with least absolute residuals methods. The fitted curve is shown in Figure 5(b) and the fitting parameters are in Table S5.

The results show the ability of our model to correlate the tool torque with power percentage. This provides a way to estimate the torque from the power percentage data recorded during the welding process. It needs only a few friction stir spot welds to obtain the data. Based on the model, torque analysis can be conducted for all the welds. The torque data can be used to analyse the defect formation and the process window.

Torque analysis

With the fitted equations, torque during FSW can be predicted from the recorded spindle power percentage data of AA2024 and AA7075. An example of the recorded power percentage data is shown in Figure S6. The spindle power increases rapidly during the plunging and start to decrease and become more stable as the tool starts to travel.

The average spindle power percentage is calculated for the travelling process for each weld. Based on the model, the corresponding average output torque from the tool can be calculated for each weld. For experiments with RPMs which is not included in the spot welding experiments, the fitted equations of the closest RPM are used. The relationship between the average torque and the RPM is shown in Figure 6(a) and (b). The power input from the torque can also be calculated based on equation (1) and the results are shown in Figure 6(c) and (d). Heat input ( $P_{torque} / V$ ) for each weld is also calculated as shown in Figure 6(e) and (f). To better illustrate the trends, only travel speeds with at least three welds are included in those graphs.

Figure 6.

(a, b) Torque–RPM relationship. (c, d) Power–RPM relationship. (e, f) Heat input–RPM relationship.

As illustrated in Figure 6(a) and (b), torque decreases with increasing RPM and decreasing travel speed. Those curves show gradual change of torque with RPM and travel speeds. Regarding power (Figure 6(c) and (d)), the values increase only slightly with higher rotation speeds and remain relatively constant overall. In certain cases, power may even decrease slightly at RPM values approaching 1500. Importantly, power demonstrates a clear dependence on travel speed, increasing as travel speed rises. For heat input (Figure 6(e) and (f)), the values are primarily influenced by travel speed, decreasing as travel speed increases and remaining largely unaffected by changes in RPM. The heat input appears to plateau around 200 $J m m^{- 1}$ , as tool failure tends to occur below 200 $J m m^{- 1}$ .

The calculated torque values, derived from the fitted model, enable quantitative analysis of defect formation. Torque, power, and heat input are categorised according to defect type, as illustrated in Figure 7. As shown in Figure 7(a) and (b), the torque data for defect-free welds has a wide distribution range. However, the torque of different types of defect tends to localise in different ranges, especially when only one type of defect occurs. For example, the torque for welds with cavity defect and the tunnel defect has barely overlapped with each other for both AA2024 and AA7075. For AA2024, the torque value $14 N m$ can differentiate between the cavity and tunnel defects. The torque for welds with only tunnel defects is always higher than welds with only cavities. The welds with both cavities and tunnel defect show the intermediate torque value range between them. For AA7075, the torque of welds with only cavities is always lower than $13.5 N m$ and the torque of welds with only tunnel defect/incipient tunnel is always higher than $13.5 N m$ . The welds with both cavities and tunnel defect/incipient tunnel show the intermediate torque value range between them. Welds with surface grooves have similar torque value range as welds with only tunnel defects. While torque distinguishes defect types, it cannot separate defective welds from sound ones. However, when the power input based on the tool torque is compared for different welds, it offers a clearer boundary between defect-free and defective welds.

Figure 7.

For AA2024 and AA7075: (a, b) torque–defect relationship; (c, d) power–fefect relationship; (e, f) heat input–defect relationship.

For the power-defects relationship shown in Figure 7(c) and (d), defect-free welds generally exhibit lower power values compared to those with defects. For AA2024, a power input below 1000 W is both necessary and sufficient to achieve defect-free welds. For AA7075, a power input below 1050 W is generally necessary and sufficient for defect-free welds, with the only exception being cavity defects. However, no distinct ranges of power values are observed among different defect types. These findings highlight the significance of power in defect formation, indicating that lower power is preferable for achieving defect-free welds.

