Influence of Wavy Bending on Microstructure and Mechanical Properties of a Rolled AZ31 Sheet

Abstract

In the present work, cross-wavy bending at room temperature was carried out to tailor the microstructure and stretch formability of rolled AZ31 sheets. Wavy bending processing generates profuse {10–12} twins and a tilt basal texture. Subsequent recrystallization annealing causes grain coarsening and enhances the intensity of twin-orientation. The combined use of wavy bending and annealing can maintain high tensile ductility and remarkably enhances the stretch formability of rolled AZ31 sheet. It can be mainly attributed to the non-basal texture in the wavy bent sheet which increases the thinning capability during in-plane tension.

1. Introduction

Mg and its alloys are currently receiving widespread attention, especially in the automobile industry, due to the low density and high specific strength [1,2]. However, the low stretch formability at room temperature largely limits the widespread application of Mg alloy sheets [3,4]. It is well known that conventional rolling and extrusion are very easy to produce strong basal texture in Mg alloy sheets. This is also the main cause of poor stretch formability in wrought Mg alloy sheets. Thus, it is important to develop new plastic-processing techniques to weaken or change the texture of Mg alloy sheets [3,5].

Asymmetry rolling [6,7] and asymmetry extrusion [8] have been developed to generate a tilt basal texture in the Mg alloy sheet. Severe plastic-processing techniques (e.g., torsion-extrusion [9] and equal channel angular extrusion [10] etc.) can also re-orientate the basal texture of Mg alloy plates. Moreover, some simple plastic deformation at room temperature can also be used to tailor texture, e.g., continuous bending [11,12], shear deformation [13] and in-plane compression [14,15] etc. These deformation ways are simple and low-cost, and thus are an important supplement to plastic-processing techniques.

For these simple deformation ways, {10–12} twins play an important role in texture control and improvement of stretch formability owing to its large misorientation (86.3°) with matrix [14,16,17]. Twin-orientation depends on twin variant selection during deformation [18]. Previous reports indicated that continuous bending, in-plane shear and in-plane compression favorably activate one twin variant in each grain and generate single twin-texture [11,12,13,14,15,19]. It has been reported that change of strain path can promote the activation of multiple twin variations to further randomize twin-texture [20]. However, it is difficult to do this through the methods mentioned above. In the present work, a wavy bending method is used to tailor texture and improve stretch formability of rolled Mg alloy sheet. It shows multi-pass wavy bending can generate multiple twin variants in each grain and a tilt basal texture in the rolled AZ31 sheet. In this paper, microstructure evolution and twinning behavior of AZ31 sheet during wavy bending and subsequent annealing were investigated. Their influences on mechanical properties and stretch formability were discussed.

2. Experimental Procedure

Initial material is a commercial hot-rolled AZ31 sheet with a thickness of 1mm. Before wavy bending, the as-rolled AZ31 sheet was subjected to annealing treatment at 300 °C for 1 h. As-rolled and annealed sheet was denoted as the AR sheet. The AR sheet was cut into the samples with a gauge size of 100 mm (RD) × 100 mm (TD) × 1 mm (ND). Here, RD, ND and TD are the rolling direction, normal direction and transverse direction of the rolled plate, respectively. Subsequently, AR samples were processed by wavy bending at room temperature. Figure 1 shows the abridged general view of wavy bending. The peak-peak separation of the waves was about 20 mm and the peak-valley distance was approximately 8 mm in the vertical direction. Bending experiments were carried out at room temperature with a speed of 2 mm/min in the RD-TD plane. The displacement along the ND during bending experiments at room temperature is 2 mm for each pass. DEFORM-3D software (V6.0, Scientific Forming Technology Corporation, Columbus, OH, USA) was employed to examine wavy bending processes that were encoded on the basis of the rigid-viscoplastic finite element method. The samples were bent for three passes with a followed 45° rotation after each pass, as shown in Figure 1b–d. There was not straightening carried out until three bending passes were finished. Effective strain contour plots of deformed sheets were simulated by finite element software (Figure 1b–d). It is found that the effective strain is inhomogeneous in each pass. With rotating the sheet pass by pass, the distribution of effective strain becomes more homogeneous after 3 passes. The wave-shaped sheets were flatted by compression (1 mm/min) at room temperature on a mechanical test machine. Some wavy bent samples were annealed at 300 °C for 1 h. The wavy bent sample and subsequent annealed sample are denoted as WB sample and WBA sample, respectively.

