Strengthening of wire arc additive manufactured aluminium alloy through interlayer rotary friction processing

To eliminate porosity and refine the grains in aluminium alloys fabricated via wire arc additive manufacturing (WAAM), a novel strengthening method – robotic rotary friction processing (RFP) – was introduced. The feasibility of RFP was assessed by processing the deposited beads under various average peak loads. During RFP, a rotating tool pressed the deposited beads to induce sufficient plastic deformation. After 3.5 kN-RFP, the WAAM densification ratio was improved to >99.5%, and pores were completely closed and eliminated owing to compression deformation and sub-grains bonding. The processed deposits exhibited effective grain refinement and high-density dislocation entanglement, which led to 22.8% and 34.2% improvements in the horizontal and vertical tensile strengths, respectively. Furthermore, the strengthening mechanism of RFP on deposited metals was clarified.


Introduction
Because of their low weight and high specific strength, aluminium alloys have emerged as promising structural materials and are extensively applied to the manufacture of various components in the aerospace industry [1].Recently, the use of wire arc additive manufacturing (WAAM) to fabricate large aluminium alloy components for aerospace applications has attracted significant attention [2].Based on the welding principle, WAAM technology uses an arc heat source to melt the metal wire, and the components are formed through layer-by-layer deposition.This technique reduces the material cost by more than 60% [3] and shortens the manufacturing cycle by more than 50% [4], allowing for the highly efficient manufacturing of large aluminium alloy components.
However, insufficient strength and mechanical anisotropy, attributed to coarse-grained microstructures and porous defects [5], are significant factors restricting the application of WAAM-based aluminium alloys.Dong et al. [6] noted that the microstructures of the deposited metal typically exhibited a columnar grain morphology with 001 and 011 textures owing to the solidification and growth of α-Al in the direction of heat dissipation.Columnar grains grow and coarsen because of the reciprocal thermal cycling of WAAM [7], which reduces the microstructural uniformity and tensile strength of the components.Regarding CONTACT Shengfu Yu yushengfuhust@hotmail.comSupplemental data for this article can be accessed online at https://doi.org/10.1080/13621718.2023.2247271.
porosity, Boeira et al. [8] reported that the rapid solidification of the aluminium alloy molten pool can hinder supersaturated hydrogen from escaping at the end of solidification and becoming trapped between layers to form pores. Nezhadfar et al. [9] showed that stress is likely to concentrate around the pores and contribute to microcrack nucleation, weakening the vertical strength and fatigue properties of the components.Many valuable strengthening methods based on interlayer plastic processing have been proposed to refine grains and reduce defects in WAAM aluminium alloy components [10].These methods break the coarse grains and close the pores by inducing plastic deformation in the deposited layers via external loading.Zhang et al. [11] used interlayer forging for the first time to break the coarse columnar grains of WAAM-deposited layers, resulting in fine-grained strengthening.Gu et al. [12] employed interlayer rolling to effectively eliminate porosity defects in 2-series and 5-series aluminium alloys processed via WAAM, noting that pore closure was due to the effective strain concentration around the pores.Fang et al. [13] adopted layer-by-layer hammering with a 50% vertical deformation to reduce the porosity of a WAAM aluminium alloy by more than 70% while obtaining an equiaxed fine-grained microstructure.However, most of these methods rely on heavy loading devices to ensure the processing load, which increases the cost of the WAAM equipment to some extent [14].Moreover, their motion mechanisms are mostly limited to 3-axes platforms [13], which have many limitations when applied to complex processing paths.
The robotic-based processing method can to some extent reduce the equipment costs, as the robot's displacement could provide processing load, eliminating the need for large specialised loading platforms [15].And the high degree of freedom of robot also offers more potentials for processing along complex paths [16].Typical applications include the interlayer friction-stir processing (FSP).He et al. [17] developed a robotic-based WAAM + FSP system, which facilitated the sufficient plastic flow of deposits by interlayer FSP to form aluminium-alloy components with low porosity.Wei et al. [18] also reported that the dynamic recrystallisation caused by interlayer FSP facilitated grain refinement in aluminium-alloy deposits.The loading force required for FSP is usually within 20 kN, which is lower than that of 30-45 kN for interlayer rolling [12], meaning a reduction in the loading-device cost.In addition, rotary friction processing (RFP) is also a latent strengthening approach to WAAM.Unlike FSP, this method involves the use of a robotic-driven rotating tool to rub and press the surface of deposits [19].The pressing force, coupled with frictional heat, induces plastic deformation of the beads, grain fragmentation within the microstructure, and pore closure.Li et al. [20] were the first to report weldingoriented RFP method.They adopted a milling machine to drive the tool rotation to press and rub the weld beads along with the welding, and observed significant columnar grain fragmentation [21].Chen et al. [22] subsequently introduced impact loads to the RFP to induce sufficient compressive deformation of the welding joints, refine the grain size, and control internal porosity defects.Yu et al. [23] performed RFP on stainless-steel deposits, and reported the elimination of internal fusion defects by uniform plastic deformation.Interestingly, as reported by Yu et al. [24], the maximum load for RFP is only within 4 kN, lower than that of interlayer rolling and FSP and more conducive to decreasing the loading-equipment cost, while the high scalability of RFP makes it possible to integrate into robots for complex-path processing [20].Nonetheless, currently there have been few studies on robotic-based RFP.More importantly, the strengthening mechanisms of RFP on aluminium-alloy deposits, such as porosity elimination and microstructural refinement, have also not been reported for WAAM.
This paper presents a novel robotic-based RFP strengthening method for WAAM deposits with the aim of refining grains and eliminating defects in aluminium-alloy components.Specifically, interlayer RFP was applied to WAAM aluminium alloy walls under different average peak loads (APLs) to (1) investigate the pore morphology, grain microstructure, and second-phase evolution, (2) test the horizontal and vertical mechanical properties of the blocks before and after RFP, and (3) elucidate the mechanism of pore elimination and microstructure strengthening.

