Keywords

1 Introduction

Industry 4.0 is the result of the twenty-first-century demand for innovation toward a revolution in products and production chains. This concept has become widely spread throughout society and is evidenced by the transformations that manufacturing processes have undergone. According to Dilberoglu [3], this new industry format integrates physical and digital systems, giving rise to smart factories. Among the concepts regarding the physical component, manufacturing is a fundamental part; conversely, it also acts as a limiting agent to the capacity of these factories. Therefore, this scenario makes the development of new non-traditional manufacturing methods a vital factor, among which additive manufacturing (AM), widely known as “3D printing”, can be highlighted.

According to Chryssolouris [2], this manufacturing technology has the potential to meet the current demand for flexibility in industries—being able to print customized products with complex geometries that are difficult to manufacture using conventional techniques. It still allows integration with digital systems—using CAD (Computer-Aided Design), CAE (Computer-Aided Engineering), and CAM (Computer-Aided-Manufacturing) software.

Metallic materials are widely used in various fields of engineering. There is also the emergence of AM technology by the fusion of metal powder bed FLPM (MPBF—Metal Powder Bed Fusion), [7]. Among these metallic materials, titanium alloy Ti6Al4V is the material focused on by this research.

Although previous works indicate the importance of understanding the generation of residual stress and distortions in Selective Laser Melting (SLM) processes, the underlying mechanisms for the generation of residual stress remain poorly understood. In order to better determine the factors that influence the accumulation of residual stress as well as the prevention of distortions, delamination, and fractures [5], a combination of parameters is sought for an ideal process window (one that meets the requirements product performance). Therefore, thermomechanical models for finite element simulation (FEM) of SLM are potentially valuable, although they are a challenge due to the complexity of the physics involved in the process [6].

A significant problem associated with SLM components is the development of high internal residual stress [1]. The repeated cycles of heating and cooling successive layers of powdered raw material during the SLM construction process produce high cooling rates and high-temperature gradients associated with the process, resulting in residual stress buildup in SLM components. Parts may fail during SLM construction or later in service due to these high internal residual stresses [4].

Therefore, the present work aims to contribute to understanding how some process parameters such as laser power, scanning speed, and hatch affect the generation of residual stresses and distortions in the printed part—both numerically (using the commercial software Simufact Additive 2020 FP1 from MSC Softwares) and experimentally. A factorial design is used to select the variation of parameters, and the analysis of variance relates the results (residual stress and distortion) with the selected variables.

2 Materials and Methods

2.1 Equipment

The manufacture of Ti6Al4V titanium alloy samples via additive manufacturing technique by selective laser melting was carried out in an OmniSint-160 SLM equipment, ytterbium fiber laser module, Rycus source 500 W of nominal power, from OMNITEK (see Fig. 1).

Fig. 1
figure 1

OmniSint-160 additive manufacturing machine

Full size image

Measurements of residual stresses of the titanium parts were carried out using the RIGAKU brand X-ray diffraction equipment, model Ultima IV (Fig. 2a) located at the Nuclear Research Institute (IPEN). To measure the thermal distortions, the optical measuring equipment ATOS Core 80—CP40/MV100 was used, with a resolution of 5 megapixels. Data was processed by the GOM Inspect 2021 software—in a partnership with the company Vtech (Fig. 2b).

Fig. 2
figure 2

a RIGAKU diffractometer, Ultima IV model, b ATOS Core 80 optical meter

Full size image

2.2 Experimental Methodology

For evaluating the impact of the variation of the SLM manufacturing process parameters, such as the distance between beams (hatch), scanning speed, and laser power, on the appearance of residual stresses and thermal distortions, 16 cylindrical specimens with 11.3 mm diameter and 10 mm height were used, according to a factorial design with the following variable factors:

  • Power (100 W–200 W)

  • Speed (500 mm/s–1500 mm/s)

  • Hatch distance (50 μm–90 μm)

Table 1 presents the order of the tests and the parameters considered for each part manufactured by the SLM manufacturer.

Table 1 Manufacturing parameters during the experimental stage
Full size table

Figure 3 shows the 16 samples of Ti6Al4V titanium printed on the printing mat.