Dog-bone shaped tension specimens with nominal gage dimensions of 8 mm × 3 mm × 1 mm were prepared for tension test. Mechanical tests were then carried out on a LD26.105 material test machine (LiShi (Shanghai) Instruments Co., Ltd., Shanghai, China) at a constant strain rate of 1 × 10⁻³ s⁻¹ (using a testing standard of ASTM E8/E8M). The mechanical properties were obtained from the average values of triplicate tests for each condition. The formability of the sheets was investigated by measuring the Erichsen value (IE). Rectangular sheets with a size of 45 mm × 45 mm were cut from the middle location of the bent sheets (see Figure 1d) and were used on a HET800 Erichsen test machine (Chongqing Weishite Electromechanical Equipment Co., Ltd., Chongqing, China) The diameter of the hemispherical punch is 20 mm, the blank force is 10 KN and the punch speed is 1 mm/min. The deformability in the thickness direction during stretch formability was evaluated by measuring the Lankford value (r-value), which a ratio of width strain to thickness strain during tension. Some tensile tests were interrupted at 10% plastic strain to obtain the r-value. The microstructure of the alloys was examined by electron backscatter diffraction (EBSD, Oxford AZtech Max2, Oxford Instruments, London, UK) analysis using an HKL Channel 5 System (Oxford system equipped in a JSM-6610, JEOL, Tokyo, Japan).

3. Results and Discussion

Figure 2 shows the EBSD maps of various samples. The AR sheet had an equiaxial grain structure with an average grain size of approximately 12 μm. The kernel average misorientation maps (KAM) value and its distribution were calculated from the EBSD data to evaluate the dislocation accumulation [21,22]. KAM map indicates a low dislocation accumulation exists in the AR sheet. Moreover, a very small amount of {10–12} twin boundaries can also be observed in the AR sheet. After WB treatment at room temperature, profuse twin lamellae can be found in the AZ31 sheet. It was confirmed that almost all twins were {10–12} extension twins. Thus, the misorientation angle of near 86° obviously increased after WB treatment. Recrystallization annealing can remove twins and dislocations in WB sample and generate an equiaxial grain structure. Moreover, annealing also causes a slight grain coarsening (approximately 24 μm). Severe grain coarsening via annealing has been widely found in the AZ31 alloys with initial {10–12} twins. It can be attributed to the low store energy in the twinned AZ31 alloys. Subsequent annealing induces thermally activated boundary migration of {10–12} twins, rather than grain nucleation [23].

KAM maps were used to further analyze the dislocation accumulation in the matrix. As shown in Figure 2, WB processing can increase the KAM value. The distribution of KAM value and the average KAM value were calculated and shown in Figure 3. It shows the average KAM value increased from 0.55° to 0.89° by WB processing. Subsequent annealing reduces the KAM value of the WB sample to 0.53° which is close to that of the AR sample. It indicates that the WBA sample contains a completely recrystallized microstructure. This result is further confirmed by the misorientation angle distribution. It shows that both the AR sample and the WBA sample had a low frequency of low angle boundaries. It is also interesting that WBA processing remarkably increases the frequency of the misorientation angle ranging from 60° to 90°.

Figure 4 shows the basal pole figures of various samples. The frequency distribution of texture orientation plotted against the tilt angle between the ND and the c-axes of grains is illustrated in Figure 4d. AR sheet exhibits a strong basal texture. The distribution of c-axis exhibits a slight dispersion from ND towards RD. This is a typical rolling texture for Mg alloys. The c-axes of the grains mainly gather at ψ [0–30°] with a frequency of above 70%. WB processing slightly reduces the pole intensity and significantly decreases the frequency of ψ [0–30°] from 72.7% to 59.5%. Although most of the grains still gather at ψ [0–30°] in the WB sample, the basal pole slightly deviated from the ND direction (approx. 11 deg.). Moreover, WB processing increases the frequency of ψ (60–90°] from 3% to 11.9%. After recrystallization annealing, the tilt basal texture component is further weakened and the texture component with ψ (60–90°] is enhanced. For WBA sample, the frequency of ψ [0–30°] is 42.9% and the frequency of ψ (60–90°] is 29.8%. It is found that both WB and WBA treatments exhibit little influence on the frequency of ψ (30–60°] and the orientation of the new texture component in the WBA sample is close to that of {10–12} twins [20]. Thus, it is considered that the generation of new texture components in the WBA sample could be related to the {10–12} twins in the WB sample.

