Enhancing Fatigue Life of Metal Parts Produced by High-Speed Laser Powder Bed Fusion Through In Situ Surface Quality Improvement

Abstract

The poor surface quality of the metal parts produced by laser powder bed fusion limits their application in load-bearing components, as it promotes crack initiation under cyclic loadings. Consequently, improving part quality relies on time-consuming surface finishing. This work explores a dual-laser powder bed fusion strategy to simultaneously improve the productivity, surface quality, and fatigue life of parts with inclined up-facing surfaces made from a novel tool steel. This is achieved by combining building using a high layer thickness of 120 μm with in situ quality enhancement through powder removal and laser remelting. A bending fatigue campaign was conducted to assess the performance of such treated samples produced with different layer thicknesses (60 μm, hull-bulk 60/120 μm, 120 μm) compared to as-built and machined reference samples. Remelting consistently enhanced the fatigue life compared to the as-built reference samples by up to a factor of 36. The improvement was attributed to the reduced surface roughness, the reduced critical stress concentration factors, and the gradually changing surface features with increased lateral dimensions. This led to a beneficial load distribution and fewer potential crack initiation points. Finally, the remelting samples produced with a layer thickness of 120 μm enhanced the fatigue life by a factor of four and reduced the production time by 30% compared to the standard approach using a layer thickness of 60 μm.

1. Introduction

Laser powder bed fusion (LPBF) is a widely recognized technique for the additive manufacturing (AM) of metal parts. The technique enables the production of parts with complex geometries by utilizing a laser to melt and consolidate the powder particles layer by layer. While the design flexibility in AM makes LPBF well suited for applications in the tooling [1] and medical sectors [2], among others, its adoption for load-bearing components remains limited. In this regard, the low as-built surface quality falls short compared to conventionally manufactured parts [3]. The typical surface irregularities of parts produced by LPBF include partially melted particles [4], ridges between the melt tracks due to inadequate hatch spacing [5], the balling of melt tracks [6], hatching patterns [7], the redeposition of ejected spatters [8], and elevated part edges [9]. These irregularities contribute to a high roughness, which largely varies with the building conditions and can typically fall within the range of 5–40 μm [10]. Controlling the surface quality is particularly challenging for the often complex-shaped AM parts, as the quality highly depends on the build orientation [11]. Specifically, the quality of up-facing inclined surfaces comprises a combination of different phenomena. Most prominently, the surface quality is affected by the staircase effect, which depends on the inclination angle and layer thickness. In addition, the presence of partially molten particles and elevated edges along the stair steps further degrades the surface quality [12]. As a result, the as-built roughness on inclined surfaces is often higher than on horizontal and vertical surfaces relative to the build plate [12,13].

It is well known that the poor surface quality of metal parts fabricated through LPBF can negatively impact their fatigue performance. Early fatigue failure is frequently associated with a high as-built roughness, surface defects, and notches [14,15].

These sites locally amplify the stress under loading, thereby increasing the risk of crack nucleation and initiation. Local stress concentration factors capture this by considering the surface valleys as notches that act as stress concentrators. This approach accounts for the notch morphology by extracting the depth-to-radius ratio for each valley and calculating the corresponding local stress values [16]. Cutolo et al. further extended this approach to compute the stress concentration factors at the critical notches that will most likely initiate fatigue failure [17]. Different studies have demonstrated the suitability of stress concentration factors as input for correlating the fatigue performance of additively manufactured metal parts [17,18].

Post-AM surface treatments are systematically applied to improve the surface quality and consequently the fatigue performance. While conventional methods such as milling, polishing, or grinding are commonly used, various alternative treatments are constantly being explored to accommodate the complex geometries of AM parts, as reviewed in [19]. Considering the considerable efforts associated with the surface finishing of intricate part geometries, improving the as-built surface quality would offer significant benefits.

In this respect, laser remelting is a promising technique to improve the surface quality directly during building. While laser remelting enables the quality of horizontal surfaces to be enhanced through the re-scanning of build layers [20], the treatment of inclined surfaces is less common as these are covered by loose powder particles. A recently introduced dual-laser powder bed fusion (DLPBF) strategy demonstrated the quality improvement of up-facing inclined surfaces using laser-induced shock waves for powder removal and subsequent laser remelting. Although these recent studies have demonstrated notable improvements in the three-point bending fatigue performance of such treated parts made from M300 and Ti64 [18] and an aluminium-based metal matrix composite [21], further research is required to explore the wider potential of this novel approach. Specifically, it remains unexplored whether this surface treatment can be effectively applied to enhance the fatigue performance of parts with up-facing inclined surfaces of variable initial surface roughnesses resulting from different building with different layer thicknesses. This is particularly relevant for recent promising approaches that utilize a high layer thickness to increase build rates but severely increase the roughness [22].

