Process Stability and Reproducibility of the Dieless Drawing Process for AZ31 Magnesium Wires

Magnesium (Mg)-based wires are in the focus of interest for numerous applications like micro-forming technologies or medical engineering. Manufacturing thin Mg-based wires is widely realized by applying a conventional multiple pass cold wire drawing process. This requires a complex manufacturing schedule of multiple passes with intermediate heat treatments to overcome work hardening, because of the cold forming process. Especially Mg and its alloys are known for their rather low formability at room temperature associated with the hexagonal close-packed lattice structure. The dieless drawing process uses local heating to initialize a localized plastic zone under an external tensile load to achieve higher reductions in diameter in a single wire drawing pass. It can therefore present a solution for a more efficient warm manufacturing process of Mg-based wires. In this study, the stability of the steady state material flow during a dieless wire drawing process and its reproducibility was investigated. For this purpose, a variation of process parameters was selected and wire manufacturing was carried out using magnesium alloy AZ31. A single and double dieless drawing process was applied. Additionally, a conventional cold wire drawing process including a die with the same forming schedule was executed as a benchmark experiment. The results of this study show, that the dieless drawing process is not only a stable process after reaching the steady state, but it is also a reproducible and accurately adjustable process. Moreover, the dieless drawing process maintains the property profile of the starting material to a large extend.


Introduction
The dieless wire drawing (DD) process was first introduced in 1969 [1] and experimentally analyzed in 1975 [2]. It is described as a non-contact process for manufacturing bars, tubes and wires by using local heating to initialize a localized plastic deformation zone. Early studies with steel wires describe the potential to continuously produce wires, but also the sensitivity of the DD process to its parameter settings and therefore its overall stability [3,4,5]. Later studies confirm the feasibility of the DD process for Magnesium (Mg)-based wires and describe the existence of a nonstable and a steady state phase [6,7,8]. Milenin et al. [9] propose a multiple pass DD process to improve the stability of deformation and increase the workability. Nevertheless, only few investigations of the steady state exist and an assessment of reproducibility and process stability have not yet been investigated.
Besides lightweight applications, Mg-based products such as wires are in the focus for application in medical engineering as resorbable implant materials with good biocompatibility, whereby the mechanical properties and the surface quality of the wires are subject to special production requirements [10]. A conventional cold wire drawing (CCD) process can produce Mg-based wires that have appropriate surface quality due to the drawing die, but the material can be contaminated by abrasion or the use of lubricants, and large reductions in diameter per drawing pass are not possible [11]. Due to the hexagonal lattice structure and the associated rather low ductility, Mg-based wires are produced in multiple pass approaches with intermediate heat treatment or preferably at higher process temperatures -e.g. with the DD process.

Dieless Wire Drawing
In Fig. 1, the principle of the DD process is shown. The wire moves into the process with the feeding velocity V0 and is drawn with the drawing velocity V1. In the heating zone 1, an induction coil heats the wire locally. Subsequently, the wire is passively cooled to room temperature by the surrounding air of cooling zone 2. The ratio between the process velocities V0/V1 in combination with the local heating initializes a localized plastic deformation zone. Under the globally applied tensile load F, resulting from the difference in velocities V0/V1, the reduction of the cross sectional area (csa) from an initial A0 to A1 after the drawing pass is initiated. Following the assumption of a constant volume, the law of continuity can be applied as [12]: Based on Eq. 1, the reduction R in csa can then be determined by the ratios of csa, as well as of process velocities, before and after the induction coil as follows [13]: This enables two ways of calculating R: From the measurement of diameters to determine the csa or the set process velocities, respectively. However, this is only valid for a single pass DD. For multiple passes, the overall reduction ratio Rov can be determined by: for n multiple passes ( ∈ ℕ).
Using Eq. 3 on Eq. 2, the achieved reduction ratio Ra from the csa determination and the set reduction ratio Rs taken from the velocity ratio are calculated for a single pass i and for an overall reduction ratio for multiple passes as follows: If the feeding velocity V0 remains the same over all n passes, Eq. 7 can be simplified to:

