1 Introduction

Additive manufacturing of metallic materials by powder bed fusion (PBF) characterize the fabrication of components based on the fusion of metallic powders disposed on the build platform [1,2,3]. This layer production occurs due to thermal energy provided by a laser beam that melts the particles to bond them, giving rise to complex shapes [1,2,3]. Maraging steel 300 is a notable material compatible with the PBF process, considering its application in different segments, such as automotive, tooling, and aerospace [4,5,6,7]. The prior properties of this low carbon steel are the high strength and microhardness [6, 7] due to the martensite matrix with intermetallic precipitates generated by aging treatment [4, 5, 8], which limits the mobility of dislocations on the microstructure [9].

Although the potential application of PBF technology is noticeable in the industry, the main barriers to its plateau of application are microcracks, pores, and high roughness [2, 10,11,12,13,14,15]. Also, components produced by PBF present high residual stresses that require monitoring [15, 16]. Tensile residual stress promotes crack propagations, resulting in lower mechanical strength [15, 16]. However, compressive residual stress can suppress this enlargement [15] and refine the dimension precision of additive manufacturing components [16].

Milling is one of the alternatives to post-process components fabricated with additive manufacturing because of its capacity to provide better surface quality, as mentioned by Fortunato et al. [7]. During milling, the material is removed by a rotative tool that also has a translatory motion along with the work table. Thus, cutting parameters defined as cutting speed (vc), feed per tooth (fz), and depth of cut (ap) characterize the process [17]. An adequate finishing, however, will depend on the correct planning and selection of cutting parameters [18].

The study of cutting parameters can be conducted with the design of experiments (DOE) approach to guarantee robust experiments. This methodology proposed by Montgomery [19] defines planning, execution, and analysis strategies to guide research and emerge as a consolidated method in literature. Lela et al. [20] examined how variations of cutting parameters in milling affect the surface roughness of steel. The most influential factors to roughness recognized were feed per tooth and cutting speed, respectively [20]. The authors also presented ways to optimize this quality feature and defined models to predict its behavior based on cutting parameters [20]. Mutua et al. [8] applied full factorial design to investigate PBF parameters and options of heat treatments to define the optimal configurations range to produce maraging steel parts.

Cutting speed increase leads to high levels of temperature, mainly, where the tool and material defined an interface [17, 21, 22]. Thus, this region assumed different behaviors during chip formation with divergent parameters [23] due to a change in mechanical responses that occurs after thermal expansion and contraction [21]. Moreover, microstructural modification induced by machining deformation [24] or a localized aging treatment [25] also influences surface finishing.

Milling parameters also influence residual stresses on the surface. Tomaz et al. [21] investigated these effects on conventional maraging steel after solution treatment. Using different flood methods, the authors found the higher effects of feed per tooth and cutting speed, respectively, on residual stress on the maraging surface [21]. Furthermore, the selection of cutting parameters also influenced the definition of compressive or tensile behavior on residual stress [21].

PBF and end milling influence the surface quality of material given the innumerous factors involved. Tan et al. [26] highlighted that the temperature gradient, different heat distribution during the laser incidence, and the cooling rate levels in the process generate a range of microstructural defects and components, such as fine and coarse cells. These PBF process characteristics can change the cutting behavior when the additively manufactured material is submitted to milling.

Thus, the roughness evaluation of a specimen produced in both is essential to take better advantage of these technologies. Moreover, it is crucial to guarantee that residual stresses will not be compromised to achieve advantageous roughness levels, considering the relevance of compressive behavior for components surface [15, 16]. To achieve materials produced by PBF with better finishing and mechanical performance combined, this work investigated the effects of cutting speed (vc) and feed per tooth (fz) used during post-processing by milling of maraging steel 300 specimens manufactured by PBF for the average roughness (Ra) and residual stresses based on DOE.

2 Experimental setup

2.1 Powder bed fusion

Maraging steel 300 specimen with 32 × 32 × 10 mm was produced by PBF using EOSINT M280 (Yb-laser) machine, EOS GmbH, Munich, Germany. Figure 1 illustrates the metallic maraging steel powders supplied by LPW (Carpenter Additive), Cheshire, UK. The particle size was calculated based on the micrographs with a magnification of × 500 using the software Image J. The average powder diameter resulted in 24.3 ± 10.2 μm.

Fig. 1
figure 1

Maraging steel powders

Full size image

Table 1 presents the chemical composition obtained by optical emission spectroscopy of the maraging specimen. During PBF, the laser power was 170 W, hatch distance was 0.1 mm, scan speed was 1250 mm/s, and the layer thickness was 0.02 mm. The nitrogen gas was used inside the chamber, and the laser performed a linear scan strategy with 67° rotation between layers.

