Keywords

Introduction

When scarce raw material resources, as well as the reduction of CO2 emissions, are the focus of interest, the demands for future-oriented, individual, and economical manufacturing processes are constantly increasing [1]. A weight reduction of 100 kg already leads to fuel savings of 0.35 l/100 km and 8.4 g CO2/km for gasoline engines [2], which is why the use of aluminum alloys and the associated weight reduction is of great importance. Since modern aluminum alloys have excellent specific stiffness and strength, steel parts can be replaced, reducing the weight by up to 40% [3, 4]. Research and development work continues to produce new or improved aluminum alloys, convinced by their outstanding performance potential. In particular, aluminum alloys with high lithium (Li) additions have unique properties superior to conventional aluminum alloys [5]. Recently, Al–Li alloys have thus attracted attention for their use in light construction and aerospace structures due to their impressive material properties, such as low density and high specific stiffness [6,7,8]. One wt.% Li reduces the density of the resulting Al alloy by about 3% and increases Young's modulus by about 6% [8]. Moreover, adding Li to Al results in fine precipitates that improve the strength of Al alloys [9]. However, the main disadvantage of Al–Li alloys is their anisotropic behavior with limited formability. It is due to different solidification phases and grain structure characteristics that occur during permanent mold casting and subsequent thermomechanical treatment [8]. As conventional methods reach their limits, processes for adapting the alloy chemistry and innovative casting and thermomechanical processing methods must be used. For example, the formability of Al–Li alloys is to be improved by combining twin-roll casting and thermomechanical treatment. Twin-roll casting is an energy-saving and efficient innovative process for producing thin strips [10, 11]. The characteristic of this process is the combination of solidification and plastic deformation of the metal in a single unit [12, 13]. This process produces near-net-shape strips directly from the molten metal [10]. In this study, the twin-roll casting of Al–Li alloys using copper shells was represented by numerical simulation using a thermofluid model implemented in ANSYS-Fluent. This simulation-based study investigates the influence of casting speed, casting temperature, strip thickness, and length of strip-forming zone on strip outlet temperature and specific size of deformation zone, thus optimizing the practical manufacturing process. The parameters determined through the simulation were used for the aluminum–lithium strip production, and the microstructure was analyzed on the basis of these parameters.

Numerical Finite Volume Model of the Twin-Roll Casting Process

The numerical finite volume model was developed. The theoretical analysis of the effects of different influencing variables of the twin-roll casting (casting speed, casting temperature, strip thickness, and length of the strip-forming zone) on the strip outlet temperature and the specific size of the deformation zone was performed. This model was implemented by using ANSYS-Fluent software. The processing of fluidic problems with the aid of fluidic simulation is divided into four work steps: preparation (model creation), pre-processing (meshing), solving (defining boundary conditions and solving model equations), and post-processing (graphical processing and presentation of results). The assumption that the melt was evenly distributed over the strip width and supplied to the strip-forming zone was made to accelerate calculations. Therefore, minor temperature and flow velocity changes along the strip width were not considered, and a two-dimensional model was developed.

Preparation

In the preparation step, the model of the twin-roll casting was created. The strip-forming zone and the copper shells were represented in a simplified form as a two-dimensional model. Figure 1 shows the dimensions of the twin-roll casting tool. The inner and outer copper shell radii are fixed, whereas the strip-forming zone length and the strip thickness are varied. The values in Fig. 1 correspond to basic parameters taken for the simulation, which amount to 50 mm for the strip-forming zone length and 3 mm for the strip thickness. Based on the drawing, a CAD model representing melt and copper shells was created.

Fig. 1
figure 1

Drawing of the twin-roll casting tool

Full size image

Pre-processing

In this step, the previously created geometry was discretized (see Fig. 2). An advantage of Fluent is that it applies to unstructured meshes that can be easily created in ANSYS. However, the consequence of this is the loss of calculation accuracy. This study used structured meshes created with the finite element preprocessor HyperMesh to avoid this problem.

Fig. 2
figure 2

Discretization of the model

Full size image

The discretization shown in Fig. 2 consists of 70,208 elements and 72,231 nodes. The discretized model was exported to ANSYS-Fluent to perform the following work step (solving).

