Investigation and characterization of a local remelting process for the improvement of surface properties resulting from extreme high-speed directed energy deposition

In the laser based directed energy deposition (DED-LB) process, a melt pool is formed by the irradiation of a laser beam on a metallic surface and additive material, either in the form of powder or wire, is fed to and melted in the melt pool. The additive material is deposited in the form of a weld bead after solidification. This process principle allows DED-LB technology to be flexibly applied as an additive manufacturing, modification, coating, as well as repair process for metallic components.^1,2 A modified variant of the DED-LB process is the extreme high-speed DED (German acronym—EHLA), which is characterized by feed rates of up to 200 m/min and coating thicknesses in the range of 100 μm.^3–5 As a comparison, the usual DED-LB feed rates are in the range of 1 m/min with coating thicknesses of >500 μm.¹ Recent developments of highly dynamic machine handlings enabled the transfer of the 2D rotational symmetric EHLA to the application in 3D (EHLA3D) so that the EHLA process can also be utilized for additive manufacturing, modification, free-form coating, and repair.^6–8 Still, as a modified variant of DED-LB, the resulting EHLA coatings and additively manufactured parts require subtractive postprocessing steps as the as-built surface quality does not meet the requirements for most industrial applications. To reduce the expense for postprocessing, the surface quality of DED-LB surfaces can be improved by a laser based remelting process (see Fig. 1), in which the surface is melted by a laser beam within a depth of typically 1–100 μm. The melt pool is smoothed out by surface tension and solidifies in the smoothed condition, resulting in a polished surface.^9–11 

Depending on the initial roughness of the component, either micro- or macro-laser remelting is applied for the polishing process. Surfaces with a spatial structural wavelength of λ < 80 μm can be effectively processed by microlaser remelting, which is characterized by a discrete pulse duration of several 100 ns and a scanning velocity of >100 mm/s.^9,11 For additively manufactured surfaces, the macrolaser remelt with a continuous wave process is more relevant as the as-built surfaces tend to have a higher surface roughness.¹² According to the state-of-the-art, the following macroremelt parameters have an impact on the resulting surface structure:

• – Scanning speed v_f: after a material specific threshold, an increasing scanning speed results in an increasing surface roughness. This can be explained by the formation of undercuts at the edge of the remelted tracks at higher scanning speeds. The depths of the undercuts increase with the scanning speed and are incorporated into the surface structure, resulting in an increased surface roughness.¹³ 

• – Beam power P_l: a too high beam power or energy input results in the evaporation of the remelted material and in an increased surface roughess.⁹ 

• – Beam diameter d_B: the applied beam diameter defines which wavelength λ of the surface structure can be remelted. The size of the beam diameter needs to be at least half of the wavelength of the surface structure. A further increase of the beam diameter results in a bigger melt pool dimension and a higher surface roughness, as extended melt pools allow an increased movement of the melt.^9,13

• – Hatch distance/remelt overlap: at a too low hatch distance of ∼96%, the extended, multiple remelted areas result in an increased surface roughness. At too high hatch distances, the roughness of the remelted track becomes the dominant surface structure, which also results in an increased surface roughness.

Typical remelt parameters for laser powder bed fusion (LPBF) surfaces are a beam diameter of 125 μm ≤ d_b ≤ 800 μm and scanning speeds v_f of up to 400 mm/s.^12,14–17 According to the literature, LPBF surfaces with an initial roughness of Sa ≈ 21 μm can be reduced to Sa < 1 μm by a remelting process.¹⁸ 

In contrast to the LPBF remelting studies, a remelting process for DED-LB can be conducted in-process with the same processing optics and without any further auxiliary equipment. Hence, the optical system consists of a moving processing optics instead of a conventional scanner type system. According to the studies, the surface roughness, Sa, which results from conventional DED-LB, can be reduced from Sa ≈ 10.3 μm to Sa ≈ 1.9 μm¹⁹ and from Sa ≈ 23.5 μm to Sa ≈ 9 μm for IN718.²⁰ The tested range of parameters was a beam power of P_L < 500 W and feed rates of v_f< 3.5 m/min. Recent studies also applied the remelt process on EHLA coated surfaces and verified the decreased surface roughness and improved corrosion resistance of the remelted surfaces. Depending on the process parameters, the surface roughness of IN625 EHLA coatings can range between 4.7 μm < Sa < 37 μm.²¹ For IN625, the initial roughness of Sa ≈ 13.3 μm could be reduced to Sa ≈ 1.8 μm by applying a remelt parameter of v_f = 7 m/min and a hatch overlap of 60%.²² The reduction of surface roughness to Sa ≈ 8.9 μm was also demonstrated for a CuNi EHLA coating with a remelt parameter of v_f = 5 m/min and a hatch overlap of 50%.²³ The current studies already demonstrate the improvement of the surface roughness by a remelting process; however, only a limited variation of the remelting parameters was investigated. In this work, the variation of the remelting parameters is expanded to further investigate the influence between the remelting parameters and the resulting surface roughness, Sa, and the surface waviness, Sa,w. Furthermore, due to limited availability of machines capable of the EHLA3D process, no remelting studies are available for the additively manufactured EHLA3D surfaces. As such, this work covers a remelt process parameter study with feed rates of up to v_f = 30 m/min applied on EHLA coatings as well as bulk specimens deposited by EHLA3D.