Parent grains and twin lamellae were extracted to draw the {0001} pole figures, as shown in Figure 5. It shows that the texture of parent grains is like that of the overall microstructure. However, the c-axes of {10–12} twin lamellae exhibit a more disperse distribution than that of parent grains. The c-axes of most {10–12} twins are perpendicular to the ND direction and have a disperse distribution from the RD to the TD. To reveal the orientation distribution of twins, some typical grains were shown in Figure 6. The {0001} pole figures correspond to the orientation of parent grains and twins. In this area, the parent grains have a close orientation (i.e., c-axis//ND), as shown in Figure 6a. However, {10–12} twins in the WB sample exhibit a very uneven distribution. In some grains, no twin is activated. In those grains with {10–12} twins, the orientation of {10–12} twins exhibits a disperse distribution. The features of twins in some typical grains are further analyzed. In grain P₁, only one twin variant (T₁₁) is activated. The T₁₁ has an orientation close to the RD in {0001} pole figure, as shown in Figure 6b. Some grains contain multiple twin variants, e.g., grain P₂. Grain P₂ contains three twin variants, as shown in Figure 6c. T₂₂ and T₂₃ have an orientation close to TD, while T₂₁ has an orientation close to RD. Moreover, T₂₂ and T₂₃ have a misorientation of 8.5° and T₂₂ and T₂₁ have a misorientation of 59.5°. It infers that T₂₂ and T₂₃ are from a twin variant pair with a misorientation angle of ~7.4°. However, T₂₂/T₂₃ and T₂₁ are from different twin variant pairs with a misorientation angle of ~60° [24]. In some grains, {10–12} twins can be re-twinned, resulting in the formation of {10–12}-{10–12} twins, e.g., grain P_3, as shown in Figure 6d. Twin T₃₁ with c-axis close to RD is the primary {10–12} twin formed in grain P₃. Twin T₃₂ with c-axis close to TD is the secondary {10–12} twin formed inside twin T₃₁. For grain T₄, both multiple twin variants (e.g., T₄₁, T₄₂, T₄₃ and T₄₅) and {10–12}-{10–12} twin (T₄₄) can be observed, which leads to disperse distribution of twin-orientation, as shown in Figure 6e. It shows both c-axes of twins and secondary twins have a distribution from RD to TD. Thus, the increase in the frequency of ψ (60–90°) in the WB sample can be attributed to the generation of profuse {10–12} twins during WB processing. Moreover, the formation of multiple twin variants and secondary twins is responsible for the disperse distribution of twin-orientation.

For the bending deformation, the outer region is under tension while the inner region is under compression and the stress axis parallels bending direction [25]. It is well known that {10–12} twinning can be favorably activated when the compression along the bending direction of the sheet, and are suppressed when then tension along the bending direction [18]. Thus, {10–12} twins can be formed in the compressed region of the bent sheet. For wavy bending, the distribution of strain on the sheet surface is non-uniform, as shown in Figure 7. It will be responsible for the non-uniform distribution of twins. Figure 7 shows that compressive strain and tensile strain alternate on the sheet surface during wavy bending. During wavy bending, the compressed regions can favorably activate the {10–12} twins and the c-axes of twins gather at compressive direction [25]. Thus, it infers that twins with c-axis//RD texture, c-axis//45° texture and c-axis//TD texture could be formed during 1st, 2nd and 3rd wavy bending, respectively. Moreover, it also found that the deformed regions of three passes exhibit a large amount of overlap. It makes the microstructure evolution more complex during wavy bending. During wavy bending, twin-growth, detwinning, twin-twin interaction and twin-dislocation interaction all may occur [24,26,27,28,29]. Typically, the overlap of compressed regions during 1st pass and 3rd pass wavy bending can promote the activation of secondary {10–12} twins [24,30]. The twins with c-axis//RD texture via 1st pass wavy bending have a favorable orientation for {10–12} twinning when compressed along the TD during 3rd pass wavy bending [31]. As shown in Figure 6, the secondary {10–12} twins usually have an orientation whose c-axis is close to the TD direction and form in primary {10–12} twins with c-axis//RD texture. In short, the formation of multiple twin variants and secondary twins is closely related to the change of loading path during WB processing. Based on the EBSD data, the area fraction of {10–12} twins in WB sample is low (approximately 20%). Moreover, the orientation of twins via WB processing has a very disperse distribution. Thus, {10–12} twins can remarkably change the orientation of the lattice, but exhibits little influence on the overall texture of the WB sample.