As for the heat input, although it is difficult to separate welds with different defects or defect-free, the heat input is higher for welds without defects than those with defects. For AA2024, heat input higher than 700 $J m m^{- 1}$ and lower than 1200 $J m m^{- 1}$ will guarantee defect-free welds. For AA7075, 1000 $J m m^{- 1}$ can prevent most of the defects with the exception of cavity formation. These results show the importance of the energy in the defect formation process. Lower power and higher heat input are desired to achieve defect-free welds.

Discussion

This study focusses on void defects in FSW. Experimental results show that different aluminium alloys have distinct process maps and defect-free regions, mainly due to differences in hot deformation behaviour. Thus, at identical rotation and travel speeds, internal variables such as temperature and flow stress vary, affecting weld quality.

For AA2024 and AA7075, similar cavities form near the top of the stir zone under high rotational speed and low travel speed-conditions that also produce higher temperatures (see Figure 4). These cavities likely result from eutectic melting and constitutional liquation around large precipitates, which alters material flow and promotes cavity formation near the shoulder. Increasing travel speed further destabilises flow, worsening cavities. Tunnel defects can occur independently or alongside cavities, indicating different formation mechanisms. In AA5052, a non-precipitate-hardening alloy, only singular cavities are observed (Figure 3(b)). These form at the interface between the shoulder- and pin-affected zones, likely due to insufficient material flow. Unlike AA2024 and AA7075, AA5052 does not exhibit both cavities and tunnel/surface groove defects simultaneously, suggesting a shared origin in material flow instability. With higher travel speeds, cavities may evolve into surface grooves or tunnel defects.

For all three alloys, defect-free welds require a homologous temperature ( $T / T_{solidus}$ ) above 0.7, consistent with previous studies.^4,7 However, high temperature alone is not sufficient, as defects especially cavities can still occur above this threshold (see Figure 4). For AA2024 and AA7075, temperatures near or above the eutectic promote flow instability and cavity formation. Even in AA5052, defects are observed at high temperatures. Validation with literatures of this conclusion is also included in the Supplementary Materials Section 4.

Torque decreases with increasing RPM due to reduced material flow stress at higher temperatures, consistent with previous studies.^20–22 Torque increases with travel speed, and the curves for different travel speeds are nearly parallel. Welds with only cavities consistently show lower torque than those with tunnel or surface groove defects, indicating that cavities form under lower shear stress and higher temperature conditions. Welds with surface grooves have similar torque value range as welds with only tunnel defects, which indicates they might have similar formation conditions.

Regarding power (Figure 6(c) and (d)), the data indicates a slight increase with increasing rotational speed at lower RPMs, followed by stabilisation or a slight decrease as the rotation speed approaches 1500 RPM. This trend is consistent with findings reported by Long et al.²¹ and Choi et al.²³ It can be attributed to the self-balancing effect between torque and RPM, where an increase in RPM leads to a decrease in torque, resulting in a relatively constant power output. At higher rotational speeds approaching 1500 RPM, the material flow stress may decrease more substantially due to the occurrence of constitutional liquation, resulting in a reduction in torque. This phenomenon likely explains the observed plateau or slight decrease in power at elevated RPM. Notably, power also depends highly on travel speed, increasing as travel speed rises. Keeping the RPM constant, the torque increases with travel speed, leading to an increase in power. This observation is consistent with the findings of Choi et al.,²³ who also reported a strong correlation between power and travel speed.