Table 1 Maraging steel chemical composition
Full size table

2.2 Characterization methods

Microstructural analysis of the as-built specimen was conducted with an optical microscope Scope A1 AxioCam iCc 5 Zeiss (Jena, Germany). A prior microscopy observation of the as-built specimen surface investigated the pattern provided by PBF, especially in terms of high roughness sources. After the polishing and etching, both transversal and top views of the specimen surface were observed for defects’ inspection.

Specimen density was measured using the Archimedes’ principle, as discussed by ASTM [27]. By this approach, the specimen had the weight measure using an analytical balance and then was immersed in a water container (ρwater = 0.9982 g/cm3) to identify the specimen density. This method is widely applied in the literature to provide density measures of parts produced by selective laser melting (SLM) [4, 26, 28, 29].

Three tensile tests were conducted for dog-bone maraging steel 300 specimens with an XZ orientation [1] using a universal machine INSTRON 3369 of Instron, Massachusetts, USA. It was considered a load speed of 1 mm/min for the dog-bone specimens with an overall length of 60 mm, a gauge length of 14 mm, and a thickness of 2.4 mm [30]. Vickers microhardness was also measured using a microhardness tester Homis HV 1000. A 4.9 N test force was applied for 15 s. Five micro indentations were performed for the average value determination [31]. The indentations’ positioning had four measurements near to the face corners and one in the center of the specimen surface. For the milled specimens’ microhardness, the measurements were distributed in the center of the channel.

2.3 Milling

The post-processing was conducted in a CNC machining center Mi 136 II, CNC Group (Santa Barbara D’Oeste, Brazil), using two-flute end mill tools model S902 with Ø6 mm, Dormer (Šumperk, Czech Republic). The tools were made from cemented carbide with helix angle λ = 30°. The depth of cut (ap) of 0.1 mm was kept constant [7, 32]. This process followed the DOE, considering two factors vc and fz with three and two levels, respectively. Using the full factorial approach with three repetitions, 18 tests were performed with the configurations shown in Fig. 2 and Table 2. The selection of the vc and the fz levels followed prior studies and the literature [7, 21, 32].

Fig. 2
figure 2

Milling conditions for experiments

Full size image
Table 2 Cutting parameters for the milling experiments
Full size table

Roughness was measured five times for each specimen, centralized in the channel of the test areas and following the feed direction, by a roughness tester Surtronic S-218 Taylor Hobson (Leicester, UK). Considering a Gaussian filter, the cutoff parameter of 0.8 mm provided the average roughness (Ra) of each specimen. Top surface as-built roughness was also performed for comparison using the same measurement direction. Statistical analysis application enabled the investigation of the effects of milling parameters on the roughness using the software Action Stat (Sao Carlos, Brazil).

The surface finishing of the machined areas was also examined using the optical microscope and under the scanning electron microscope (SEM) Fei Quanta 250, Thermo Fisher Scientific (Massachusetts, USA). At last, residual stresses of the as-built and milled specimens were measured by X-ray diffraction with a Cr-kα tube, a voltage of 25 kV, and a current of 7 mA using the Stresstech G2R machine (Rennerod, Germany). The measurements followed the same direction used for the roughness evaluation, which was also the feed direction, as shown in Fig. 3.

Fig. 3
figure 3

Residual stress evaluation. Red arrow shows the measurement direction

Full size image

3 Results and discussion

3.1 PBF manufactured specimens characterization

As shown in Fig. 4, the building orientation used (XY) provides a better finishing for the top of the specimen due to the remelting conducted for the last three layers. This process consists of carrying out a second laser beam incidence above the material to melt this region again in PBF [33, 34]. The roughness drops 68% on the remelted face when compared to the side face (Fig. 4). Thus, the remelted face (top) was considered for the specimen characterization due to a better roughness condition for material applications.

Fig. 4
figure 4

Roughness of different faces of maraging steel specimen. Red arrows show the average roughness measurement directions

Full size image

One of the most important specifications of the PBF process for surface characteristics and quality is the laser energy density, especially because of its dependence on other relevant parameters [2]. The relation involving scan speed (v), laser power (P), and hatch distance (h), in which the laser energy density (Ed) can be determined, is expressed by Eq. 1 [14]. The laser energy density per volume (Edvol) can be calculated when the layer thickness (t) is included in the denominator [6, 8, 26].

$$ {E}_d=\frac{P}{vh}\kern1em \left[\frac{J}{mm^2}\right] $$
(1)

In this study, an energy density Ed = 1.36 J/mm2 and Edvol = 68.00 J/mm3 were used. These values of energy density can provide low porosities levels according to other studies [8, 26], because of a nearly fully dense structure achieved for maraging steel. This characteristic was also identified in this study based on the relative density of 99.47 ± 0.04%.