This work advances a DLPBF strategy towards the high-speed building of up-facing inclined metal parts with an enhanced fatigue life made from M789, a novel tool steel powder. In situ surface quality improvement through laser remelting was applied to parts produced with different building strategies, namely a standard strategy using a layer thickness (LT) of 60 μm (LT60), a high-productivity strategy using a high LT of 120 μm (LT120), and a hull-bulk strategy using an LT of 60/120 μm (HB60/120). A three-point bending fatigue campaign was conducted to evaluate the fatigue life of such treated samples in comparison to reference samples with as-built (AB), electrical discharge machined (EDM), and mechanically polished (P) surface conditions. A comprehensive analysis of the surface morphology is presented, quantifying the surface roughness, the critical stress concentration factors, and the lateral size of the dominant surface features. The strategy’s capability to simultaneously enhance productivity, surface quality, and fatigue life is demonstrated.

2. Materials and Methods

2.1. Sample Manufacturing

The samples were fabricated using an innovative DLPBF machine setup. The setup consisted of a commercially available ProX^® DMP 320 machine (3DSystems, Leuven, Belgium) equipped with a continuous-wave laser with a nominal spot size of d_1/e² = 90 μm and a central wavelength of 1070 nm. The machine was further equipped with a nanosecond pulsed-wave laser with a nominal spot size of d_1/e² = 50 μm and a central wavelength comparable to the continuous-wave laser. The laser optics setup enabled sequential processing with either the continuous-wave or pulsed-wave laser. Slicing and hatching were performed using 3DXpert (Oqton, Bremen, Germany). The material selected was a novel tool steel powder, BÖHLER AMPO M789, from voestalpine Böhler Edelstahl GmbH & Co KG (Kapfenberg, Austria). The chemical composition is specified by the supplier in [23].

Figure 1a depicts the geometry that was used to fabricate the parts using three different layer thicknesses (LT): a standard LT of 60 μm (LT60), a high LT of 120 μm (LT120), and a hull-bulk LT of 60/120 μm (HB60/120. The LT60 samples were manufactured using the parameters recommended by 3D Systems [24]. To increase the productivity, the samples were manufactured by employing a recently proposed high-speed LPBF strategy using LT120 [22], which comes at the expense of reduced surface quality. Additional samples were manufactured through an HB60/120 strategy adapted from [22], aiming to restore the surface quality to the LT60 level by combining LT60 for the hull region of the samples and LT120 for the bulk. The hull thickness was set to 1000 µm, with an overlap of 150 µm between the hull and the bulk. In addition, 12 up-safety layers were used to reach a stable LT60 up-facing surface quality.

All samples underwent the same post-processing after LPBF. The vertical side surfaces were milled to focus the investigations on the curved surface (Figure 1b). The horizontal top and bottom surfaces that were in contact with the rollers were ground to ensure parallelism (Figure 1b,c). The samples underwent a heat treatment consisting of solution annealing at 1000 °C for 1 h followed by aging at 500 °C for 3 h in a nitrogen atmosphere. Consequently, in this study the term ‘as-built’ refers to the surface condition. This heat treatment yielded consistent hardness values across the investigated sample conditions of H_LT60 = (51.7 ± 0.5) HRC, H_LT120 = (51.5 ± 0.5) HRC, and H_HB60/120 = (51.7 ± 1.2) HRC, each obtained from 10 measurements using the Wilson 4-JR equipment with a 150 kgf load. The hardness values are consistent with those available for solution-annealed and aged M789 material [24,25,26]. Average sample densities of ρ_LT60 = (99.73 ± 0.03)%, ρ_LT120 = (99.74 ± 0.03)%, and ρ_HB60/120 = (99.69 ± 0.03)% based on eight samples per condition were measured using the Archimedes principle with a theoretical density of 7.715 g/cm³.