Experimental procedure
The developed robotic WAAM-RFP system is illustrated in Figure 1(a).It comprises a WAAM unit, RFP unit, and two-axis positioning platform.The WAAM unit employed a Fronius TPS 4000 power supply for the fused deposition of the aluminium alloy wire, whereas the RFP unit used a FAEMAT electrospindle to clamp and rotate the self-developed tool.After the deposition of each layer, the rotating tool applied pressure to the deposited layer to perform the RFP step.The interlayer RFP paths for single-pass and multi-pass deposited beads, including the processing and travelling paths, are displayed in Figure 1(b) and c, respectively.The tool bottom ends were sequentially lowered (performing pressing), raised, and brought forward along the processing and travelling paths.This cycle was repeated continuously (Figure 1(b)).The length of a single travelling path was set to 25% of the tool end diameter to ensure partial overlap between adjacent compression areas.The multi-passes are horizontally offset single passes which are then joined in an 'S' scan, as shown in Figure 1(c), with an offset lap of 25% of the tool end diameter.Throughout the displacement process, the RFP tool was maintained at a rotational speed of 2300 r min -1 .Since the path planning could ensure the overlap of adjacent compression areas, two adjacent travelling paths could be set to be at any angle, meaning that RFP has better advantages in processing along complex paths, such as high-curvature curves, oscillation curves, and multi-acute-angles paths.
The deposited material was an ER5B06 aluminium alloy wire, and the substrate was 1060 pure aluminium.To perform WAAM, the 0881 CMT-P expert database process waveform was utilised; the parameters are detailed in Table S1. Figure 1(d) shows the pressure-load curve measured during a single downward rotation of the RFP.Upon contact of the tool with the deposited layer and subsequent downward pressing, the load increased, leading to compression and deformation of the deposited layer.When the deposited layer underwent 68% and 77% compressive deformation in height direction, the APL stabilised at 1.8 and 3.5 kN, respectively.The 1.8 and 3.5 kN-RFP was used to strengthen the walls along a single-pass path; the 3.5 kN-RFP was used to strengthen the blocks along a multi-pass path.
A 20 mm × 10 mm × 8 mm specimen was extracted from the central area of the single-pass forming wall to analyse the pore defects, as depicted in Figure 1(e).The pore morphology was observed using a GXIAN TZ optical microscope (OM) system, and the roundness of the pores and their diameters were calculated under equal-area conditions using ImagePro software.The block density was measured using a G-DenPyc automated densitometer, and the densification ratio was determined using the standard density of 2.65 g cm -3 of a commercial 5B06 aluminium alloy.The deposited metal microstructures were observed using a Quanta 650 scanning electron microscope (SEM) and analysed using electron backscatter diffraction (EBSD).A Zwick ZHV30 Vickers-hardness tester was used to test the hardness gradient on the wall cross sections (100 g load and 15 s packing pressure time).According to the DIN EN ISO 6892-1, horizontal and vertical tensile-property tests were performed on the multi-pass formed blocks using a Shimadzu AG-IC universal testing machine, as shown in Figure 1(f).