Solving

In this step, the simulation was done in Fluent after the boundary conditions and material properties were adjusted. The melt temperature at the top of the strip-forming zone and the vertical component of the entry speed were set as boundary conditions. In addition, the speed of the copper shells was assumed as the rotation speed. Furthermore, it was considered that the copper shells are rigid; thus, the influence of their elastic deformation on the roll gap size can be neglected. The convective heat transfer from the copper shells to the coolant was described as a boundary condition by entering the heat transfer coefficient and the coolant temperature. Regarding the high turbulence caused by the lateral coolant supply and the rotation of the rolls, a heat transfer coefficient of 10,000 W/(m2 K) results [14]. For the coolant temperature, 20 °C (293.15 K) was selected. The thermophysical properties of the aluminum–lithium alloy 2099 and the copper shells are shown in Table 1, and the boundary conditions are in Table 2. The properties of the material in a semi-solid state were calculated using the mixture rule.

Table 1 Thermophysical material properties for the numerical simulation
Full size table
Table 2 Boundary conditions
Full size table

Subsequently, the boundary conditions and the material properties were set in the Fluent software, and the calculation was performed. The results of a complete analysis are shown in Fig. 3.

Fig. 3
figure 3

Temperature (left) and liquid phase (right) distributions

Full size image

After the solving of the simulation was completed, the work step post-processing was carried out.

Post-processing

Post-processing is required to analyze the results. A well-known post-processor for analyzing Fluent's results is CFD. The influence of the process parameters on the casting process can be investigated by comparing the specific length of the deformation zone. The specific length is calculated using the following equation:

$${{l}_{d}}^{^{\prime}}=\frac{{l}_{d}}{{l}_{SFZ}}$$

where \({{l}_{d}}^{^{\prime}}\) is the specific length of the deformation zone, \({l}_{d}\) is the length of the deformation zone, and \({l}_{SFZ}\) is the length of the strip-forming zone. The length of the deformation zone can be calculated by the coordinates of the kissing point, where the solidus temperature is reached. The melt is completely solidified in the deformation zone so that the liquid mass fraction in the deformation zone is zero. The influence of the process parameters can now be simulated and displayed by colorful pictures and curves using these points.

Experimental Procedure

After the numerical study, a trial on the twin-roll casting of an aluminum–lithium alloy 2099 was performed. The chemical composition of the cast strip measured with an optical emission spectrometer Q4 TASMAN from Bruker AXS GmbH is given in Table 3. The experiment was carried out using a lab-scaled vertical twin-roll caster with shells made of a copper alloy CuCr1Zr. The shell geometry corresponds with the specifications given in Fig. 1. Other characteristics of the utilized twin-roll casting unit are described in [15]. The numerical simulation results adjusted the twin-roll casting parameters to reach a moderate plastic reduction and avoid the overloading of the twin-roll caster. Casting speed of 7.2 m/min, casting temperature of 730 °C, strip thickness of 3 mm, and length of the strip-formation zone of 40 mm were used in the experiment. The melt was shielded by argon in the furnace and in the tundish during the pouring. The mean specific roll-separating force during the trial amounted to 1 kN/mm.

Table 3 Chemical composition of the aluminum alloy 2099 used in the study
Full size table

The microstructure of the as-cast material was studied by light optical microscope (LOM) observations using the Zeiss Axio Observer 7 microscope (Jena, Germany) in bright field and circular polarized light differential interference contrasts. These contrasts were used to observe and analyze precipitate dispersion and grain structures. Samples for these observations were polished with SiC papers from P800 up to P4000 FEPA grade. Struers OP-S suspension of colloidal silica was used to finalize polishing. Samples for precipitate dispersion observations were etched with an 0.5% solution of HF in ethanol for 10 s. Samples for grain structure observations were electrochemically polished by Barker’s solution at 10 °C for 300 s. Observations were done from the rolling (RD) and transversal (TD) directions of the material. Grain size and secondary dendrite arm spacing (SDAS) were analyzed using the ImageJ software (National Institute of Health, USA).

Results and Discussion

The thermofluid model was used to determine the influence of casting speed, casting temperature, strip thickness, and length of strip-forming zone on strip outlet temperature and a specific length of the deformation zone. The results at a casting speed of 2–8 m/min (Fig. 4) have shown that the strip outlet temperature increases and the specific length of the deformation zone decreases with increasing casting speed. The contact between the shells and the melt is reduced with increasing casting speed; thus, the melt solidifies more slowly. The cooling of aluminum–lithium alloy in copper shells is rapid, resulting in a relatively high deformation zone length and a low strip outlet temperature.