2.2. Dual-Laser Powder Bed Fusion Strategy

The proposed DLPBF strategy was specifically developed for the treatment of up-facing inclined surfaces. The DLPBF strategy was based on previous findings in [18], and in this study was further developed to enable the treatment of samples made from a novel tool steel, M789, with variable initial surface states resulting from building using different layer thicknesses. The strategy consisted of three steps, as illustrated in Figure 2. Firstly, the samples were built through LPBF (Figure 2a). Secondly, after completing the last layer, pulsed-laser-induced shock waves were used to remove the powder that was covering the up-facing curved surface (Figure 2b). The full scan sequence for the powder removal scan passes is illustrated in Figure 2b. The treatment happened in different sections with the focal plane manually adjusted by Δf (Figure 2b). This adjustment was necessary since treating an inclined surface with a static focal plane leads to constant laser defocusing along the surface slope, reducing the efficiency of the shock waves for powder removal [18]. Thirdly, laser remelting of the newly exposed up-facing surface was performed using four scan passes with the scan orientation indicated in Figure 2c. The processing parameters were selected based on previous research on the M789 material [27], using for powder removal an average laser power of 45 W, a scan speed of 500 mm/s, a hatch spacing of 40 μm, a pulse repetition rate of 100 kHz, and a pulse duration of 30 ns, and for remelting a laser power of 200 W, a scanning speed of 500 mm/s, and a hatch spacing of 80 μm.

To evaluate the effectiveness of the proposed DLPBF strategy, the fatigue performance of the laser remolten (LR) samples was compared to reference the samples in as-built (AB), electrical discharge machined (EDM), and mechanically polished (P) surface conditions. Both post-AM surface treatments were performed by the company Deceuninck (Hooglede-Gits, Belgium).

2.3. Surface Characterization

The surface quality was assessed through tactile profilometry with a Mitutoyo Formtracer CS-3200. It was evaluated on the up-facing curved surface based on 100 profiles spaced by 20 μm across an evaluation length of 10 mm, using a probe with a 2 µm tip radius. The primary profiles were filtered according to ISO 16610–21 [28], with short and long cut-off wavelengths of λ_s = 2.5 μm and λ_c = 2.5 mm, respectively.

The surface parameters [R_a, k_t,crit, R_al, R_dq] were calculated based on the aforementioned tactile profiles. The arithmetic mean roughness Ra and root mean square gradient R_dq were calculated according to ISO 21920–2 [29]. The critical stress concentration factors k_t,crit were evaluated according to the method described in [16]. This method treats the profiles a(x) as a function of cosine components using Equation (1), with A_i, λ_i = 1/f_i, and ϕ_i being the ith cosine amplitude, wavelength, frequency, and phase, respectively. The profiles were filtered with a short wavelength cut-off filter of λ_s = 25 μm with a Gaussian filter according to ISO 16610-21 [28] before the corresponding stress concentration profiles k_t(x) were calculated with Equation (2). The stress concentration factor at each valley’s location, defined as the local minimum in a(x), were extracted from k_t(x). The critical notches were identified as those with a probability exceeding 95% in the generalized extreme value distribution according to [17].

$$a\left(x\right)={\displaystyle \frac{A_0}{2}}+\sum _{i=1}^n{A}_i\cos ({\displaystyle \frac{2\pi x}{{\lambda }_i}}+{\varphi }_i)$$

(1)

$$k_t\left(x\right)=1-\sum _{i=1}^n{\displaystyle \frac{{4\pi A}_i}{{\lambda }_i}}\cos ({\displaystyle \frac{2\pi x}{{\lambda }_i}}+{\varphi }_i)$$

(2)

The lateral dimensions of the dominant surface features were determined by the autocorrelation length R_al according to ISO 21920–2 [29], defined as the horizontal distance over which the autocorrelation function decays to 0.2 [30]. A smaller R_al indicates dominant fine features with rapid spatial variation, while a larger R_al denotes dominant coarser features with more gradual changes. The results presented in

Table 1. Quantification of surface parameters for the investigated surface conditions, including roughness Ra, critical stress concentration factor k_t,crit, autocorrelation length R_al, and root mean square gradient Rdq. The results represent mean values and the corresponding standard deviation between the samples.

