The Benefit of the Process Combination of Wire Arc Additive Manufacturing (WAAM) and Forming—A Numerical and Experimental Study

Abstract

Wire arc additive manufacturing (WAAM) involves the deposition of weld beads layer by layer using an electric arc energy source. However, during this procedure, the properties of each layer may differ because of unequal thermal distribution, resulting in a difference in microstructure and, therefore, mechanical properties in between the layers. This negative effect can be compensated for by combining WAAM with a subsequent forming process to introduce dynamic recrystallization, which allows a more homogeneous microstructure distribution within the material. This paper investigates numerically and experimentally the hybrid process of combined WAAM and forming of fine-grained mild steel (FGMS) SG3/G4Si (1.5130) to achieve a high degree of recrystallization in all layers of the WAAM-deposited material. Three different possible combinations of WAAM and forming are considered regarding the sequence and setup of the processes to show their influences on the recrystallization behavior. It was found that combining welding and forming allows recrystallization of up to two layers; however, the top layer is not recrystallized. Preliminary simulation results show that this can be resolved by adding a top roller to induce plastic strain after welding, leading to recrystallization in the top layer. The found results promise a certain controllability of the recrystallization behavior.

1. Introduction

Wire arc additive manufacturing (WAAM) offers high deposition rates, high material utilization, and the ability to generate large parts as compared to other manufacturing techniques. The process is based on layer-wise deposition of metal using an electric arc in conjunction with a protective shielding gas (e.g., pure argon) to melt off the wire. Fine-grained mild steel (FGMS) such as SG3/G4Si1 (1.5130) is widely used for steel construction and low-temperature applications and a commonly used material for WAAM. Due to its unique chemical composition and alloy design, the weldability is excellent compared to other steel materials. During welding, a significant change in microstructure occurs caused by rapid inhomogeneous cooling and annealing phenomena. This results from the heating cycles during the welding of multiple beads, leading to undesired microstructure evolution and unsatisfying mechanical properties in the as-welded condition. Usually, post-heat treatment at a higher temperature level is used to cause austenite formation with subsequent slow cooling to obtain a more homogenous microstructure and improved mechanical properties. Recrystallization, another desired effect, occurs when a certain amount of heat is applied to a material with prior plastic strain, leading to a change in the microstructure such as phase fractions, grain size, and grain orientation. Recrystallization is classified as static recrystallization (SRX), characterized by a decoupled plastic strain and heat input—typically observed in cold forming with subsequent heat treatment; meta-dynamic recrystallization (MDRX), characterized by close cycles of plastic strain and heat input—typically observed in multi-step hot forming with longer cooling times between forming steps; and dynamic recrystallization (DRX), characterized by high simultaneous plastic strain and heat input [1]. To improve the WAAM-deposited material, one approach uses the welding heat and applies plastic strain to the deposited material in combination. The principle of this ‘hybrid’ process is to induce recrystallization and, therefore, the desired microstructure evolution during material deposition with repetitive heating cycles.

The basis of the development of the hybrid process is the WeldForming process (Figure 1a), a combination of welding and forming [2]. The thermal energy induced by welding is utilized in the subsequent forming process to initiate microstructural transformation during hot forming. This leads to the introduction of the microstructural effects of crystal recovery and recrystallization (Figure 1b) [3].

Behrens et al. studied the process of laser welding with subsequent hot forming and found that the process combination could improve the mechanical properties [5]. Blohm et al. showed for plasma arc coating and cross wedge rolling an improvement of the coated area with a more homogenous distribution of coating thickness [6]. There has been some preliminary work on the simulation of this hybrid process of WAAM and forming, such as the work done by Zhao et al. on welding with rotation compression to control the performance and shape of the semi-solidified state of the weld pool. Within this work the effect of the surface morphology of the weld bead and altering the stress distribution is investigated [7]. However, this is preliminary work and the results indicate further machining procedures are required. A study by Adams et al. on the inline WeldForming process exhibits validation of the numerical model followed by microstructural analysis [8].