The observed power trends in the experiments are further supported by the analytical model for FSW proposed by Patricio et al.²⁴ Based on the model, the power calculated from the torque is:

\hat{P} = \hat{M} ω

(7)

where

\hat{P}

is the power,

\hat{M}

is the total torque from the pin and shoulder and

ω

is the angular velocity of the tool. Using the equations from the article, the power can be expressed as:

\hat{P} = 2 π \frac{k_{0} Δ T_{0}}{η K_{0} (Pe)} d {1 + \frac{f}{3} \frac{a}{d} [{(\frac{b}{a})}^{3} - 1]}

(8)

Pe = \frac{V a}{2 α_{0}}

(9)

where

k_{0}

is the thermal conductivity,

Δ T_{0}

is a constant temperature related to melting temperature of the material,

η

is the total heat efficiency of the process (can be considered constant for specific materials),

a

is the pin radius,

b

is the shoulder radius,

d

is the thickness of welded plates,

f

is a torque factor related with plunge force,

V

is the travel speed and

α_{0}

is the thermal diffusivity of the plate. Those parameters can be considered as constant for specific materials and welding configurations. Therefore, the power primarily depends on

K_{0} (Pe)

. For the same material and tool, the power is mainly a function of the welding travel speed. This aligns with the experimental results shown in Figure 6(c) and (d): the power does not change significantly with higher RPM, but it does increase with higher travel speed.

As indicated in the power analysis, a threshold power of approximately 1000 W for AA2024 and 1050 W for AA7075 was identified. This observation suggests that power applied by the tool may serve as a qualitative indicator of welding difficulty. This conclusion can be further supported by data from literatures, which are included in the Supplementary Materials Section 4. Conceptually, this can be understood by considering the function of the tool during FSW: regardless of whether weld defects are present, the tool must shear and displace the material ahead of it to maintain forward motion. Higher power output implies a greater energy requirement per unit time to sustain this movement, reflecting increased resistance. Conversely, when lower power is sufficient to facilitate tool advancement, it indicates a less demanding process, which is generally associated with improved weld quality and defect minimisation. Therefore, lower power input is favourable for achieving defect-free welds. Nonetheless, this interpretation remains qualitative; further quantitative investigation is necessary to establish a predictive relationship between power input and defect formation.

For heat input (Figure 6(e) and (f)), values decrease with increasing travel speed and are largely unaffected by RPM. The most significant drop of HI occurs at lowest travel speeds, after which heat input gradually plateaus around 200 $J m m^{- 1}$ . This trend aligns with the analytical model by Patricio et al.²⁴ and is consistent with previous studies.^20,25 However, a distinct drop followed by a rise in power and heat input with increasing RPM is observed, particularly at the 50 mm/min. For AA2024 in Figure 6(a), torque decreases sharply once rotation speed exceeds 650 RPM as the temperature rises above the eutectic point in Figure S3(a), leading to a drop in power and heat input. Beyond 1000 RPM, torque stabilises likely also due to constitutional liquation and the continued increase in RPM then drives power and heat input to rise again. This mechanism explains the observed drop and rise behaviour in Figure 6(c) to (f) and is consistent across both AA2024 and AA7075. A more detailed, quantitative analysis of the combined effects of RPM and travel speed on torque would provide deeper insight into power and heat input variations in FSW, which is not sufficiently addressed in the literature.

Overall, the findings from this study provide key insights into void defect formation in FSW of aluminium alloys. Distinct process maps for each alloy underscore the influence of material properties and process conditions. The proposed torque estimation method enables real-time monitoring using spindle power data. Identified thresholds for temperature, torque, power, and heat input offer practical guidelines for defect-free welding. Among these, power and torque are most effective: maintaining power below the threshold prevents void defects, while torque distinguishes defect types. This approach is adaptable to different alloys and setups with minimal calibration. Furthermore, adopting a power-based control mode may help ensure the process remains within the defect-free thresholds. Future work will focus on further understanding the mechanisms behind defect formation and build analytical models for FSW defect assessment.

Conclusions

Bead-on-plate FSW process envelopes of three types of aluminium alloys AA2024, AA5052 and AA7075 in 5 mm plate were investigated by conducting 159 welds with measurement of temperature and spindle power. For each alloy and conditions tested, the optimal parameters to achieve defect-free welds are shown in Figure 2. A novel methodology for estimating torque from spindle motor power is developed and validated against direct torque measurements. A fitting model is established to correlate spindle power with output torque from the tool. This approach enables real-time, online monitoring of torque during the FSW process. The major outcome and conclusions on defect formation conditions are as follows.