	Surface Condition
Surface Parameters	AB			LR			P	EDM
	LT60	LT120	HB60/120	LT60	LT120	HB60/120	LT60	LT60
Ra/μm	14.0 ± 1.4	25.8 ± 4.2	15.7 ± 1.8	5.6 ± 1.0	18.7 ± 6.5	6.5 ± 0.7	0.21 ± 0.03	0.70 ± 0.04
kt,crit	2.14 ± 0.09	2.38 ± 0.20	2.27 ± 0.10	1.24 ± 0.02	1.60 ± 0.21	1.33 ± 0.04	1.03 ± 0.01	1.27 ± 0.01
Ral/μm	128 ± 11	307 ± 97	260 ± 50	474 ± 64	466 ± 125	508 ± 24	1182 ± 715	45 ± 15
Rdq	0.43 ± 0.02	0.48 ± 0.07	0.41 ± 0.02	0.12 ± 0.01	0.17 ± 0.01	0.12 ± 0.01	0.08 ± 0.0	0.19 ± 0.0

for k_t,crit and R_a represent mean values with corresponding standard deviations based on the samples included in the fatigue tests. The results reported for R_al and R_dq are based on five samples per condition.

The chamfer observed on the mechanically polished samples was characterized through tactile surface profile measurements. The chamfer height was defined as the maximum height difference between the part’s edge and the investigated surface within the middle section. The characterization was conducted using 10 tactile profiles spaced by 100 μm taken from the middle section.

2.4. Fatigue Testing

The three-point bending fatigue tests were performed on an Instron Electropuls E10000 machine equipped with a 10 kN dynamic load cell. The tests were performed with a stress ratio R of 0.1 and a frequency of 30 Hz. The samples were subjected to sinusoidal compression–compression loading, resulting in cyclic tensile stress on the surface under investigation (Figure 1a,c). The tests were performed at three stress levels until sample failure or run-out at N_f = 2 × 10⁶ load cycles without failure. The stress levels were chosen as a percentage of the materials’ yield strength, with σ_y,LT60 = 1758 MPa based on the testing of three round type 4 tensile specimens according to the ASTM E8 standard, with the values in agreement with [24], and σ_y,LT120 = 1737 MPa [22]. For the HB60/120 condition, the same stress levels were applied as those for the LT60 standard. The applied loads F required to achieve the maximum stress levels S_max were calculated using Equation (3), with the area moment of inertia I of a beam determined from Equation (4):

$$F={\displaystyle \frac{8·S_{max}·I}{l·h}}$$

(3)

$$I={\displaystyle \frac{w·h^3}{12}}$$

(4)

The bending span l was set at a constant of 43.8 mm. The sample dimensions (width w, height h) were measured at the critical cross-section. The fatigue data is presented with SN curves for the investigated sample conditions produced with a standard LT60 (AB, LR, P, EDM) and a high LT120 (LR, AB), based on a minimum of 10 and 8 samples for each condition, respectively. Additional fatigue data was collected for the samples produced through a hull-bulk strategy (HB60/120), with five samples tested for both the AB and LR conditions. The number of cycles to failure reported represent the arithmetic mean values and corresponding standard deviations for the samples tested at each respective stress level. The statistical significance of the differences in fatigue life between the sample conditions was assessed using p-values, with the reported improvements meeting a threshold of p < 0.05 (95% confidence level), and p < 0.32 (68% confidence level) for the polished samples.

3. Results

3.1. Surface Quality

Figure 3a–d present the surface profiles for the investigated sample, while Table 1 lists the calculated surface parameters and their corresponding standard deviations. The AB samples exhibited the highest surface roughness of R_a = 14.0 μm. LR reduced the roughness to R_a = 5.6 μm, which corresponds to a reduction of 60%. Mechanical polishing and electrical discharge machining achieved significantly lower roughness values of R_a = 0.2 μm and R_a = 0.7 μm, respectively.

The k_t,crit trends observed for the investigated sample conditions were similar to those for R_a (Table 1); however, the differences between the investigated surface conditions were less pronounced. The AB and P reference samples displayed the highest and lowest values of k_t,crit = 2.1 and k_t,crit = 1.0, respectively. Notably, the LR samples exhibited a similar k_t,crit = 1.2 to the EDM samples (k_t,crit = 1.3), despite exhibiting a R_a value that was eight times higher. While the roughness parameter takes only the depth of the valleys into account, k_t,crit considers the notch morphology, including both the valleys’ depth and radius. Laser remelting redistributes molten material from roughness peaks to valleys. The molten material partially fills the valleys, resulting in a smoothened surface morphology compared to the initial AB state, which is shown by a reduction of k_t,crit by 40%.