Mutua et al. [8] also verified a high-quality surface when a laser energy density per volume of 71.43 J/mm3 was used. Thus, the Edvol = 68.00 J/mm3 adopted in this study had a great potential to provide low surface roughness. Few spherical elements stayed along the remelted as-built surface, as shown in Fig. 5a. These units are the remaining metallic powders derived for a melting issue, considering the diameter of ≈ Ø 40 μm (Fig.5b).

Fig. 5
figure 5

a Surface of the as-built condition and b an unmelted powder identified on the surface

Full size image

The as-built condition presented few defects inherent to PBF, especially melting flaws and a rough aspect with an average roughness of Ra = 3.30 ± 0.36 μm. After metallography preparation and etching, it could be observed a characteristic microstructure of maraging steel 300 produced by PBF. Figure 6a shows the semielliptical melt pools [8, 28] with some interior pores spread in the specimen, even with the high density measured. Moreover, the superior view of the specimen, in which the evaluation of the average roughness and residual stress took place, presented the rotational scan strategy adopted for the process with 67° between consecutive layers (Fig. 6b).

Fig. 6
figure 6

Microstructure of (a) build-oriented face (side) and (b) superior view of maraging steel 300 as-built specimen

Full size image

Figure 7 shows the stress-strain diagram of a tensile test conducted on the as-built specimens. Based on the three repetitions performed, the average ultimate tensile stress (UTS) was 1047.8 ± 56.9 MPa, and the elongation at break was 18.2 ± 1.2%. All the parameters analyzed confirmed the high mechanical properties of maraging steel 300 specimens manufactured by PBF, especially the high ductility, considering the high elongation at break and the diagram shape [1] (Fig. 7).

Fig. 7
figure 7

Stress vs. strain diagram of maraging steel 300 produced by PBF. BO refers to the specimen building orientation, whereas LD refers to the loading direction during the tensile tests

Full size image

3.2 Milled specimens

3.2.1 Roughness

The analysis of variance of roughness results identified two outliers (runs 12 and 18) based on the standardized and studentized residuals analysis, considering that both values exceeded the acceptable interval from −3 to 3 [19]. Residuals reflect the differences between the experimental values and the fitted data [19]. Thus, both measures were disregarded for further investigations. Table 3 presents the values provided by the final evaluation using ANOVA [19]. Considering a significance level of 0.05, feed per tooth and cutting speed were significant for average roughness on the surface, with the p values of 0.000 and 0.007, respectively. Also, statistic F exhibited the same relation and emphasized the major effect of feed per tooth on roughness. The main effects’ graphs show these behaviors (Fig. 8). The predominant effect of fz was also highlighted in other studies [20, 21].

Table 3 ANOVA for surface roughness
Full size table
Fig. 8
figure 8

Main effects’ graphs of feed per tooth and cutting speed on roughness

Full size image

In the case of the cutting speed effect, a smooth decline on roughness (~ 0.1 μm) occurred when vc was adjusted from 150 to 250 m/min (Fig. 8). On the other hand, Ra remained almost constant at 1.4 μm between 250 and 350 mm/min. A roughness average of 0.34 μm could reach a value of 2.50 μm when feed changed from 0.02 to 0.08 mm/tooth (Fig. 8).

The chip formation mechanism involved in the additively manufactured maraging steel was shown in Fig. 9. On materials produced with conventional methods, the crack propagation is expected to occur along the cutting line, according to Hashimura et al. [35]. However, considering the melt pools configured with the laser scanning and some pores generated during the PBF process, the crack growth can suffer deviations from this pattern, following in some cases the melt pool boundaries [36] or even with the layer shearing [8] (Fig. 9). Moreover, regarding the maraging steel 300 properties, the material had a ductile behavior with high strength and elongation at break (Fig. 7), which improve its capacity to great plastically deform before fracture.