Porosity
OM images of the porosity distribution in the deposits in different states are shown in Figure 2. Interlayer RFP was effective in decreasing porosity.As illustrated in Figure 2(a), in the non-RFP state, the specimen has large pores with dimensions above 50 μm, resulting in a densification ratio (DR) of only 88.34%.Figure 2(b,c) show the pore distribution after interlayer RFP at APL of 1.8 and 3.5 kN, respectively.The pores number reduced significantly, and the DR increased from 88.34% to 98.65% and 99.72% after 1.8 and 3.5 kN-RFP, respectively.In addition, the pores were compressed into a flattened shape as the RFP load increased, with only a few fine pores remaining at 3.5 kN (Figure 2(c)).This could be attributed to the combination of rotational shear forces and vertical downforce during RFP, leading to compressive deformation of the pores with the deposited beads [24].A higher load facilitated the transfer of more strain to the metal, rendering the pores more susceptible to compression.
Figure 3 shows the average pore diameters and roundness.The average pore size decreased from 67.85 to 12.41 μm, and the roundness decreased from 76.29% to 44.31% after interlayer RFP at 3.5 kN. Figure S1 further depicts the frequency distributions of  diameter and roundness of individual pore.Most of the pores in the non-RFP beads were in the size range of 60-70 μm.The proportion of pores smaller than 10 μm increased significantly after the interlayer RFP.This shift was accompanied by an increase in the degree of pore deformation, indicating that most pores underwent effective compressive deformation and elimination.This is consistent with the findings of Gu et al. [12], who reported that the continuous plastic deformation of the deposited metal concentrated the local effective strain around the pore, facilitating the compression of the pore to a flattened state and eventually closing it completely.
Moreover, a narrow residual gap may remain after the pores are closed, which must be connected to eliminate porosity completely [24].The dynamic reversion/recrystallisation of the deposited metal during RFP may be a potential method for eliminating residual gaps.The EBSD analysis of the grain distribution around the pores with and without 3.5 kN-RFP is displayed in Figure 4(a,b), respectively.Figure 4(a) demonstrates the existence of fully recrystallised grains around the typical large pores in the non-RFP metal, which can be attributed to the static recrystallisation of the original α-Al dendrites because of repeated heating during continuous WAAM deposition, leading to the formation of irregular polygonal recrystallised grains.In contrast, subgrains were always present around the small deformed pores, as shown in Figure 4(b).This is because of the constant plastic deformation of the deposited metal during RFP, which induces dynamic reversion to generate low-angle boundaries (LABs) of the subgrains, which could further transform into dynamically recrystallised grains by absorbing dislocations.As the interfaces of a pore close to form a residual narrow gap, continuous nucleation, growth, and linking of subgrains around it may occur, eventually leading to the complete elimination of pores.