The differences in the surface morphology among the investigated samples are evident from the tactile surface profiles displayed in Figure 3a–d. The reference EDM samples showed abruptly changing, spiky surface features (Figure 3b). A similar morphology was observed for the AB reference samples, but with higher amplitude and lower spatial frequency (Figure 3a). In contrast, the LR samples exhibited a gradually changing surface profile characterized by low spatial variation (Figure 3a). This is reflected by the autocorrelation length R_al, which quantifies the lateral size of the dominant surface features. The feature size increased for EDM, AB, and LR, with corresponding R_al values of 45 μm, 128 μm, 470 μm (Table 1). For the P samples, the significantly larger autocorrelation length of 1182 μm suggests the absence of dominant surface features. These observations were further supported by the root mean square gradient parameter R_dq, which showed an increasing local surface slope across the surface conditions, following the sequence P, LR, EDM, and AB samples (see Table 1).

Figure 3c,d and Table 1 further display the surface morphology for the samples produced with a high layer thickness of LT120 and a hull-bulk strategy of HB60/120. The AB samples produced by the LT120 strategy exhibited the highest observed R_a = 26 μm and k_t,crit = 2.4, representing an increase of 84% and 11% compared to the standard LT60 approach. This aligns with previous work that demonstrated significant improvements in productivity with this LT120 strategy, albeit at the expense of the surface quality [22]. The AB samples produced using the HB60/120 strategy exhibited R_a = 16 μm and k_t,crit = 2.3, which widely restored the surface quality to LT60 level. Laser remelting improved the surface quality for both sample conditions, shown by the reduction of R_a and k_t,crit by 28% and 33% (LT120) and 59% and 41% (HB60/120). Moreover, Table 1 presents the R_al and R_dq values for the investigated conditions, reinforcing the previous observation that laser remelting consistently resulted in a smoothened surface morphology displayed by an increased feature size (R_al) and reduced local surface slope (R_dq) compared to the AB reference samples.

3.2. Fatigue Performance

Figure 4a presents the three-point bending fatigue test results for the investigated sample conditions produced with a standard layer thickness (LT) of 60 μm. The LR samples showed the best performance at the comparably low and intermediate stress levels of 967 MPa and 1143 MPa, followed by the P, EDM, and AB samples. At the highest stress level of 1317 MPa, the P samples exhibited an improved performance over the LR samples.

The AB samples exhibited the poorest fatigue performance across the investigated stress levels. They failed consistently below N_f < 100 × 10³ load cycles at N_f = (64 ± 4) × 10³ (at 791 MPa), N_f = (15 ± 5) × 10³ (at 967 MPa), and N_f = (10 ± 4) × 10³ (at 1143 MPa). In comparison, the LR samples recorded consistent run-outs without failure after N_f = 2·10⁶ load cycles at an applied stress level of 967 MPa. Moreover, the LR samples only failed after N_f = (367 ± 310) × 10³ load cycles at a stress level of 1143 MPa, corresponding to a fatigue life improvement compared to the AB reference by a factor of 36.

The EDM reference samples showed a consistent fatigue performance, failing at the stress levels of 967 MPa, 1143 MPa, and 1317 MPa at N_f = (79 ± 23) × 10³, N_f = (36 ± 12) × 10³, and N_f = (22 ± 14) × 10³ load cycles, respectively. In comparison, the LR samples outperformed the EDM reference sample across the respective stress levels, showing an improvement in fatigue life by up to a factor of 10.

Despite the large scatter in the data from the P reference samples, they demonstrated a notably increased fatigue life compared to the AB and EDM reference samples across all stress levels. Furthermore, the P reference samples exhibited an enhanced fatigue life compared to the LR samples at the high stress level of 1317 MPa by almost a factor of four. In contrast, the LR samples showed an enhanced fatigue life at the stress levels of 967 MPa and 1143 MPa. At corresponding stress levels, the LR samples were consistently run-out without failure at N_f = 2 × 10⁶ load cycles, exhibiting a fatigue life that was improved by a factor of three.