Fig. 9
figure 9

Chip formation mechanism

Full size image

Regarding the deviations among the roughness’ values observed for the 150 m/min, 250 m/min, and 350 m/min (Fig. 10), there was not a significant difference to consider when fz = 0.02 mm/tooth. However, the vc increase provided soft lower roughness when fz = 0.08 mm/tooth with a range of 0.18 μm, which justified the presence of a weak vc significance (Table 3). Fortunato et al. [7] also observed this vc trend when fz = 0.08 mm/tooth during milling of as-built additive manufactured maraging steel specimens. It was expected a roughness improvement with increasing vc for conventional materials [17, 22, 23, 37]. Chen [37] highlighted that this mechanism could occur because of the less plastic behavior of the material when higher cutting speeds are used in machining. Due to the enhanced maraging steel capacity to plastically deform and its microstructure/chip formation interaction, this mechanism did not occur for the investigated cutting conditions.

Fig. 10
figure 10

Surface roughness results after milling

Full size image

Roughness was expected to express a growth trend when feed per tooth increased based on literature for conventional [21] and additively manufactured materials [38], and the present study reinforced this trend. Feed per tooth could significantly (Table 3) reduce the roughness when considering the lower level (0.02 mm/tooth) for all cutting speeds, as shown in Fig. 10. The milling-induced changes on microstructure [24, 39] also could explain the differences for both fz results. The interaction between vc and fz did not express statistical significance for surface roughness, considering the p value > 0.05 (Table 3). Tomaz et al. [21] found a significant interaction when investigated the conventional maraging steel after milling with lubrication, but the materials’ microstructure defined with different manufacturing processes together with the different conditions of the machining could influence the difference regarding this study.

By visual inspection, it was verified that deformed material got stuck in the surface, mainly when vc was 250 m/min and 350 m/min. SEM images confirmed the presence of these elements (Fig. 11a and b). Microstructural heterogeneities induced by PBF could enable exceptional intense plastic deformations in some surface regions [23], which can explain these adhered elements on the surface. This occurrence in the intermediate and high cutting speeds also enlarges that the high mechanical performance of the maraging steel, as aforementioned, still presents when using these levels. The size flow phenomenon, which consists in the size smearing of materials disposed near the tool edges along with feed motion [40], was observed for all parameter combinations (Fig. 11c).

Fig. 11
figure 11

SEM images of the milled maraging steel 300 emphasizing the (a) adhered materials, (b) microcracks inherent to residual elements, and (c) the size flow phenomenon

Full size image

Moreover, the disparities observed among parameters were not related to vibrations, due to the absence of chatter marks [23]. Figure 12 shows the best (a) and worst (b) surface conditions obtained in the study. All conditions showed feed marks generated by feed per tooth (Fig. 9) with different intensities, as verified by Biondani et al. [40] and Jeelani and Ramakrishnan [23]. Higher fz defined more pronounced tool marks. Additionally, milling conducted with the low levels of vc (150 m/min) and fz (0.02 mm/tooth) expressed a different pattern, as shown in Fig. 12c, in which microvoids were generated on the surface. Cavities are also common for low vc during machining of conventional alloys due to the broken chip fragments, according to Jeelani and Ramakrishnan [23]. Also, the PBF effects on the cutting mechanism show the possibility of microvoid generations on the milled surface when pores are present during the chip formation (Fig. 9).

Fig. 12
figure 12

Surfaces obtained for (a) the lower roughness (fz = 0.02 mm/tooth and vc = 250 m/min); (b) the higher roughness (fz = 0.08 mm/tooth and vc = 150 m/min); and (c) the different pattern identified for fz = 0.02 mm/tooth and vc = 150 m/min

Full size image

The cutting parameters combination that showed the low surface roughness was fz = 0.02 mm/tooth and vc = 250 m/min (Fig. 10). The drop from Ra = 3.30 ± 0.36 μm of the remelted as-built condition to the optimized Ra = 0.31 ± 0.03 μm enabled a reduction of 90.5% on surface roughness when the milling was performed in the better configuration among the tests. At last, one new tool was used for each machining face resulting in minimum tool wear.

3.2.2 Residual stress

Figure 13 show the residual stress measured for the milled specimens. The evaluation was done to the feed per tooth fz = 0.02 mm/tooth due to the improved roughness results found for this condition.

Fig. 13
figure 13

Residual stress of maraging steel

Full size image

It is well known that the PBF process generates higher residual stresses, and mainly tensile residual stresses are observed [4, 15, 16]. Nevertheless, the remelting of the top layers induced compressive stress on the specimen. This observation could be related to the austenite-martensite transformation [16] during the PBF process. Compressive residual stresses are beneficial to mechanical properties, mainly to reduce the probability of crack open [16]. Compressive residual stresses identified for the remelted as-built specimen were reduced from − 394 ± 18 MPa to an average of − 25 ± 3 MPa, − 57 ± 8 MPa, and − 103 ± 4 MPa when performing milling the vc 150 m/min, 250 m/min, and 350 m/min, respectively. Becker and Dimitrov [4] observed high tensile residual stresses of 720 ± 142 MPa and 455 ± 91 MPa for the single and double scan strategy, respectively. The discrepancy among the remelted and milled specimens’ results was related to the surface conditions.