1. Only void defects including cavities, tunnel (including incipient tunnel) and surface grooves were observed. Cavities predominantly form in the high-RPM region of the defect map, while tunnel defects are typically located in the lower right area with low RPM and high travel speed. These defect types may also overlap in certain cases. For AA2024 and AA7075, distributed cavities are identified, while singular cavity near on the advancing side (AS) is observed for AA5052.

2. Several process conditions are shown to be necessary or sufficient for defect-free welds. Firstly, high temperature conditions ( $> 0.7 T_{solidus}$ ) are necessary but not sufficient to produce defect-free welds for all alloys tested. Secondly, a threshold of torque values ( $14 N m$ for AA2024 and $13.5 N m$ for AA7075) can be applied to differentiate welds exhibiting only cavities from those with only tunnel/incipient tunnel, surface groove defects, or their combinations. Welds with only tunnel/incipient tunnel defects, only surface grooves, or their combinations consistently exhibit higher torque values than those with only cavities, while welds containing both cavities and tunnel defects display intermediate torque values. It suggests that torque can be used as a reliable indicator to distinguish between cavities and tunnel/surface groove defects. However, torque measurements alone do not distinguish between defective and defect-free welds. Finally, low torque-based power input below 1000 W is both necessary and sufficient for AA2024 to achieve defect-free welds under the tested conditions. For AA7075, a power input below 1050 W is generally necessary and sufficient for defect-free welds, with only one observed exception. This suggests that power input calculated from torque is a reliable indicator for detecting void defects. High heat input values ( $> 1000 J m m^{- 1}$ ) will largely result in defect-free welds with several exceptions of cavities defect.

These findings highlight the potential of using power input and torque value as practical metrics for monitoring welding quality and ranges of parameters for defect-free welds are suggested. Using the technique described in this article and the suggested thresholds values, real-time prediction of defects is possible.

Supplemental Material

sj-pdf-1-stw-10.1177_13621718251413024 - Supplemental material for Process envelope and void defect analysis in FSW of AA2024, AA5052 and AA7075 aluminium alloys

Supplemental material, sj-pdf-1-stw-10.1177_13621718251413024 for Process envelope and void defect analysis in FSW of AA2024, AA5052 and AA7075 aluminium alloys by Xinrui Liu, Adrian P. Gerlich and Patricio F. Mendez in Science and Technology of Welding and Joining

Footnotes

Acknowledgements

The authors gratefully acknowledge the support from Dr. Gentry Wood from Apollo Machine and Welding Ltd. The FSW tools used in the experimental procedures were manufactured at Apollo Machine and Welding Ltd.

ORCID iD

Xinrui Liu

Author contributions

Xinrui Liu did Conceptualisation, Methodology, Investigation, Writing – Original Draft. Adrian P. Gerlich did Writing – Review & Editing, Resources. Patricio F. Mendez did Writing – Review & Editing, Supervision, Resources, Funding acquisition.

Funding

The authors disclosed receipt of the following financial support for the research, authorship and/or publication of this article: This work was supported by Canadian Welding Bureau (CWB) Welding Foundation (RES0056120) and Natural Sciences and Engineering Research Council of Canada (NSERC) (RES0069951).

Declaration of conflicting interest

The authors declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.

Supplemental material

Supplemental materials for this article are available online.

References

Kim

Fujii

Tsumura

, et al. Three defect types in friction stir welding of aluminum die casting alloy. Mater Sci Eng A 2006; 415: 250–254.

Lakshminarayanan

Malarvizhi

Balasubramanian

. Developing friction stir welding window for AA2219 aluminium alloy. Trans Nonferrous Met Soc China 2011; 21: 2339–2347.

Qian

Sun

, et al. An analytical model to optimize rotation speed and travel speed of friction stir welding for defect-free joints. Scr Mater 2013; 68: 175–178.