Materials and Experiments
As feedstock material, a wire of Mg alloy AZ31 with a thickness of 1 mm is used for all experiments. A cast ingot of the alloy was machined into cylindrical shape for the hot wire extrusion process. A direct extrusion process was applied with a billet and container temperature of 350 °C and a ram velocity of 0.2 mm/s to receive wires, using a die with 4 nozzles positioned equidistant from the center. Details on the setup can be obtained from earlier work in Nienaber et al. [14]. The flexible experimental setup for dieless wire drawing has been described in detail in an earlier work in Braatz et al. [15] and is extended by an automatic force value export and a demountable die holder for the cold drawing at room temperature experiments of the presented study. This enables multiple process setups including DD and CCD.
For the experiments, a high frequency induction generator with a maximum power capacity of 10 kW and a frequency of 1383 kHz is used, generating temperatures in the range from 200 to 400 °C by variating the input power of the induction heating coil from 1.1 to 1.6 kW. For measuring the process temperature, a radiation based thermal imaging (IF) camera is used with an accuracy of ± 10 °C. A single torque sensor on the drawing side servo motor measures the acting torques in Nm during the process, which are then converted into process forces in N and exported for every second of process time. For calculating the achieved reduction ratio Ra of the wire in Eq. 4 and 5, a minimum of five measuring points for the ingoing diameter d0 before the coil and for the outgoing diameter d1 several measuring points (every 5 to 10 cm) after the coil are taken. During the experiments for the DD process, the feeding velocity V0 is fixed to 10 mm/s and the drawing velocity V1 is set to 11.25 or 12.5 mm/s to achieve reductions in csa of 10 % or 20 %, respectively. For the CCD process, the drawing velocity is set to 10 mm/s and a die is chosen from the die set that is closest to the targeted reduction ratio R. In the setup the wires are preloaded at 15 ± 2 N, the duration of every experiment equals a 5 m sample size and each set of process parameters is applied three times.
The mechanical properties are investigated by carrying out tensile tests at ambient air temperature using an universal testing machine Zwick05 with a maximum load of 5 kN. At a constant initial strain rate of 10 -3 s -1 , stress-strain diagrams and the related stress and strain properties are obtained. Standard metallographic procedures with picric acid etching are used for measuring the grain size with an Olympus light microscope and a Keyence laser scanning microscope is used for surface analysis.

Results and Discussion
Process stability. The process stability of the DD process itself is addressed by continuous analysis of the process parameters and along the length of the wire, for each experiment.  Fig. 2 shows the variation in process forces, process temperature and the reduction ratio over the length of the wire for two representative DD processes with 10 % and 20 % reduction ratio. The gaps in the force and reduction ratio curves are due to calibration paint reference points for the emissivity calibration of the IF camera. For all DD experiments, it is observed that a steady state occurs after approx. 17 sec. For all single pass experiments with a csa reduction of 10 %, a maximum standard derivation of ± 0.014 mm in diameter and ± 3.1 N in force is measured. For all experiments with 20 % csa reduction, a maximum standard derivation of ± 0.003 mm in diameter and ± 4.5 N is obtained. The results have to be compared to a variation in diameter of ± 0.002 mm of the extruded wire, which is used as feedstock material, and a variation in process forces for CCD experiments of ± 2.35 N. Thus, the variation in diameter after a DD process is only slightly increased. This indicates an overall stable process during the steady state with even higher homogeneity regarding diameters for experiments with 20 % reduction ratio.
Process reproducibility. Secondly, the reproducibility of the DD process is investigated in comparison to the CCD process. Each experiment with fixed parameters is repeated three times to receive mean values and standard derivations of the process parameters for each parameter setting. In Fig. 3, for each single pass DD experiment, the achieved reduction ratio Ra, calculated from Eq. 4, are compared with the set reduction ratio Rs, calculated according to Eq. 6. Focusing on each parameter set, similar results are obtained with respect to the achieved reduction ratio and variation in diameter. Temperature variations of up to 220 °C for the same reduction ratio do not appear to affect a change in the reduction ratio or the magnitude of the deviation. Taken all experiments at Rs = 11.11 %, the mean achieved reduction is 10.14 ± 2.3 %. Moreover, for all experiments at Rs = 20 %, the mean achieved reduction is 19.43 ± 0.45 %. Comparing experiments of the same parameter sets, the DD process shows the same mean values of reduction ratio, as well as force and temperature values with similar standard derivations. This indicates similar results for DD process experiments with the same process parameter settings and therefore a good reproducibility.

392
Achievements and Trends in Material Forming

Key Engineering Materials Vol. 926
For calculating the flow stress, the average process force F and the determined final csa A1 are used. Fig. 5 shows the flow stress σf over the process temperature for all single and multiple (double) passes DD experiments for different reduction ratios. In general, a tendency of lower process flow stresses for higher temperatures can be observed. Higher reductions also lead to an expected increase of the process forces or the flow stresses, respectively, which is consistent for the higher temperature range applied in this study. It is interesting to note that the flow stresses for a single drawn wire and the flow stresses for a pre-drawn wire (shown in red) do not differ much from each other for similar reductions and temperatures. This suggests that the property profile of the wires is not significantly affected after the first drawing pass of the DD process.
Mechanical properties. The mechanical properties of the drawn wires are studied by comparing their stress-strain diagrams and their related mechanical properties. In Fig. 6, the tensile test results for the starting material and the single pass DD experiments are plotted. For each parameter set (each carried out three times), at least three tensile tests were carried out and the mean tensile yield strength (TYS), ultimate tensile strength (UTS) and fracture strains are determined. One representative stress-strain-curve for each parameter set is shown. The average values and standard deviations for TYS, UTS and fracture strain for one parameter set show little variation within the relatively high standard deviations achieved. The fracture strains appear only slightly reduced compared to the original extruded wire before drawing. The impression of decreasing TYS and UTS with increasing input power fits to the concurrent increase of the processing temperature, but is not clear due to the high standard deviations. Still, even with a temperature variation of up to 220 °C for one reduction ratio, it is remarkable that there is no distinct impact on the resulting mechanical property profile, which remains rather stable during the drawing passes.