The two general process combinations, 1. forming + AM (forming first) and 2. AM + forming (AM first), each have their own set of advantages. Combination 1, for example, is beneficial if the material has high strength or limited forming behavior or if multistep forming processes with expensive dies have to be performed, e.g., for Ti6Al4V [9], TiAl [10], or nickel super alloys [11], which limit the complexity of a part’s formed geometry. After forming, a near net shape with specific complex geometric features (e.g., fixation points [12]) can be generated using AM that, if necessary, can be machined and does not have to obtain the superior mechanical properties as for the rest of the part. This will reduce tool complexity and costs, as well as the amount of required forming steps. Combination 2 also allows processing of materials that are challenging to form by generating the near net shape geometry by means of AM. After AM, a single subsequent forming step can lead to the final geometry and simultaneously improve the microstructure mechanical properties because of plastic-strain-induced microstructure evolution, as it occurs, for example, for Ti6Al4V forged parts [9]. In any case, forming can either provide a final geometry but requires a semi-final AM-made part before or provide a semi-final part that has to be extended by AM. This means only rather simple forming geometries like cuboid or cylindrical bodies are relevant for such hybrid processes, regardless of the process combination order. So far, no high-complexity hybrid parts have been examined regarding one of the process combinations.

In general, the layer-wise process allows a step-by-step combination of depositing and forming, which can be used to also generate complex parts. To show the basic principle with different forming methods, a setup with a small narrow roller attached to a CNC machine [13] or machine hammering [14] was examined in the work of Martina et al. A grain refinement and texture modification could be obtained and the residual stress was diminished because of induced compressive stresses by the forming operations.

This paper presents the preliminary process optimization of the hybrid WAAM and forming process by means of numerical simulations and experimental work. Three different combinations are considered, varying the sequence and setup of the processes. The effect of compression and rolling procedures on weld beads is considered in terms of the fraction of recrystallization and is then investigated experimentally through hardness measurement and light microscopic analysis.

2. Materials and Methods

Four strategies are used to represent the process combination of welding and forming (see Figure 2) for a single-bead multilayer wall. This means the wall is one bead wide and consists of stacked beads on top of each other.

The first strategy (Figure 2a) includes the process combination of welding and forming a single layer single-bead. This process was examined in [8] and the obtained experimental data was used to develop and validate a simulation model for the simple process combination. The simulation model was calibrated by thermal measurements with thermo couples regarding cooling during welding and forming. In addition, the microstructural evolution during cooling after welding and during hot forming was determined by means of dilatometer experiments and metallographic examinations.

The second strategy (Figure 2b) is considered as a decoupled process and is a simplification of the in situ (target) process. The decoupled process consists of welding four layers, cooling down to room temperature, a subsequent forming process, and, afterwards, welding one additional layer with two variations of the welding speed. Forming at room temperature (cold forming) will cause plastic strain and dislocations. The heat input acts as a heat treatment to enable static recrystallization after the forming step.

The third strategy named the inline or coupled process (see Figure 2c) is examined by FEM (finite element method) simulations. Here, one welding bead is deposited with a direct subsequent forming process by two lateral rolls using the stored welding heat (hot forming). Depending on the distance between the torch and the rolls, the varying cooling time affects the solidification behavior respective of the microstructure evolution of the welded bead. In this case, dynamic recrystallization takes place. The torch-to-rolls distance is not varied in this work but will show the basic effect of this experimental setup. It was set to the critical value to allow the full solidification of the melt pool.

The fourth strategy for the coupled process (see Figure 2d) is also examined by FEM simulations. An additional top roll is mounted behind the two lateral rolls to enable plastic deformation of the top side of the last welded layer, which otherwise may not exceed the required plastic strain for (meta-)dynamic recrystallization as it happens for the third strategy.

Strategies 2–4 are examined numerically. Only strategy 2 (the decoupled process) is additionally examined by experiments to calibrate the simulation model and to prove the expected influence. In terms of plastic strain, the different forming processes of compressing and rolling covered by the FEM simulations are considered to be roughly equivalent.