Annealing treatment slightly weakens the tilt basal texture and enhances the pole intensity of twin-texture, as shown in Figure 4c,d. In the previous reports, {10–12} twins were quite stable during annealing and the annealing at 300 °C cannot remove the twin structure [32]. However, in the present work, annealing at 300 °C induces a complete static recrystallization and removes almost all twins. In fact, the thermal stability of twins is dependent on the store energy [33]. Figure 3 shows that the wavy bending increases the KAM value. It is considered that repeated wavy bending increases the store energy to reduce the thermal stability of twins. Moreover, the twin–twin interaction owing to the activation of multiple twin variants can also promote static recrystallization [32,34]. The influence of annealing on texture is dependent on twin size and dislocation stored for AZ31 alloys with initial twins [33,35]. In the present work, most twins are smaller than their parent grains. It has been reported that when the stored energy is low, the narrow twin lamellae will be consumed by a large adjacent matrix during recrystallization annealing [35]. It indicates that high store energy in the WB sample could remain the texture of twins with small size after recrystallization annealing [33]. Enhancement of twin-texture after annealing could be attributed to the growth of new grains with twin-orientation.

The typical true stress-strain curves are shown in Figure 8 and the detail mechanical properties are listed in

Table 1. Yield strength (YS), peak strength (PS) and uniform elongation (UE) of various samples.

Mechanical Properties	YS (MPa)			PS (MPa)			UE
	RD	45°	TD	RD	45°	TD	RD	45°	TD
AR	131 ± 3	152 ± 7	177 ± 5	310 ± 2	290 ± 3	309 ± 5	0.21 ± 0.5	0.18 ± 0.2	0.17 ± 0.3
WB	152 ± 5	136 ± 4	94 ± 6	307 ± 3	302 ± 4	294 ± 5	0.13 ± 0.3	0.13 ± 0.3	0.12 ± 0.6
WBA	99 ± 5	111 ± 3	128 ± 4	294 ± 2	277 ± 3	298 ± 2	0.22 ± 0.5	0.18 ± 0.3	0.18 ± 0.2

. AR sheet exhibits a certain degree of yield anisotropy. Yield strength in TD direction is higher than those in the RD and 45° direction. It can be attributed to the slight dispersion of basal texture towards the RD, as shown in Figure 4a. Figure 9 shows that the average Schmid factor (SF) for basal slip with the lowest CRSS for AZ31 alloys under various loading directions. It indicates that SF for basal slip under tension along the RD is larger than those under other loading directions for the AR sample. WB processing slightly increases the yield strength in RD direction from 131 MPa to 152 MPa, but reduces those in 45° and TD direction. Especially, yield strength of TD direction is decreased by 83 MPa. WB processing can introduce dislocations and twin boundaries which can all generate hardening effect on yield strength. The reduction in yield strength can be attributed to the texture change via WB processing. As shown in Figure 9, WB processing increases the SF of basal slip for loading along the 45° and TD direction. Moreover, tension along the TD has a deformation feature of detwinning [36], as shown in Figure 8b. As shown in Figure 6, c-axes of {10–12} twins have a disperse distribution from the RD to the TD. It is considered that the activation of detwinning at the beginning of deformation plays a key role in the low yield strength along the TD direction [37]. However, obvious detwinning behavior has not been found in the tension curves along the RD and 45° direction. It is known that the twins with c-axis//TD texture are mainly formed in the compressed region during the 3rd pass (i.e., final pass). In this case, the reverse tension can easily activate the detwinning [38]. However, {10–12} twins with c-axis//RD and c-axis//45° texture suffer twin-twin and twin-dislocation interaction during 2nd pass and 3rd pass wavy bending. These could suppress the detwinning [39,40].