Microstructure
Figure 5(a-c) shows SEM images of the deposited metal microstructures in different states.From Figure 5(a), the coarse and white second phases are present in the non-RFP deposits, and their composition are shown in Figure 5(d).The laminal phase (region a) has a higher Mg content, whereas the irregularly shaped phase (region b) contains Mn and Fe aggregates, the former being β-Al 3 Mg 2 and the latter being Al 6 (Mn, Fe) with solid-soluble Fe, as evidenced by the Al-Mg and Al-Mn phase diagrams.After interlayer RFP, the original coarse second phase was significantly refined and more dispersed, as illustrated in Figure 5(b,c).With an increase in the RFP load, the refinement of the second phase became more pronounced, and after 3.5 kN interlayer RFP, the second phase was spot-like.As both β-Al 3 Mg 2 and Al 6 (Mn, Fe) are brittle, hard phases and have strain coefficients that differ from those of the α-Al matrix, when the deposited metal was plasticly deformed during RFP, local strains accumulated at the second-phase interface, leading to the second-phase breakage into small fragments.
The inverse pole figures (IPFs) and pole figures (PFs) of the grain structures in different states are shown in Figure 6(a-c).The original coarse α-Al grains were broken and refined owing to plastic deformation during RFP, and no significant texture was observed in the deposited metal.The maximum uniform density multiple (UDM) value of 5.18 in the deposited state decreased to 3.37 and 2.61 after RFP with 1.8 and 3.5 kN, respectively, indicating an increase in grain fragmentation with random orientation.Figure 6(d-f) shows the frequency distributions of the misorientations, where the fraction of LAB with a misorientation of 0°< θ < 15°increased from 17% to 62% and 69%, respectively, after being subjected to 1.8 and 3.5 kN-RFP, respectively.This increase is attributed to the dynamic reversion/recrystallisation of the deposited metal.During plastic deformation, frictional heat and load action induce high strain energies and dislocations.Strain-driven slippage and climbing of dislocations caused the formation of fine subgrains with LABs in the fragmented grains, which continued to absorb dislocations to form recrystallised grains.Ultimately, grain fragmentation and dynamic reversion/recrystallisation induce grain refinement in the deposited metal.
Figure 7 presents a comparison of dislocations in the deposits with and without 3.5 kN-RFP, showing fewer dislocations in the non-RFP beads owing to the residual strain from solidification and formation, as shown in Figure 7(a).Many dislocations were observed after RFP, with some dislocation lines bypassing the second phase (Figure 7(b,c)).As the total dislocation level is linearly related to the geometrically necessary dislocation (GND), the GND density, ρ GND , can be used to   quantitatively characterise the dislocation level [13] as follows: where θ and μ are the average misorientation and scan step size measured via EBSD, respectively, and b is the Burgers vector.Processing of the EBSD data in Figure 6 reveals that ρ GND increases from 4.42 × 10 12 m -2 in the deposited state to 1.63 × 10 13 m -2 and 1.95 × 10 13 m -2 after RFP at 1.8 and 3.5 kN, respectively.As the deposits undergo severe plastic deformation during RFP, strain transfer promotes dislocation movement and accelerates proliferation.The fragmentation of the original grain boundaries and formation of subgrain boundaries led to the accumulation of local strain, resulting in the formation of dislocation arrays at the multivariable boundaries and dislocation networks within the grain, as shown in Figure 7(d).The increase in the dislocation density due to the RFP is conducive to improving the strength of the deposited metal through dislocation strengthening [14].