Ambrosio

Aldanondo

Wagner

, et al. A semi-empirical model for peak temperature estimation in friction stir welding of aluminium alloys. Sci Technol Weld Joining 2022; 27: 491–500.

Mukherjee

DebRoy

. Conditions for void formation in friction stir welding from machine learning. npj Comput Mater 2019; 5: 1–8.

Shi

Chen

Yang

, et al. Thermal-fluid-structure coupling analysis of void defect in friction stir welding. Int J Mech Sci 2023; 241: 107969.

Abboud

Dubourg

Racineux

, et al. Experimental methodology to identify optimal friction stir welding parameters based on temperature measurement. J Manuf Mater Process 2024; 8: 137.

Guan

Cui

Liang

, et al. The response of force characteristic to weld-forming process in friction stir welding assisted by machine learning. Int J Mech Sci 2023; 253: 108409.

Pei

. Shear localization modeling of friction stir welding and applications in process window estimation. Ph.D. Thesis, 2017.

10.

Dialami

Cervera

Chiumenti

. Defect formation and material flow in friction stir welding. Eur J Mech A/Solids 2020; 80: 103912.

11.

Das

Bag

Pal

. A finite element model for surface and volumetric defects in the FSW process using a coupled Eulerian–Lagrangian approach. Sci Technol Weld Joining 2021; 26: 412–419.

12.

Chowdhury

Chen

Bhole

, et al. Effect of pin tool thread orientation on fatigue strength of friction stir welded AZ31B-H24 Mg butt joints. Procedia Eng 2010; 2: 825–833.

13.

Zhang

Cao

Larose

, et al. Review of tools for friction stir welding and processing. Can Metall Q 2012; 51: 250–261.

14.

Mills

. Recommended values of thermophysical properties for selected commercial alloys, chapter Al-2024-T4. Cambridge: Woodhead Publishing, 2002.

15.

Davis

. ASM specialty handbook – aluminum and aluminum alloys, chapter 30.11 2036 (2.6Cu–0.45Mg–0.25Mn). Materials Park, Ohio: ASM International, 1993.

16.

Mehta

Chatterjee

. Monitoring torque and traverse force in friction stir welding from input electrical signatures of driving motors. Sci Technol Weld Joining 2013; 18: 191–197.

17.

Pittner

, et al. Simultaneous measurement of tool torque, traverse force and axial force in friction stir welding. J Manuf Process 2013; 15: 495–500.

18.

Roshandel

Mahmoudi

Kahourzade

, et al. Efficiency maps of electrical machines: a tutorial review. IEEE Trans Ind Appl 2023; 59: 1263–1272.

19.

Lee

Kim

Nam

, et al. Loss distribution of three-phase induction motor fed by pulsewidth-modulated inverter. IEEE Trans Magn 2004; 40: 762–765.

20.

Pew

Nelson

Sorensen

. Torque based weld power model for friction stir welding. Sci Technol Weld Joining 2007; 12: 341–347.

21.

Long

Tang

Reynolds

. Process response parameter relationships in aluminium alloy friction stir welds. Sci Technol Weld Joining 2007; 12: 311–317.

22.

Cui

Chen

Robson

. A model relating tool torque and its associated power and specific energy to rotation and forward speeds during friction stir welding/processing. Int J Mach Tools Manuf 2010; 50: 1023–1030.

23.

Choi

Morrow

Pfefferkorn

, et al. Welding parameter maps to help select power and energy consumption of friction stir welding. J Manuf Process 2018; 33: 35–42.

24.

Mendez

Tello

Lienert

. Scaling of coupled heat transfer and plastic deformation around the pin in friction stir welding. Acta Mater 2010; 58: 6012–6026.

25.

Ambrosio

Wagner

Garnier

, et al. Influence of welding parameters on the microstructure, thermal fields and defect formation in AA7075-T6 friction stir welds. Weld World 2020; 64: 773–784.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

3.43 MB

0.00 MB