2.1. Experimental

The WAAM experiment is conducted with a Fronius welding power source 5000 CMT using the CMT process (cold metal transfer). The used welding wire is the fine-grained structural steel SG3/G4Si1 (1.5130) with a diameter of d = 1.2 mm. To protect the melt pool against oxidation and to improve the electric arc behavior, a gas mixture with 82% argon and 18% carbon dioxide is supplied at a gas flow rate of 15 L/min. The stick-out of the wire (distance from workpiece to contact tip) is 15 mm. The weld beads with a length of 150 mm are deposited on a metal sheet made of S235JR with the dimension of 200 × 70 × 10 mm (l × w × h). The interlayer temperature is 150 °C. The chemical composition of the two materials is displayed in

Table 1. Chemical composition of welding wire SG3/G4Si1 (1.5130) and substrate S235JR, Fe balanced.

Chem. Composition [%], Fe Balanced	C	Si	Mn	P	S	N	Cu
Wire G4Si1 (1.5130) [15]	0.08–0.12	0.9–1.1	1.6–1.8	-	-	-	≤0.3
Substrate S235JR [16]	Max. 0.17	-	Max. 1.40	Max. 0.035	Max. 0.035	Max. 0.012	Max. 0.55

.

The metal sheet substrate is not preheated. The welded bead is cooled down by natural convection by the contact with the forming tool. The applied WAAM parameters considering two variations of welding speed are shown in

Table 2. WAAM welding parameters. Standard welding speed is 0.4 m/min, the variation of the welding speed is only realized for the fifth (last) layer.

Wire Feed Rate vf [m/min]	Welding Speed vw [m/min]	Voltage V [V]	Current I [A]	Power P [kW]	Specific Energy Input E [kJ/mm]
3.5	0.2	14.8	110	1.63	0.49
3.5	0.4	14.4	108	1.55	0.23

. The welding speed is determined by the velocity of the welding torch. The induced specific energy input E is proportionally linked to the welding speed.

The forming process for the second strategy (decoupled welding and forming) is realized by compressing with two flat forming tools under cold forming conditions. The forming parameters are shown in

Table 3. Forming parameters to obtain certain plastic strain.

Initial Wall Width b0 [mm]	Resulting Wall Width b1 [mm]	Global Plastic Strain $\overline{\mathsf{\phi }}$ [-]
5.8	4.0	≈0.4

.

The global plastic strain in the weld seams can be calculated using Equation (2).

$$\overline{\mathsf{\phi }}=\ln \left(\frac{{\mathrm{b}}_0}{{\mathrm{b}}_1}\right)$$

(1)

The global plastic strain $\overline{\mathsf{\phi }}$ is determined from the initial wall width b₀ of the weld seams and the remaining width b₁ of the weld seams after the forming process.

Forming by rolling is not considered in the experiments of this work but in the simulation. Only the resulting initial plastic strain will be considered as simplification.

Material testing considers metallographic analysis and hardness measurements to indicate the recrystallization. The welded walls are cut in half by a band saw at the central area perpendicular to the longitudinal direction to obtain the cross-section of the layers (see Figure 3a).

The preparation procedure, metallographic analysis, and hardness measurement are realized as stated in [3]. Hardness measurements are done from the bottom (first layer) to the top (fifth layer) in the center of the cross section at a distance of 0.3 mm for each position (Figure 3b).

Light microscopy was used to determine the recrystallized microstructure percentage using the line-section method according to DIN EN ISO 643 [4].

2.2. Simulations

Simulations are performed using the software tools SIMUFACT welding (V16.0) and SIMUFACT forming (V16.0). The mechanical (determined with upsetting tests, see [3]) and thermal material data calculated using JMATPRO by Sente Software for mild steel are used for SIMUFACT for all simulations. A JMAK material model is used to predict microstructure development and to determine the fraction of recrystallization. The model parameters were determined at temperatures of 900, 1000, and 1100 °C for a plastic strain rate of 1.0 s⁻¹ up to a plastic strain of 0.7 [3].

The welding simulations utilize a 3D coupled thermo-mechanical model. The GOLDAK double ellipsoid is used as a moving heat source. The formula used for heat density distribution, heat source parameters, and the initial and boundary conditions are taken from [3]. The metal sheet is fixed during welding. The real seam geometry was modeled by incorporating the “deactivated element method”. The heat transfer to the environment at room temperature is considered with 50 W/(m²∙K). Heat transfer during welding (e.g., due to arc and shielding gas) is not considered.