Mechanical properties
The hardness gradients of the wall cross sections formed under various conditions are shown in Figure 8(a).The initial hardness before RFP was in the range of 82-97 HV0.1, with an average value of 89.4 HV0.1.Following RFP at 1.8 and 3.5 kN, the average hardness values increased to 107.7 and 114.1 HV0.1, respectively.Owing to the grain refinement and second-phase fragmentation caused by the RFP, both the grain boundaries and fine second phase impeded dislocation movement under the test load, thereby increasing the hardness of the deposited metal.The ultimate tensile strength (UTS), yield strength (YS), and elongation (EI) before and after RFP at 3.5 kN are shown in Figure 8(b).The mechanical properties of the deposited metal were notably improved, with the horizontal (H) and vertical (V) UTS increasing from 303.6 and 274.2 MPa to 372.7 and 368.1 MPa, respectively, corresponding to increases of 22.8% and 34.2%, respectively.The EI of the RFP metal shows decreasing trend, which may be due to the hindrance of dislocation movement caused by high-density dislocation pile-up, weakening the plastic deformability of the deposits [13].Moreover, mechanical anisotropy was eliminated after RFP strengthening, and the difference between the UTS and YS in different directions was within 10 MPa. Figure 8(c,d) show the vertical fracture morphologies before and after RFP at 3.5 kN, respectively.Dimples were observed in all the fractures, demonstrating ductile fracture characteristics.No significant pores containing dendritic projections were observed after RFP, and numerous tear edges were present, demonstrating the high strength of the deposits.

Discussion on the strengthening mechanism
The mechanical properties of WAAM aluminium alloy components are significantly affected by pore defects.
During the interlayer RFP, the surface layer of the deposited metal is the first to be subjected to frictional heat, caused by the shearing forces and downward pressure from the rotating tool [24].The resulting thermal and strain input are transmitted to the subsurface, causing compressive deformation of metals and their internal pores.Hydrogen pores are the most induced defects in aluminium alloy-deposited metals, as proven by Hauser et al. [5].Owing to the pressure destabilisation of the hydrogen molecules contained within the pores, hydrogen dissociates as [H] when the pores are compressed multi-directionally [12].The diffusion/dissolution of [H] into the solid metal constantly reduced the internal pressure within the pores, facilitating their closure.Subsequently, the dynamic reversion/recrystallisation process leads to the nucleation, growth, and joining of subgrains around closed pores, ultimately eliminating them.The presence of pores reduced the unit load-bearing area of the deposited metal and degraded its mechanical properties.As reported by Qi et al. [9], stress concentrations accumulate around pores and contribute to microcrack nucleation, whereas dispersed pores shorten the crack extension distance, causing premature fractures.The deposits DR improves owing to the significant reduction in porosity from RFP, which contributes to an increase in the unit bearing area and improvement in the mechanical properties of the formed block.Furthermore, the improved microstructure of the deposited metal contributed to the enhancement of mechanical properties.Owing to the plastic deformation caused by the RFP, the coarse β-Al 3 Mg 2 , Al 6 (Mn, Fe) second phase, and α-Al grains were fragmented.The small, hard, and brittle second phases and fine grains can be strengthened through the Orowan mechanism and high grain-boundary fraction, respectively, impeding dislocation movement to produce the second phase and fine grain strengthening, respectively.The simultaneous increase in the dislocations number caused by the plastic strain also resulted in dislocation strengthening.If a linear relationship exists between the multiple strengthening mechanisms, then the overall yield stress change, σ s , of the RFP deposits can be calculated as follows [13]: where σ o , σ g , and σ d are the contributions from Orowan strengthening of the second phase, finegrained strengthening, and dislocation strengthening, respectively.Based on the Orowan equation, σ o can be expressed as follows [14]:  where M is the Taylor coefficient (M = 3); G is the shearing modulus (taken as 26 GPa for aluminium alloys); b is the Burgers vector (0.3 nm); and λ 0 and r are the second-phase spacing and dimensions, respectively.σ o is inversely proportional to the size of the second phase.However, despite the fragmentation of second phases after RFP, their size remained larger than 100 nm.By quantitatively analysing on SEM images, the average level of secondphase spacing and size in the 3.5 kN-RFP metal were ∼ 6.64 and ∼ 1.38 μm, respectively, resulting in σ o being less than 1 MPa.This suggests that second-phase refinement is not the main mechanism of microstructure strengthening.
The contributions of the fine-grained and dislocation strengthening mechanisms can be expressed based on the Hall-Petch and the Taylor formulas, respectively: where k is a constant describing the relationship between the stress and grain size of the aluminium alloy (0.78 MPa mm -1/2 ); α is a constant describing the deformation hardening (taken as 0.24) [13]; and d and ρ are the mean grain size and GND density, respectively.Based on EBSD data, the average grain size of 3.5 kN-RFP metal was 5.21 μm.The approximate calculations indicated that σ g and σ d were higher than 10 and 70 MPa, respectively, after RFP at 3.5 kN.Meanwhile, the YS of the deposited metal increased by more than 80 MPa (Figure 8(b)).Thus, it can be inferred that grain refinement and improved dislocation strengthening are the main microstructural strengthening mechanisms.
While the robotic-based RFP has shown good strengthening effects in the applications of line walls and blocks, it should be noted that some complex structures with overhanging features may be faced in practical applications.In this condition, the internal stress may not be transmitted well due to the lack of support below.The potential solution is to better apply loads by adjusting the RFP-tool posture and the components position.As the main motion mechanism is a robot, the RFP-tool pose can be more conveniently changed on the sixth axis of the robot, and it is also expected to avoid interference in complex-structural processing by multi-axes path planning.Further researches are needed on the synergistic regulation of RFP pose and multi-axes processing path to achieve the RFP strengthening for typical complex structures.