Annealing treatment induces a completely recrystallized microstructure and further weakens texture intensity, as shown in Figure 4c. Randomization of texture in the WBA sample further increases the SF of basal slip for in-plane tension, as shown in Figure 9. Thus, annealing can remarkably reduce the yield strength and increase tensile ductility. However, it is interesting to find that annealing increases the yield strength along the TD. As shown in Figure 2c, annealing retains twin-texture, however, it removes twin structure. The disappearance of the {10–12} twin structure leads to a transition from a detwinning predominant deformation to a slip predominant one during tension along the TD [41]. This change of deformation mechanism accounts for the annealing hardening phenomenon. Finally, the WBA sample exhibits a lower yield strength and comparable uniform elongation with the AR sample. Low yield strength in the WBA sample can be attributed to the grain coarsening and the formation of tilt basal texture and twin-texture.

The Erichsen test was carried out to evaluate the stretch formability of annealed AZ31 sheet. As shown in Table 1, the WBA sample has a comparable uniform elongation with the AR sample. However, WBA processing largely enhances the stretch formability. Erichsen value is increased from 2.82 mm to 3.56 mm by WBA processing, as shown in Figure 10. It has been reported that stretch formability of the AZ31 sheet is largely dependent on the ductility and deformability in the thickness direction during in-plane tension [3].

Table 2. r-value of various samples.

Samples	AR			WBA
	RD	45°	TD	RD	45°	TD
r-value	2.9 ± 0.3	3.5 ± 0.5	3.7 ± 0.5	1.7 ± 0.2	2.3 ± 0.4	1.5 ± 0.3

shows that WBA processing reduces the r-values along various tensile directions. Thus, it is considered that high Erichsen value in the WBA sample could be attributed to the enhancement of deformability in the thickness direction. For the rolled AZ31 sheet with strong basal texture, basal slip and {10–12} twinning with low CRSS is hard to activate during in-plane tension [42]. The prismatic slip is usually the dominated deformation mode to accommodate the length strain and width strain, but cannot generate thickness strain. The thickness strain only can be coordinated by pyramidal slip and contraction twinning with the very large CRSS, resulting in a very poor thinning capacity [14,15]. This is also the main reason for the low stretch formability of Mg alloys with strong basal texture. As shown in Figure 2 and Figure 4, WBA processing slightly increases the grain size and generates non-basal texture components. Formation of non-basal texture can increase the contribution of basal slip and {10–12} twinning to plastic deformation [43]. Moreover, prismatic slip, basal slip and {10–12} twinning in the grains with non-basal texture can accommodate the thickness strain and enhances the thinning capability during in-plane tension [14]. Moreover, Chino et al. [44] also reported that the grain coarsening could increase the thinning capability by promoting the twinning. For these reasons, WBA processing can also enhance the stretch formability of the rolled AZ31 sheet.

Present work has revealed that wavy bending processing can effectively tailor the texture of the rolled AZ31 sheet and improve the stretch formability. The advantages of this method are simple operation and low cost. However, poor microstructure uniformity in the WB sheet is the main problem restricting this method. Moreover, increasing the active amount of {10–12} twins during wavy bending can enhance the effect of texture control. In further work, controlling microstructure uniformity and twinning activation by optimizing the wave bending process will be interesting topics.

4. Conclusions

(1) Wavy bending processing generates profuse {10–12} twins and a tilt basal texture. The c-axes of {10–12} twins in the WB sample exhibit a disperse distribution from the RD to TD. It can be mainly attributed to the activation of multiple twin variants and generation of secondary {10–12} twins during wavy bending processing.

(2) Annealing at 300 °C induces a completely recrystallized microstructure and causes a slight coarsening of the grains in the wavy bent sample. After annealing, all twin lamellae are removed, but twin-orientation is enhanced.

(3) WBA sample has a comparable tensile ductility and lower yield strength compared with the AR sample. Moreover, WBA processing can largely enhance the stretch formability of the AZ31 sheet. Erichsen value is increased by 26% by WBA processing.