Higher cutting speeds defined more compressive residual stress (Fig. 13). A strong negative correlation between both variables was obtained, considering the Pearson coefficient of − 0.84. The increase of cutting speed provided more intense plastic deformations, identified with the presence of deformed materials attached on the surface [23]. This mechanism could enable the generation of less tensile residual stress when the cutting speed increased, considering the vc range adopted [41]. Hence, the use of vc = 250 m/min and vc = 350 m/min for maraging steel milling showed a way to improve surface roughness taking the benefit to induce compressive residual stress on the final part.

3.2.3 Microhardness

The polished as-built specimen reached a higher level of 399 ± 10 HV. The remelted face showed an average of 329 ± 40 HV. The decrease in microhardness with the remelting followed the literature [34]. Aging heat treatment generates higher hardness for maraging steel specimens [8, 29]. Remelting of the last three layers also induces a localized aging treatment [33], but the hardening mechanism provided by this effect was less expressive based on the microhardness results. Moreover, the second laser scanning performed with the remelting can define coarser grains in the microstructure due to a different kind of heat flow and solidification time [42]. Thus, it could also induce a lower microhardness, as emphasized by Bhardwaj and Shukla [42].

The microhardness evaluation after milling was done to the fz that had a better performance for roughness improvement (fz = 0.02 mm/tooth). Milled specimens expressed almost the same microhardness average for the three levels of cutting speed. The results were 412 ± 9 HV, 414 ± 9 HV, and 409 ± 19 HV for vc = 150 m/min, vc = 250 m/min, and vc = 350 m/min, respectively. Figure 14 presents the comparison of microhardness among all specimens. The higher deviation obtained for the remelted face in comparison to the others is related to the limitation imposed by the roughness and surface defects for microhardness evaluation in the prior [43]. Focusing on the as-built (polished) and the machined specimens, the Ra diverged approximately 0.12 μm among the milled specimens, so the impact of this factor for microhardness evaluation was disregarded.

Fig. 14
figure 14

Average Vickers microhardness

Full size image

Closer microhardness results among the three milled specimens emphasize that cutting speed did not provide the required thermal conditions for the occurrence of aging with the feed per tooth fz = 0.02 mm/tooth. The presence of different levels of compressive residual stress on the surface before and after milling could also affect the microhardness [39]. Regarding the lower differences for the compressive residual stresses among the milled surfaces with different cutting speeds (Fig. 13), both the residual stress and the microstructural and phase changes resulted from the material plastic deformation during milling [7, 24] could influence the microhardness [39]. The latter, work hardening [39], together with the effect of the polishing, could also explain the possible smooth average microhardness increase observed for the polished to milled specimens. The lack of relation between cutting speed and surface microhardness was also evidenced by a Pearson coefficient of − 0.14.

4 Conclusion

Effects of cutting speed and feed per tooth used during milling of maraging steel 300 manufactured by PBF on the average roughness and residual stresses were investigated. The main findings emphasized the combined improvements in surface roughness and residual stresses can be achieved with adequate cutting parameters’ selection. Considering the almost fully dense maraging steel produced by PBF, high mechanical properties could be achieved, especially the ultimate tensile stress of 1047.8 ± 56.9 MPa. Residual stresses obtained for a remelted surface showed compressive residual stress that is beneficial for maraging applications. However, roughness expressed values above 3.00 μm that needed to be enhanced.

Milling used as post-processing provided a roughness decrease from 3.30 μm (as-built) to 0.31 μm, when fz = 0.02 mm/tooth and vc = 250 m/min was adopted. Among fz and vc, feed per tooth was more significant to refine roughness levels. Cutting speed also affected roughness, but the interaction between fz and vc did not present statistical significance for roughness. The melt pools, layer construction, and porosity generated during the additive manufacturing can affect the chip formation.

Focusing on the best feed per tooth condition (0.02 mm/tooth), residual stress evaluation emphasized the generation of compressive residual stress for all cutting speeds used. vc = 350 m/min resulted in the higher compressive residual stress (− 103 ± 4 MPa). The feed per tooth combined with the cutting speed improved the surface roughness and the compressive residual stress of the specimens, showing the importance of considering both these parameters in the milling process planning of PBF maraging steel parts.