Figure 4b further presents the fatigue performance of the samples produced with a high layer thickness of 120 μm (LT120) and a hull-bulk strategy of 60/120 μm (HB60/120). The LR samples demonstrated the dominant performance, with the HB60/120 outperforming LT120, followed by their respective AB reference samples.

The laser remelting of the samples produced with a high layer thickness of LT120 consistently achieved an enhanced fatigue life compared to the AB references. While the AB reference samples made from LT120 showed early failure with N_f = (7 ± 2) × 10³ at a stress level of 955 MPa, one LR sample ran out at N_f = 2 × 10⁶ load cycles and other samples failed at N_f = 50 × 10³ and N_f = 68 × 10³ load cycles. Hence, this corresponded to an improvement in fatigue life by a factor of eight. Similarly, the reference AB samples produced through the HB60/120 strategy failed consistently after N_f = (17 ± 4) × 10³, whereas their LR sample counterparts failed at N_f = (434 ± 267) × 10³, representing an improvement in fatigue life by a factor of 25. Furthermore, the significant performance increase achieved through the remelting of the LT120 and HB60/120 samples is well demonstrated by their ability to compete with the reference samples produced using the LT60 approach. In fact, they exhibited a similar or improved fatigue life compared to the reference EDM samples and outperformed the AB LT60 reference. Specifically, the laser remelting samples that were built using LT120 had a fatigue life enhancement compared to the AB LT60 reference by up to a factor of four.

4. Discussion

The proposed DLPBF strategy enables significant improvements in fatigue life compared to the investigated reference samples. It is evident that the clear reduction in surface roughness R_a compared to the as-built AB reference by up to 60% contributes to the increase in performance. However, the R_a parameter does not sufficiently explain the increased performance compared to the reference samples that underwent surface finishing as they have a roughness that is smaller by several orders of magnitude. Similarly, it does not explain the improved performance in comparison to the specific AB sample conditions that exhibit a similar roughness. Given the similar sample quality, as evidenced by the consistent values for density, hardness, and yield strength across the investigated sample conditions, this strongly suggests that differences in the surface morphology are responsible for the observed fatigue performance.

The critical stress concentration factor k_t,crit considers the morphology of surface valleys through the ratio of depth-to-radius. Surface valleys act as micro-notches that locally increase the stress and hence the risk of crack initiation from this site. Consequently, a deep and steep valley morphology corresponds to a high k_t,crit value, whereas a shallow and gradual morphology corresponds to a low k_t,crit value. The LR samples displayed a slightly lower k_t,crit value compared to the reference EDM samples, despite exhibiting an R_a value larger than a factor of eight. This is in line with the surface profiles depicted in Figure 3, illustrating a smoothened, gradually changing surface morphology for the LR samples (Figure 3a). In contrast, the EDM sample exhibited numerous abruptly changing surface features that are likely to act as stress concentrators under loading (Figure 3b). Examining this further, the lateral size of the dominant surface features was quantified using the autocorrelation length R_al (Table 1). The LR exhibited wider surface features compared to the EDM samples, which displayed a dominant feature width of R_al = 45 μm and R_al = 474 μm, respectively. In addition, the EDM samples exhibited steeper local surface slopes by 58%, quantified by R_dq (Table 1). Hence, the EDM samples were densely packed with abruptly changing surface features, corresponding to a surface feature density that was increased by almost a factor of 11. This is linked to a higher number of potential crack initiation points and an unfavorable load distribution, both of which raise the likelihood of early failure.

Despite the large scatter observed for the data from the polished reference samples, it is evident that they display a significantly increased fatigue life at the highest stress level compared to the LR samples, whereas the LR samples are superior at the lowest stress level. The scatter is likely attributed to the presence of a chamfer located at the vertical side surface of the investigated curved surface, underscoring the challenges associated with the post-AM surface finishing of complex geometries. The manual polishing operation resulted in a maximum height reduction of 155 μm to 590 μm for the smallest and largest observed chamfers, respectively, captured by tactile profile measurements. Consequently, the reduced cross-section led to a local stress increase at the chamfers of approximately 6% and 27%, respectively, which likely reduced the fatigue life and increased the scatter. Moreover, following previous explanations, it is hypothesized that the gradually changing surface morphology of the LR samples is beneficial compared to abruptly changing fine features observed on the P surfaces (Figure 3a,b). This is supported by a k_t,crit value that approaches that of the P reference samples, despite having a greater roughness by a factor of 28.