The forming simulation utilizes a 3D coupled thermo-mechanical kinematic model, where the welded beads can be deformed freely. The heat transfer coefficient between the table and sheet is set to 2000 W/(m²∙K) and between forming tools and sheet to 3000 W/(m²∙K). Heat generation by forming is considered as well. Friction between the metal sheet and forming tools is considered. According to the conducted experiments, the relevant minimum global plastic strain of approx. 0.4 will be realized by a forming gap of 4 mm.

Only for the rolling process (strategies 3 and 4), welding speed v_w is coupled to the rolling speed v_R after gripping the metal sheet substrate, which can be calculated using the equation:

v_w = v_R = 2 × π × n_R × R

(2)

where n_R is the rotational speed of the rolls and R is the radius. The welding speed is then varied to the desired values by changing the rotational roll speed. The FEM rolling process is simulated with a radius of R_L = 7.5 mm for the large lateral rolls and R_S = 5 mm for the small top roll, which are not heated or cooled. The welding torch distance to the lateral rolls is 5 mm and to the small top roll, 20 mm.

3. Results and Discussion

As mentioned, the first strategy (one single bead, single layer) was tested experimentally and the results were used for the calibration of the simulation model (see Section 2). The results of the other strategies are presented in this section.

Strategy 2

The metallographic analysis for the second strategy (single bead multilayer wall) is shown in Figure 4. The fifth layer shows for both welding speeds of 0.2 (see Figure 4a) and 0.4 m/min (see Figure 4b) the same as-cast microstructure with large fractions of bainite/acicular ferrite and large grains, while only minor ferrite with large grains was developed. For low-alloyed steels with a low carbon content, the formation of bainite with acicular ferrite and polygonal ferrite has already been reported [17,18]. For both welding speeds, the microstructure shows no significant difference, as the cooling rates are quite similar. Due to the lack of initial plastic strain, no recrystallization occurred.

Layer 4 with induced initial plastic strain shows full recrystallization for both welding speeds, which in contrast to layer 5 is characterized by the formation of a more homogenous and finer microstructure. Because of recrystallization, a measurable change in hardness is expected. At layer 4 for a welding speed of 0.2 m/min, the formation of a primarily ferrite/bainite microstructure takes place (see Figure 4c). The microstructure consists of large ferrite and bainite/acicular ferrite grains with a small fraction of perlite. The large amount of ferrite will be responsible for the lower hardness. The corresponding microstructure for a welding speed of 0.4 m/min (see Figure 4d) shows a much finer grain structure of ferrite and bainite. Compared to the welding speed of 0.2 m/min, a higher cooling rate is present. As layer 4 is directly affected by the deposited fifth layer, a higher average temperature by a welding speed of 0.2 m/min leads to recrystallization and grain growth. In comparison, the welding speed of 0.4 m/min shows recrystallization but only smaller grains, indicating a shorter time for grain growth after full recrystallization.

Layer 3, with the largest distance to the heat source (welding torch), in contrast to layers 4 and 5, shows a remarkable influence regarding welding speed. The welding speed of 0.2 m/min shows a recrystallized fine grain structure of ferrite, bainite, and perlite (see Figure 4e). Compared to layer 4 (welding speed of 0.4 m/min), the grain growth was more pronounced, indicating a higher average temperature and a longer persistence at recrystallization temperature. At a faster welding speed of 0.4 m/min, layer 3 was not affected by the heat input as it shows an inhomogeneous microstructure with very large grains of ferrite and small bainite/acicular ferrite and perlite grains (see Figure 4f). The induced deformation is still visible, which indicates that no recrystallization occurred. The reduction in the fraction of bainite compared to layer 5 is attributed to a lower cooling rate.

The initial plastic strain, temperature distribution, and obtained fraction of recrystallization are shown in Figure 5. The global plastic strain (see Figure 5a) exceeds the required magnitude of 0.4, which causes recrystallization.

The faster welding speed of 0.4 m/min shows an almost evenly distributed heat within the bottom layers 1, 2, and 3 (see Figure 5b). In layer 4, the temperature distribution ranges from approx. 850 °C (bottom of the layer) to 1300 °C (top of the layer). Yet, a nearly full recrystallization of the fourth layer was achieved, which indicates that the recrystallization temperature was exceeded (see Figure 5d). The third layer shows only partial recrystallization of an average of 50% ranging from nearly 0% (bottom of the layer) to approx. 80% at the interface to the fourth layer, where the initial strain was 0.4. In this case, the temperature dropped too quickly at the bottom of the layer, which would not provide enough time for recrystallization. There is no recrystallization in the fifth welded layer because it remains undeformed. The two bottom layers (1 and 2) show no recrystallization as well, as the temperature did not exceed the recrystallization temperature.