Conclusions
The main findings of this study are summarised as follows: (1) A novel technique for strengthening WAAM aluminium alloy, known as robotic-based interlayer RFP, was developed.This method uses a robotdriven rotating tool to induce the compressive plastic deformation, porosity defects closure, and grain refinement of deposited metals.(2) The elimination of pores was attributed to compressional closure and subgrain association through dynamic reversion/recrystallisation.At an RFP load of 3.5 kN, the densification ratio of the deposited metal reached 99.72%.
( This work has proven the effectiveness of roboticbased RFP on grain refinement and porosity elimination of the WAAM aluminium alloy.The UTS of processed deposits was improved to more than 360 MPa, which exceed the mechanical property requirements of the 5-series aluminium alloy in most industrial applications.This provides an alternative new method for strengthening WAAM components.The developed robotic-based RFP technology is expected to build a low-cost and high-scalability hybrid production line integrated with robotic swarm, achieving flexible and high-quality WAAM for large aluminium-alloy components.

Disclosure statement
No potential conflict of interest was reported by the authors.

Figure 3 .
Figure 3. Average diameter and roundness levels of pores under different conditions.

Figure 8 .
Figure 8.(a) Hardness gradients of wall cross sections in different states.(b) Mechanical properties of deposited metal with and without 3.5 kN-RFP.(c, d) SEM images of tensile fracture morphology: (c) non-RFP and (d) 3.5 kN-RFP.
) RFP causes significant fragmentation of the coarse β-Al 3 Mg 2 , Al 6 (Mn, Fe) second phase and α-Al grains within the deposited metal.The primary microstructural strengthening mechanisms included grain refinement and high-density dislocation strengthening.(4) Porosity elimination, grain refinement, and highdensity dislocations induced by RFP significantly enhanced the mechanical properties of the WAAM aluminium alloy.After RFP at 3.5 kN, the horizontal and vertical UTS values reached 372.7 and 368.1 MPa, respectively, corresponding to improvements of 22.8% and 34.2%, respectively, and the mechanical anisotropy was eliminated.