**3. Results**

### *3.1. Process Parameters and Prediction Model*

The first stage of process parameter development included the AM of 27 samples. All data and predicted porosity results are included in Table A1. The modification ranges of the parameters were as follows:

• *PL* from 160 to 240 W;


The groups of parameters in which the requirement *Pρ* < 0.3% was achieved were characterized by energy densities ranging from 95.2 to 121.2 J/mm3. Additionally, we ben observed that the measured porosity values were different despite reaching similar values of energy density in the given groups. The 1.25, 1.26, and 1.27 parameter groups (Table A1) taken from the available literature [23,27] did not achieve the required porosity regime. Equation (1) was used to estimate porosity values for these cases. As variables, three parameters were considered: *<sup>x</sup>*1—laser power, *<sup>x</sup>*2—exposure speed, and *<sup>x</sup>*3—hatching distance. Using the method of least squares, the coefficients βm and βmn (for m = 1, 2, 3; *n* = 1, 2, 3) were established. Additionally, by constructing the analysis of the variance table, it was possible to determine the statistical significance and create Equation (3):

$$\begin{array}{l} y = 52.149000 - 0.018000 \text{x}\_1 - 0.067000 \text{x}\_2 - 525.55000 \text{x}\_3 + 0.00032000 \text{x}\_1^2 + 0.000031 \text{x}\_2^2 \\ + 1499.389000 \text{x}\_3^2 - 0.000099 \text{x}\_1 \text{x}\_2 - 0.475000 \text{x}\_1 \text{x}\_3 + 0.449000 \text{x}\_2 \text{x}\_3 + 0.580000 \end{array} \tag{3}$$

According to a different number of considered cases for *x*1, *x*2, and *x*3 variables, the statistical significance was valid for only one part of the equation (exposure speed, *p* = 0.015). The value of the R<sup>2</sup> coefficient for this exact case was equal to 0.83. Based on the established minimum of a function (3), it was possible to estimate the parameter groups, ensuring the same density as the parent material. Additionally, a determinizing estimation of the process parameters groups was made. This approach allowed us to obtain one additional group of parameters, with which the theoretical porosity value of the produced

sample would be close to that produced using the extreme of function (3). The obtained process parameter groups are shown in Table 4.

**Table 4.** Process parameter values were obtained by following the statistical model based on the results of the first experiments.


Before validating the obtained parameter groups, we decided to create 33 full factorial designs to distribute the measurement points equally in the considered space, which was restricted by the given variables (x1, x2, and x3). Based on the obtained results (included in Table A1), it was possible to identify three process parameter value for each variable. The selection was made according to the lowest measured porosity achieved using an exact parameter group. In the case of the hatching distance, as an additional third variable, a value of 0.120 mm was taken into account.

This approach allowed for the generation of 27 different parameter groups (Table A2). Moreover, based on Equation (3), two additional groups of parameters 2.28 and 2.29 (Table 4) were estimated. Samples manufactured through the selected parameter groups were examined using the same procedure as was performed in the first part of the experiment (Section 2.1). Ten out of a total of 29 parameter groups achieved the porosity condition (*Pρ* < 0.3%). Subsequently, 2.29 groups (extreme of function (4)) indicated the minimum value of porosity ( *<sup>P</sup>ρ*2.29 = 0.08%) among all the tested samples.

Another level in the selection of the process parameters in the initial statistical model was supplemented by the examination of porosity measurements. This influenced the changes in the regression and *p* coefficient values (Table 5). Based on the *p*-factor, the most significant factors in the equation among the variables were the scanning velocity and the two combinations of velocity with laser power, as well as the hatch distance.


**Table 5.** Regression and *p* coefficient values.

The R<sup>2</sup> value equal to 0.83 did not change significantly compared to the initial value designated in the first part of this study (despite the increase in the statistical significance coefficient number). The final form of the statistical model can be described by the following Equation (4):

*y*= 29.829000 − 0.011000*<sup>x</sup>*1 − 0.041000*<sup>x</sup>*2 − 294.833000*<sup>x</sup>*3 + 0.000140*x*21 + 0.000022*x*22 <sup>+</sup>570.583000*x*23− 0.000078*<sup>x</sup>*1*x*2− 0.007000*<sup>x</sup>*1*x*3 + 0.278000*<sup>x</sup>*2*x*3+ 0.460000 (4)

> The extreme of the function (4) has been considered in Table A2 as parameter group 2.30. However, the validation indicated a lack of repeatability in the results. This phenomenon was related to the adoption of a hatching distance that was too small, which caused overlapping of the exposure lines and, as a result, increased the number of voids. Hence, it was essential to consider all technical aspects related to the AM process manually because the statistical model did not take into account such limitations. Finally, based on the all obtained results, it was possible to identify five process parameter groups that were used in the further analysis:

• Group "2.3"—the highest value of the energy density;


Statistica software was used to express the statistical model's answers as a graphical image (only when a constant value was assumed for one of the variables). The answer surfaces generated when a constant value was used for the hatching distance parameter are shown in Figure 3.