The slower welding speed of 0.2 m/min shows a higher average temperature and also a nearly evenly distributed heat (see Figure 5c), but only for the layers 1 and 2. Layers 3 and 4 show temperatures of ~850–1300 °C (layer 3) and ~1300–1500 °C (layer 4) within the layer. These higher temperatures cause a larger area of nearly completely recrystallized material covering the third and fourth layer (see Figure 5e). The second layer shows partial recrystallization within a range of 0% (bottom of the layer)) up to 90% (top of the layer). Even the first layer has a relatively evenly distributed fraction of recrystallization of about 30%. The fifth layer also shows no recrystallization because of the lack of the required plastic strain. Therefore, a higher heat input by reducing welding speed leads to a nearly full recrystallization of 2.5 layers, which can be considered for the optimization of the process regarding process parameters.

Recrystallization causes a new microstructure formation, resulting in a softening of the material, which is indicated by a drop in its hardness. The fraction of recrystallization (experimental and simulated) for welding speed 0.2 m/min and measured hardness distribution along the wall height for strategy 2 is shown in Figure 6. The simulated fraction of recrystallization shows good agreement with the experimental results for both welding speeds.

It can be seen that the third and fourth layer at a speed of 0.2 m/min experience nearly full recrystallization due to the high temperature over a longer time. The fourth layer is not completely recrystallized on top of the layer due to partial melting and therefore no remaining critical plastic strain. Layer 2 shows a large gradient of recrystallization. It is visible that the required temperature for recrystallization is not distributed over the whole layer. On the one hand, the top of the layer has a higher temperature level and enough time for recrystallization. On the other hand, the bottom of the layer is cooled too fast, which results in minor recrystallization. Layer 1 remains completely without any recrystallized areas.

The measured and simulated fraction of recrystallization show a significant effect on the hardness. Layer 4 shows nearly complete recrystallization for both welding speeds, which corresponds to a hardness of about 170 HV1. Layer 3 shows nearly complete recrystallization of about 90% for the slow welding speed of 0.2 m/min, leading to a low hardness of about 160 HV1. For 0.4 m/min in layer 3, the hardness rises to 240 HV1 because the fraction of recrystallization drops to about 50%. Layer 2 with a welding speed of 0.2 m/min shows only minor recrystallization of about 50%, indicated by a hardness of 200 HV1. The welding speed of 0.4 m/min at this point does not show any significant recrystallization, which corresponds to a hardness of 250 HV1. The bottom layer 1 has 0% recrystallization for 0.4 m/min and about 20% for 0.2 m/min, which leads to a hardness of 260 HV1 and 230 HV1, respectively. The fifth layer shows no recrystallization but moderate hardness levels. This comes because of the faster cooling directly after welding, causing an as-cast-like microstructure with larger fractions of bainite and acicular ferrite, which has a higher hardness than to ferrite. The effect of recrystallization is not present here.

Strategy 3 and 4

To compare the different forming processes (compressing and rolling), the initial plastic strain induced by forming is simulated. The results are shown in Figure 7. Here, the magnitude of approx. 0.4 as the minimum plastic strain is visible for compressing (strategy 2; see Figure 7a), while the minimum global plastic strain for rolling (strategies 3 and 4) is 0.6 (see Figure 7b). The difference in the magnitude of plastic strain is caused by the different material flows. Due to friction, compressing causes a lower plastic strain as the material flow is inhibited in the middle area. Continuous rolling as an inline process allows material flow in both axial and perpendicular (vertical) directions. This leads to compressive and shear strain close to the surface, resulting in a higher global plastic strain and a more homogeneous plastic strain distribution, which is a benefit of the coupled process.