In an effort to simultaneously improve productivity, surface quality, and fatigue performance, LR was applied to samples that were built with a high layer thickness of 120 μm and a hull-bulk strategy of 60/120 μm, respectively. As displayed in Figure 4b, such treated samples demonstrated a significantly improved performance compared to the investigated AB reference samples, including those made from the standard LT60 approach. Furthermore, the LR samples displayed a similar or significantly enhanced performance compared to the EDM reference samples. Given that the LR samples made from LT120 had an R_a about 27 times higher than that of the EDM reference samples, the performance is likely attributed to the gradually changing surface morphology, as displayed in Figure 3c, in contrast to the abruptly changing surface features (Figure 3b). This is supported by the LR samples made from LT120 having a comparable k_t,crit value and a reduced density of dominant surface features by a factor of 10 according to R_al (Table 1) compared to the EDM reference samples. Similar explanations apply to the performance improvement for the LR HB60/120 samples.

The production time for the proposed DLPBF strategy includes the build time through LPBF and additional time for the surface treatment. It is worth nothing that the production time depends on the number of samples to be treated, their size, and their design. In this case, the build time for a job including 16 samples using the standard LT60 approach was 170 min, compared to 94 min with the LT120 strategy. This corresponds to a reduction in build time by 81%. The surface treatment per sample required up to ~130 s for powder removal, ~30 s for remelting, and ~120 s for manually adjusting the focal plane. For the LT120 job, where eight samples were laser remolten, the resulting production time was 131 min. Hence, despite the additional surface treatment, this represents a 30% shorter production time compared to the standard LT60 approach. Simultaneously, the LR samples produced using LT120 exhibited an enhanced fatigue life compared to the AB LT60 samples by up to a factor of four. Moreover, as the DLPBF treatment happens directly in the AM machine, it has the potential to avoid or limit the need for post-AM surface finishing along with the associated costs and time. Despite its advantages, it is important to point out that this scanning strategy involves several manual operations to adjust the focal plane to avoid efficiency losses (see Figure 2), making it impractical for treating larger components. These limitations were recently addressed, enabling the fully automated treatment of large and complex-shaped components directly during building [27].

5. Conclusions

This work further developed a dual-laser powder bed fusion strategy that demonstrated simultaneous improvements in the productivity, surface quality, and fatigue life of parts with up-facing inclined surfaces made from M789—a novel tool steel powder. The proposed strategy combines building using a high layer thickness of 120 μm and subsequent in situ surface quality improvement. This is achieved through the selective removal of the powder covering the inclined surfaces via laser-induced shock waves and subsequent laser remelting. The main findings are as follows:

• Laser remelting demonstrated significant improvements in surface quality and hence fatigue life when applied to samples with variable initial surface states resulting from building strategies using different layer thicknesses. The fatigue life was consistently improved across the investigated as-built conditions, reaching a performance increase of up to a factor of 36.
• This significant improvement in fatigue life compared to the as-built reference samples was attributed to a smoothened surface morphology after remelting, shown by a reduction in surface roughness R_a and critical stress concentration factor k_t,crit by up to 60% and 40%. Moreover, the smoothened surface morphology exhibited gradually changing features that were wider by up to a factor of four, as displayed by the autocorrelation length R_al, compared to the abrupt and rapidly changing surface features observed for the as-built reference samples.
• Laser remelting enabled notable improvements in fatigue life compared to the electrical discharge machined reference samples by up to a factor of 10, despite having a roughness 8 times greater. This was similarly quantified by the smoothened surface morphology of the laser remolten samples. Moreover, the laser remolten samples demonstrated an enhanced fatigue life compared to that of the mechanically polished reference samples at relatively low and intermediate stress levels, whereas the mechanically polished samples dominated at higher stress levels.
• In an attempt to combine increased productivity, surface quality, and, hence, fatigue performance, laser remelting was applied to samples fabricated using a high layer thickness of 120 μm. While this approach significantly reduced the build time compared to the standard strategy using a layer thickness of 60 μm, it also considerably deteriorated the as-built surface quality. The laser remelting samples produced using a high layer thickness of 120 μm demonstrated an improvement in fatigue life compared to the as-built reference samples produced using the standard layer thickness of 60 μm by up to a factor of four, while simultaneously reducing the production time by 30%.