Indicating the model answer in such a way allows one to observe the range of the process parameters in which the porosity of the AM parts would be lower than 0.3%—the so-called "technological window". Such an estimation allows one to characterize the given material from the point of view of density without the need to conduct the whole spectrum of this kind of research. Additionally, five groups of process parameters selected for further research were shown on the answer surfaces (Figure 3). Process parameter groups 2.11 and 2.29 were located in the areas in which the statistical model estimated porosity values lower than 0.3%. Process parameter groups 2.3 and 2.12 were located close to the areas mentioned earlier.

**Figure 3.** *Cont.*

**Figure 3.** Answer surfaces generated for the constant value of the following hatching distance parameters: (**a**) hd = 0.0110 mm and (**b**) hd = 0.0100 mm.

Figure 4 shows a chart of the porosity values as a function of energy density. It can be observed that the growth of the energy density positively affected the reduction of voids in the material volume. It can also be observed that changes in the pores' shapes were related to the increase in energy density (this is shown in the figures located in the chart in Figure 4). The use of low-energy density values caused the formation of irregular void shapes, which could be connected to the "lack of fusion" phenomenon [20].When the energy density value was greater than 80 J/mm3, the share of spherical shapes (caused by gas porosity) and irregular shapes (caused by lack of fusion) of voids could be observed. After exceeding 100 J/mm3, the voids mainly exhibited gas porosity characteristics. For a similar value of energy density, different authors observed similar defects in the structure of the material [20,26,27]. No cracks were found in the microstructure.The point at which the line denoting the acceptable value of the porosity crossed the trend line (determined by measurement) indicated a value of energy density equal to 104 J/mm3. This is an empirically determined value that should be used in the AM of 21NiCrMo2 steel to achieve the condition *Pρ* < 0.3%. For the experimentally indicated energy density (104 J/mm3), the estimated porosity values with the use of Equation (4) were as follows:


These results led us to conclude that the statistical model created in this study enables us to estimate the porosity values of the AM parts made of 21NiCrMo2 steel.

**Figure 4.** Porosity values for AM samples as a function of the energy density observed during the microscopical investigation.

### *3.2. Microstructure and Chemical Composition*

PBF-LB/M and conventionally made samples were subjected to microstructural analyses (in both cross-sections). Images of the registered microstructures are shown in Figure 5. The parent material (Figure 5a) was characterized by a typical ferritic-pearlitic structure [32]. In the case of the PBF-LB/M samples, the microstructure was standard for this kind of AM process. The XY (Figure 5b) plane of the AM samples revealed exposure paths, whereas in the YZ plane, a layered structure of the deposited and melted material was visible. In addition, in Figure 5b,c, examples of porosity present in the structure are marked with green arrows. Because of the diverse cooling rates, there were observable differences in particular exposure lines, which led to a visible differentiation of the obtained microstructure. The main reason for this phenomenon is that the heat conductivity from the melt pools' outlines was more significant than that from their center, as proven by Schmitt et al. [33]. The overall microstructure showed similarity to a martensitic-bainite structure. Because of the presence of high-temperature gradients, additional heat treatment by means of stress relief was carried out (following the guidelines shown in Table 3). Hence, we registered the loss of the layered structure and made the microstructure finer but it still exhibited the form of an acicular microstructure (Figure 5d,e). These effects are typical of recrystallization annealing. The authors of [20] also demonstrated a similar phenomenon for materials produced using PBF-LB/M after the same type of heat treatment. This effect may be due to the presence of a large number of microstructural defects in the structure in its as-built state, which may lower the temperature required for the complete rebuilding of the microstructure of the material.

**Figure 5.** 21NiCrMo2 microstructure images, 2% nital etched, obtained conventionally (**a**) and via the PBF-LB/M process in both planes: XY (**b**) and YZ (**c**); after being subjected to heat treatment (AHT) and PBF-LB/M—XY (**d**), and YZ (**e**) with the visible heterogeneous micro-area of concentrated alloy elements (**f**).

Additionally, as shown in Figure 5f, in a part of the microstructure we observed a heterogeneous micro-area of concentrated alloy elements. This phenomenon was observed in both analyzed cross-sections (XY and YZ, marked with red squares and with arrows in Figure 5b,c), independently of the process parameter groups used. The dimensions of those structural heterogeneities were close to the width of a particular exposure path and a depth of 1–3 layers. According to the available research [20,27,28], such inclusions are formed during the melting process, and their chemical composition is similar to the rest of the material. In Figure 6, an EDS element distribution is identified in the mentioned micro-area.