394
Achievements and Trends in Material Forming  For Fig. 7, representative stress-strain-curves for every parameter set for DD and CCD experiments are plotted. In the comparison of the DD and the CCD experiments in general, it stands out, that the DD experiments have slightly increasing TYS, mostly unaffected UTS and slightly decreasing fracture strain with increasing reduction ratio compared to the starting material, while the CCD results have significantly higher TYS and UTS and substantially lower fracture strain. These changes in TYS, UTS and fracture strain are well related to the work hardening of the material during the drawing passes of the CCD process. Focusing on the CCD experiments, the fracture strain is decreasing slightly and the TYS is increasing slightly over all reduction ratios, but the UTS increases significantly. Generally, the higher the reduction ratio, the higher the UTS. However, achieving the same overall reduction by adding a second pass, e.g. reaching 20 % by applying two passes of 10 % + 10 %, results in lower UTS. This shows that the higher the reduction in csa in a single pass, the more pronounced the work-hardening effects in the AZ31 wires, and that with multiple passes the overall reduction is still decisive. However, UTS for 10 % + 10 % remains visibly below a single pass of 20%. For 10 % + 20 % and 20 % + 10 %, the former, with a higher R value in the second pass, is more ductile than the latter.
It is interesting to note that although the elongation at fracture remains far below the values of the starting material, the various drawing steps do not significantly change it. The results also show that only one single pass with a reduction ratio of 10 % significantly changes the mechanical properties of the wire compared to the starting material. Focusing on the DD experiments, the TYS shows no significant changes, while the UTS changes slightly and the fracture strain varies with different reductions in csa. The UTS changes by a magnitude of 20 MPa and the fracture strain varies from 9.5 % (R = 20 % + 20%) to 14.9 % (R = 10 %). Generally, the lower the overall reduction, the less the difference in fracture strain compared to the 17 % of the extruded AZ31. Contrary to the CCD results, the DD curves show no significant changes for achieving 20 % reduction in a single pass or with multiple passes, whereas the order of reduction for Rs = 28.9% (10 % + 20 % or 20 % + 10 %) seems to play a role. This demonstrates that the DD process maintains the property profile of the extruded wire to a large scale while reducing its diameter.
Wire characterization. Surface and microstructure analysis of the wires are shown by means of metallographic analysis for measuring the surface roughness and the grain size.  and Sz (maximum magnitude) for the extruded and the drawn wires for single pass 10 % and 20 % and multiple pass 20 % + 20 % reduction ratio. After the DD process, the surfaces of the drawn wires are slightly rougher then the starting material, whereas the CCD wires have a significant smoother surface than the DD wires and the extruded wire, demonstrating the positive influence of a form giving tool (die).   9 shows the arithmetic surface roughness Sa over the process temperature for the extruded, the CCD, and DD drawn wires. Following the findings from Fig. 8, the CCD process generally improves the surface roughness significantly while the DD process influences the surface roughness of the wires only slightly. Differences in Sa values and its deviations for different reductions in csa for the CCD may be explained by manufacturing inaccuracies of the die. For the DD wires, a tendency of slightly less rougher surfaces for higher temperatures can be observed. Higher reduction ratios seem to have an influence on the surface roughness in general and the order of reduction (e.g. 10 % + 20 % or 20 % + 10 %) seems to play a role in this context, too. This confirms that the die provides a significant surface improvement, but also indicates that the DD process does not significantly degrade the surface quality of the starting material.
As seen in Tab. 2 and Fig 10, the metallographic results show a typical AZ31 microstructure [14], which is completely recrystallized and homogeneous, for the extruded material. The drawing experiments reveal also recrystallized microstructures, which are clearly developed if the reduction R is low at 10 % and the temperature is higher than 300 °C. If the reduction R is higher at 20 %, developing grain structures with serrated boundaries can be revealed at the same temperature, too. This applies also for the double pass experiments. A temperature range of 325 to 350 °C reveals rather homogeneous microstructures with average grain size of 7 to 8 µm. Higher temperature, e.g. in Fig.  10 m), shows increasing grain size, here up to 12 µm. Multiple pass DD experiments seem to very slightly increase the average grain size to 7 to 9 µm. Grain size analysis for CCD drawn wires reveal a deformed version of the initial microstructure of the extruded wires, where the etching impact increases on other obstacles than the grain boundaries due to the work hardened condition of the wires. The average grain size appears unchanged as expected in Fig. 10 b) and 10 c). Higher CCD reductions of the csa are strongly deformed and do not offer the possibility of grain size quantification (Fig. 10 d) -g)).  Figure 10. Micrographs from longitudinal sections for AZ31 wires for Table 2.