Fig. 3
figure 3

Printed Ti6Al4V samples

Full size image

The parts were cut from the printing base using wire EDM, which greatly minimizes the possibility of generating residual stresses during cutting.

2.3 Computational Methodology

Numerical analysis was performed using the Simufact Additive 2020 FP1 software from Hexagon MSC Software, which uses CAD models to apply the same manufacturing conditions, defining a thermomechanical simulation. Such settings are established directly by the graphical user interface and the software calibration process. To minimize the complexity—caused by the wide range of data that influence the process—it’s necessary a pre-calibration step that uses the inherent strain method, marked by the definition of the volume expansion factor (VEF). The material’s properties can be selected in Simufact Material 2020 FP1 that will supply our model with standard data for that material or can be defined manually using the datasheet provided by the supplier or data collected experimentally.

VEF is responsible for correcting the effects of thermal expansion and contraction, since the element generated by the discretization of the domain in voxels (hexahedral finite elements—Fig. 4) consists of more than one layer of material. For selecting this factor, a geometry pre-selected by the application is generally used, the cantilever (which facilitates the characterization of the state of deformations). However, as such specifications would already be provided by the experimental study, a convergence analysis of the factor was carried out for an arbitrary sample, through the relationship with the data obtained numerically and experimentally—aiming at the approximation of the thermal distortions, so that, if there is an equivalence in the displacements, it can be stated that the thermal deformation was correctly captured by the software and the residual stresses can thus be calculated, since they derive from deformations caused by the temperature gradient.

Fig. 4
figure 4

Discretization of the continuous domain of geometry

Full size image

When defining the metal powder bed melting process, the stages of the process were also defined, namely: construction of the part and separation of the base. The base separation was set up straight away. For the construction of the part, it is necessary to insert the constant parameters (the width of the laser beam and the scanning strategy) and variables for the study, namely: laser power, scanning speed, and distance between scanning vectors (hatch)—being the same parameters varied for the experimental part.

3 Results and Discussion

3.1 Experimental Thermal Distortions

Initially, the sample (Fig. 5a) is subjected to scanning in the ATOS Core 80 equipment of its entire external surface (Fig. 5b) resulting in the geometry of Fig. 5c, which is treated using the GOM Inspect software.

Fig. 5
figure 5

a 10 T printed sample, b optical measurement process, c scanned geometry of the sample

Full size image

After scanning each printed piece, the GOM Inspect software is used to perform the analysis of metrological parameters. Initially, the geometry scanned by a cylinder is superimposed with the nominal dimensions of the part, which allows evaluating the distortions resulting from the manufacturing process. For this research, only the mean diameter was defined as the distortion response parameter.

Figure 6 shows a dimensional comparison of the entire scanned surface of sample 10T in relation to the nominal cylindrical body, with nominal dimensions of 10 mm height by 11.3 mm diameter. Warm colors indicate larger dimensions and cool colors indicate smaller dimensions than the nominal value.

Fig. 6
figure 6

Representation of distortions of the 10 T printed sample

Full size image

3.2 Experimental Residual Stresses

In this step, the samples had their residual stresses measured. Measurements were taken on the side of each piece. In Fig. 7a, b, it is possible to visualize the part being measured by the RIGAKU diffractometer, model Ultima IV, in the lateral position.

Fig. 7
figure 7

Measurement of the 10 T sample by diffraction a external view of the machine, b measurement position

Full size image

3.3 Numerical Simulations

After the printing parameters setup steps according to factorial planning for each sample, it was possible to measure the thermal distortions and calculate the residual stresses. Distortions were measured at a total of 15 points distributed on the cylindrical surface, 5 points at a height of 2.5 mm from the base, 5 points at a height of 5.0 mm, and finally, 5 points at a height of 7.5 mm Sample. Figure 8 shows some of the points measured in the simulation stage. The measurement of the residual stress was made by a single point of the same coordinate that was considered in the experimental stage in the first phase, and later the other points were considered as explained later.

Fig. 8
figure 8

Distortion measurement points in the simulation stage

Full size image

Table 2 presents the distortion and residual stress results from the experimental stages and simulations for each of the 16 specimens with the respective relative percentage deviations between the values obtained.