The main alloy elements which were concentrated in the localized micro-areas were chrome, nickel, and molybdenum. The significant value of the standard deviation of the chemical composition analysis (Figure 7) indicates that a meaningful diversity of alloy elements was shared in these structures. The natural after-effect of this phenomenon was a reduction of the share of iron (Figure 6c). There is a need to conduct more detailed analyses of these phenomena in order to identify the reason for the formation of those heterogeneities.

### *3.3. Tensile Testing and Hardness Measurements*

Dog-bone-shaped samples were made using the SLM 125HL device. The sample's orientation was selected according to Figure 2. The orientation was limited to one position only, because from the point of view of material strength, this was the most advantageous position. Some specimens were chosen to undergo stress-relief heat treatment directly

after the PBF-LB/M process (without the removal of samples from the substrate plate). That kind of approach allowed us to avoid the potential deformation of the manufactured parts. Conventionally made, PBF-LB/M-produced (made using the five selected parameter groups), and heat-treated PBF-LB/M samples were tested. The tensile test results are shown in Figure 8. The porosity of samples subjected to static tensile tests was also measures and it did not differ from the values obtained in the case of cubic samples.

**Figure 6.** EDS element distribution identified in the 21NiCrMo2 steel area of structural heterogeneity: (**a**) micro-area image, (**b**) share of iron, (**c**) share of molybdenum, (**d**) share of chrome, (**e**) share of nickel, and (**f**) share of manganese.

**Figure 7.** A bar chart of the exact alloy element percentages in the heterogeneous micro-area.

**Figure 8.** Stress-strain curves of the PBF-LB/M samples (as-built and after heat-treatment—AHT) compared with the parent materials (after normalization).

The PBF-LB/M-made samples, in the as-built state, were characterized by a lack of a visible yield point, independent of the process parameters used. Generally, the material in the as-built state was characterized by the lowest deformation and the highest strength, which was directly related to the martensitic-bainitic microstructure of the material. The visible difference between each parameter group was registered in the strain values during the fracturing of the samples (in the as-built state). Parts obtained with the use of parameter groups 2.30 and 2.29 were characterized by smaller strain values than those of the counterparts made with the use of groups 2.3, 2.11, and 2.12. Additional heat treatment caused a change in the tensile testing curve. Through this post-processing step, it was possible to obtain a visible yield point, but at the same time, there was a significant slope in terms of the registered UTS. Excluding the samples from group 2.30, a proportional increase in the strain of the PBF-LB/M samples was visible. Additionally, parts obtained employing the parameters from group 2.30 were characterized by the highest porosity (0.54%—Table A2), which directly affected the decrease in the total strain during tensile testing. This visibly affected the total strain values, whereas the change in UTS was negligible. The parent materials in the normalized conditions were characterized by a lower UTS and a higher strain than as-built PBF-LB/M parts. This phenomenon was strictly related to microstructural material conditions. In Table 6, we present the compiled tensile test results obtained for all tested parts.

The tensile strength of the PBF-LB/M samples was close to 1012 MPa, whereas the yield point assumed values in the range of 925–970 MPa. A slight difference between given process parameter groups was also maintained in the case of samples subjected to additional heat treatment. This indicated a minor deformation strengthening effect, which is a negative phenomenon in parts dedicated to machine design. Unlike the samples obtained using parameter group 2.30, a relatively low strain value was registered during the tensile testing of parts from group 2.29. This encourages us to constantly verify the strength properties. Additional heat treatment causes a decrease in the UTS and YS to the average values of 743 MPa and 698 MPa, respectively, and an increase in the total strain to an average of 17.3% (excluding group 2.30). Under the same conditions, the 16MnCr5 materials exhibited UTS = 730 MPa and YS = 658 MPa [26].



For the hardness testing, the HV1 Vickers methodology was used. A selection of the HV1 was strictly related to the minimal tip spot hole area. The registered hardness values are shown in Table 7.


**Table 7.** HV1 hardness measurements of the parent material and PBF-LB/M (as-built and AHT) material.

The as-built material was characterized by an average hardness of 331–357 HV1, and this was independent of the process parameter group used and the energy density used (100.3–133.3 J/mm3). The maximal difference between measured values in both tested cross-sections (XY and YZ) was equal to 12 HV1, which could prove a lack of anisotropy in the case of this exact material property. The fact that the highest hardness of the material was observed in the as-built state was directly caused by the presence of a fine-grained

martensitic-bainite structure in contrast to the ferritic-pearlitic structure of the parent material. A decrease in the hardness was registered after additional heat treatment, with the value of 260 HV1. This was the effect of microstructural changes, which also confirms the occurrence of phenomena related to recrystallization. In both as-built and AHT conditions, PBF-LB/M material was characterized by a higher hardness value than that of the parent material, which is typical in such comparisons. The material in both states was characterized by a higher hardness than PBF-LB/M 20MnCr5 and 16MnCr5 [19,20], and this may be due to the inclusion of elements which improve hardenability in 21NiCrMo2 steel.