Fig. 4
figure 4

Influence of the casting speed on the twin-roll casting conditions at casting temperature of 730 °C, strip thickness of 3 mm, and length of strip-forming zone of 50 mm

Full size image

Furthermore, the results have shown that with increasing casting temperature (690–750 °C in steps of ten), the strip outlet temperature increases, and the specific length of the deformation zone becomes smaller (Fig. 5).

Fig. 5
figure 5

Influence of the casting temperature on the twin-roll casting conditions at casting speed of 6 m/min, strip thickness of 3 mm, and length of strip-forming zone of 50 mm

Full size image

A variation of the strip thickness from 2 to 5 mm (Fig. 6) with otherwise constant parameters leads to an increase in strip outlet temperature and a decrease in the specific length of the deformation zone. The variation of the strip thickness has the most substantial effect on the twin-roll casting conditions changing.

Fig. 6
figure 6

Influence of the strip thickness on the twin-roll casting conditions at casting temperature of 730 °C, casting speed of 6 m/min, and length of strip-forming zone of 50 mm

Full size image

A variation of the strip-forming zone length provides an inverted course. The results have shown that with the increasing length of the strip-forming zone (40–70 mm in steps of five) the strip outlet temperature decreases, and the length of the deformation zone increases (Fig. 7).

Fig. 7
figure 7

Influence of the length of strip-forming zone on the twin-roll casting conditions at casting temperature of 730 °C, casting speed of 6 m/min, and strip thickness of 3 mm

Full size image

With the knowledge gained from the simulation, thin strips were produced and their microstructure investigated. As-cast material is relatively coarse-grained. The grain size of the material is two orders of magnitude higher than the solidified eutectic cells. Grains have a bimodal distribution, with equiaxed grains forming at the surface or in the center of the strip and columnar grains forming the bulk of the strip (Fig. 8). Equiaxed grains have a diameter of (260 ± 60) μm. The columnar grains have an equiaxed cross section. They are (220 ± 20) μm wide and are (530 ± 160) μm high. The high scatter of grain height is mainly influenced by the presence or lack of center/surface grains. Such grain microstructure is typical for the strips of Al–Li alloys, which do not contain scandium and are twin-roll cast under the condition of a moderate roll-separating force [16].

Fig. 8
figure 8

Grain structure of the as-cast material. Observation from the RD (left) and TD (right)

Full size image

Constituent particles formed during casting are concentrated into dendrites formed around eutectic cells. The size of the cells is characterized by SDAS. SDAS varies only slightly from surface to center of the strip, and the difference falls within the experimental scatter of both measurements (Fig. 9). The average SDAS is (5.8 ± 0.8) μm as evaluated from both observation directions. SDAS can be higher in some globular regions of the solidified strip. There is an apparent lack of major surface bleeding and centerline segregations in the observed strips. Some segregation occurs along grain boundaries close to the surface of the strip. The dendritic cells are elongated, and the primary axis of elongation differs from grain to grain but is consistent across individual grains.

Fig. 9
figure 9

Particle dispersion at the surface (left) and in the center (right) of the strip. Observation from the TD

Full size image

On the one hand, a homogeneous distribution of finer grains over the cross section would be advantageous for the mechanical properties, especially for the ductility. On the other hand, a slight deformation during the process reduces the anisotropy of the mechanical properties. Analysis of the mechanical properties of the strips and their anisotropy as well as further process optimization for their improvement will be the subjects of further investigations.

Conclusions

Within the scope of the numerical study using the developed thermofluid model implemented in ANSYS-Fluent, dependencies of thermal and deformational states from the main technological process parameters such as casting speed, casting temperature, strip thickness, and length of strip-forming zone for the twin-roll casting of the aluminum–lithium alloy in copper shells were examined. Based on the established correlations, a laboratory twin-roll casting trial with moderate plastic reduction of the material was designed.

The aluminum–lithium strip produced using twin-roll casting has a coarse-grained structure with grain sizes ranging from 200 to 500 μm. Grain distribution is bimodal, with equiaxed grains in the center and at the surface and columnar grains forming the rest. Grain size is two orders of magnitude higher than the secondary dendrite arm spacing of the solidified eutectic cells. The observed dendrite arm spacing is the same across the strips. No significant segregates or dense particle clusters were detected.