Table 2 Results of experimental and simulated distortions and residual stresses
Full size table

3.4 Analysis of Variance

Figure 9a, b show the result of the analysis of variance for the experimental stage. Figure 9a shows the Pareto chart of the mean diameter with the reference line for statistical significance (α = 0.05). The factors that present values lower than the reference line (to the left of the line) do not present statistical significance in the result of the response studied. Figure 9b presents the main effects of the factors on the mean diameter.

Fig. 9
figure 9

a Pareto chart for mean diameter, b Main effects chart for mean diameter—experimental stage

Full size image

The result of the analysis indicated that the power and hatch distance factors, respectively, influence the variation in the average diameter of the printed parts at the 95% confidence level. The velocity factor was not statistically significant, at an adequate confidence level (95%), for any discussion or correlation with the result of the mean diameter variation.

Analyzing Fig. 9b, a trend toward an increase in diameter is observed with increasing power, which may be related to an increase in local temperature at the time of fusion of the powder in the layer. However, with increasing scanning speed and hatching (hatching at greater intensity), the mean diameter decreased. The variation of these two factors changes the cooling rate of the layer, which may have caused a reduction effect on the final diameter dimension.

Figure 10 shows the Pareto chart of the residual stress measured at a single point on the side of the printed parts in the experimental stage with the reference line for statistical significance (α = 0.05).

Fig. 10
figure 10

Pareto chart for the lateral residual stress—experimental stage

Full size image

The result of the analysis indicated that none of the power, velocity, and hatch factors statistically influence the residual stress on the side of the part at the 95% confidence level. This is most likely due to the difficulty in obtaining a homogeneous stress distribution condition in such a small geometry piece.

To allow a more reliable verification, once the relationship between the printed part and the simulated part was proven for the measured point shown in Table 2, we decided to measure more points along the radius and also at the height of the part, which allowed working with an average value of residual stress (Fig. 8).

The average residual stress values can be seen in Table 3 and compared with the stress values measured at a single point. A great difference is observed, which makes the computational simulation step even more important in this sample characterization process. The residual stresses on the side presented their lowest values, when sufficiently close to the bottom and top. These results made it possible to perform the analysis of variance, observing this time significance in the results.

Table 3 Comparative results between average and “one point” residual stress
Full size table

Figure 11a, b show the Pareto chart of the residual stress obtained on the side face of the parts simulated in the numerical step considering the average value measured. Figure 11b shows the graphs of the main effects of the parameters on the residual stress result.

Fig. 11
figure 11

a Pareto plot for mean lateral residual stress, b Main effects plot for mean lateral residual stress—numerical step

Full size image

The result of the analysis indicated that all three factors, laser power, speed, and hatch distance, statistically influence the residual stress of the part at the 95% confidence level.

Analyzing Fig. 11b, it is possible to verify a decreasing tendency of the residual voltage on the lateral face with the increase of the power. However, by increasing sweep speed and hatch distance, the residual stress of the side face increased.

4 Conclusions

After carrying out the studies, the modeling of the distortions showed to present good repeatability while the residual stresses and their measurements in a single point present significant uncertainty. The software can thus be concluded to be able to accurately predict the distortion and, qualitatively, the residual stresses measured on the surfaces studied.

It was also found that the average diameter of Ti6Al4V titanium parts was greatly influenced by the laser power followed by the hatch distance. The increase in power increased the average diameter of the samples, while the increase in speed and hatch decreased their diameter.

The residual stress measured on the lateral face through a single point showed great variation when comparing the experimental and numerical results. This is due to imprecision in the measured coordinate and also to difficulties in stabilizing the X-ray focus during the acquisition of measurements in the experimental stages. The reduced dimensions of the printed pieces and high porosity are believed to have generated this difficulty in measurement.

In turn, the simulated mean residual stress measured on the lateral face of the Ti6Al4V titanium parts showed a statistical behavior, being greatly influenced by the laser power, followed by the sweep speed and the hatch distance. Increasing laser power decreased residual stress, while increasing speed and hatch tended to increase the mean lateral residual stress.