For the third and fourth strategies, the simulation results are presented in Figure 8. The results for the third strategy (coupled (inline) process (two lateral rolls) for each layer) in Figure 8a show nearly the same temperature distribution as for strategy 2. Therefore, the thermal effects will be equivalent. It can be seen that a moderate fraction of recrystallization with an even distribution could be obtained for the third and fourth layers. Only the top layer 5 shows no significant recrystallization within a large area. The reason is that the two lateral rolls only flatten the side of the wall, which leads to minor recrystallization in a small zone close to the side of the wall—the top area experiences no plastic strain. The bottom layers 1 and 2 do not show any recrystallization either. Here, the time at the recrystallization temperature level was not long enough to initiate recrystallization, and the recrystallization temperature was not reached at all.

The results for the fourth strategy (optimized coupled (inline) process (two lateral rolls + one top roll)) in Figure 8b show a very high fraction of recrystallization for the top layer 5 and a quite homogenous distribution. Only the area at the side of the wall does not show recrystallization, where no plastic strain is induced by the top roll. As the local plastic strain is quite inhomogeneous in layer 5, a large gradient of the fraction of recrystallization is present close to the side of the wall. Layer 4 shows a large recrystallized fraction as well, with a quite homogenous distribution. Layer 3 shows a moderate fraction of recrystallization, that is slightly lower than for strategy 3. This can be explained by the heat transfer from the welded wall to the top roll, which cools the wall down. In this case, this cooling effect reduces the temperature faster, shortening the time required for recrystallization. The distribution of recrystallized fractions for layer 3 is comparable to strategy 3. Layers 1 and 2 of strategy 4 do not recrystallize, as they did in strategy 3, because of the same effect of a too low temperature and time at recrystallization temperature.

Depending on the forming process and its variations regarding tool setup, different plastic strain magnitudes and distributions were achieved, which also affected the fractions of recrystallization. In general, full recrystallization for multiple layers could be obtained. The static recrystallization can also be improved for pre-induced plastic strain by changing the heat input. The combination of welding and forming was proven suitable for adjusting the microstructure evolution, fraction of recrystallization and material hardness. Control and optimization can be realized by altering the parameters of heat input and plastic strain as well as the tool setup to obtain the desired microstructure and mechanical properties, which makes the process combination more viable for more complex applications. Further research is needed to extend the range of optimization because the process combination is affected by multiple process parameters, such as the number and order of process sequences.

For all strategies, the reduction in the waviness of the wall is also a benefit as the geometric accuracy is increased and otherwise required machining can be reduced or prevented.

4. Conclusions

This work focused on the recrystallization behavior, microstructure evolution, and hardness distribution of welded and deformed FGMS, considering four possible process combinations: 1. separated welding and forming of a single bead; 2. separated welding and forming of single bead multilayer wall; 3. coupled welding and forming for each layer excluding top roll; and 4. coupled welding and forming for each layer including the top roll.

It could be shown that:

• Initial plastic strain in four layers and subsequent heat input (welding cycle) cause full static recrystallization but only in the fourth layer.
• Recrystallization and hardness show an inversely proportional interaction. Fully recrystallized material has a hardness of about 160 HV1, whereas non-recrystallized material has a hardness of between 230 HV1 and 260 HV1. The hardness drop is addressed to the formation of a finer homogenous microstructure with an increased fraction of softer ferrite because of recrystallization.
• An increase in heat input by lowering the welding speed affects the range of static recrystallization. A higher heat input leads to higher average temperatures along the cross section, which exceed the critical temperature for a full recrystallization of up to 2.5 layers from the top.
• Coupling welding and forming for each layer allows recrystallization of up to two layers, but the top layer remains mainly unrecrystallized.
• Adding a top roll and inducing plastic strain after welding leads to (meta-dynamic/static) recrystallization in the top layer.
• Optimizing the process (sequence welding/forming, setup, and process parameters) allows recrystallization of multiple layers to improve efficiency.

Future work will consider:

• Examining the effect of different plastic strains and strain distribution on the amount of recrystallized fraction;
• Variating the number of sequence welding/forming and tool setups to determine the major influence of process strategies and to find the optimal number of sequences;
• Examining the influence of pre- and post-heat treatment on welded and formed layers to achieve better control over microstructure development;
• Other materials such as stainless steel and hard facing filler wire with higher ultimate strength;
• Multi-material application to determine the interaction of phase faces on the integrity of the joint and forming behavior.