All experiments were conducted on a modified five-axis CNC prototype machine developed by Makino Asia Pte Ltd (see Fig. 2). The machine can operate the processing optics with a maximum feed rate of v_f = 30 m/min in x–y–z.

The optical setup consists of a Beo D70 processing optics from Trumpf GmbH and is combined with a LDF4000-8 beam source from Laserline GmbH. The laser focus diameter of d_focus = 400 μm results from an optical aspect ratio of 2:1, a laser fiber core diameter of d_fiber = 200 μm, and a beam parameter product of BPP = 8 mm mrad. The beam diameter used for the process can be set up by the motorized collimation system of the processing optics. The powder nozzle was a Highno 4.0 from HD Sonderoptiken and is specified with a stand-off distance of 9 mm.

The additive powder material for this study was 316L with a specified particle distribution between 15 and 45 μm. The substrate plates were hot rolled 1.4301 steel with a thickness of 10 mm.

In scope of this work, coating as well as bulk specimens were deposited by EHLA3D with 316L stainless steel. After the measurement of the reference surfaces by white light interferometry (WIM), the surfaces were used for the remelting process parameter study. The resulting remelted surfaces were analyzed by WIM and the surface roughness, Sa, and waviness, Sa,w, were compared to the reference values. The remelted probes with the lowest achieved surfaces roughness were also analyzed by metallographic cross sections for a qualitative analysis of the resulting microstructure.

As the first stage, the probe specimens were deposited by using the following EHLA3D parameters (Table I).

To evaluate the surfaces of coatings as well as the additively manufactured surfaces, the following probe specimens were defined and deposited (Table II).

For both types of specimens, a bidirectional path planning strategy was applied with an additional 90° cross-hatching for the coating specimen (see Fig. 3).

In the second stage, the surfaces of the specimens were remelted with a variation of the following remelting parameters (see Table III). For each set of process parameters, the indicated variation of beam power was applied with an increment of 100 W. Only the process parameter set resulting in the lowest Sa value was considered for the evaluation of the parameter study and used as a data point. For the hatch distance, the overlap of the beam diameter is used, which also results in the track overlap. A variation of the incidence angle of the beam will have an influence on the resulting surface roughness.²⁴ However, the incidence angle of the beam is perpendicular to the surface in this study, as the processing optics is moved laterally.

Each remelted surface had a dimension of 17 × 17 mm² (see Fig. 4).

The resulting reference and remelted surfaces were analyzed with a Nexview^TM NX2 white light interferometer from Zygo. For each probe, the surface roughness, Sa, and the waviness, Sa,w, were measured. For a more representative evaluation of the surface, the surface related roughness, Sa and Sa,w, instead of the line roughness, Ra, and waviness, W, were selected for the evaluation. This ensures a direction independent evaluation of the measurement area. By measuring the surface roughness, Sa, and waviness, Sa,w, both ranges of surface structures are considered in the evaluation, as different bandpass filters are applied for each surface value. As such, spline type bandpass filters were applied to extract the surface roughness, Sa, and waviness, Sa,w. The following measurement configuration was applied for the white light interferometer measurements (Table IV).

Due to the number of data points, one surface measurement was conducted for each remelted surface. The deviation of the measurement is estimated by σ = ±1.5 μm with the applied measurement configuration.

The surface specimen with the lowest achieved surface roughness, Sa, was additionally evaluated with a metallographic cross section.

The deposition of EHLA3D coatings and bulk specimens results in different surface qualities. The top surface of a coating is mainly characterized by the topography of the weld beads while the side surfaces of the wall specimen had an increased roughness, which likely result from the characteristic particle overspray during the deposition. Due to this, an overall higher roughness and waviness value are measured for the additively manufactured wall specimens (see Table V and Fig. 5).

For the evaluation of the effect of the hatch distance on the resulting surface roughness, a hatch distance of 50% and 80% track overlap was tested. Within the small defined variation, a bigger remelted track overlap generally resulted in a lower surface roughness (see Fig. 6).

However, an increasing hatch distance also increases the waviness, Sa,w, in the case of the coating surface while the waviness of the wall surface remains nearly unaffected. A possible explanation of this effect can be deduced from the existing, conventional remelting studies, in which multiple, remelted tracks decrease the surface roughness. However, too high numbers of remelted areas result in a higher roughness. This might be the case for the surface specimen and surface structures with higher structural wavelengths, which would explain the increase of thewaviness.

The variation of the applied beam diameter indicates that the lowest Sa values are achieved with a beam diameter of d_B = 2 mm for both surface specimens. According to the waviness evaluation, the waviness increases with an increasing beam diameter for the coating surface while it decreases for the wall surface (see Fig. 7).

According to the literature of laser macro remelting, a possible explanation of the effect of the beam diameter on the surface roughness Sa is that the set beam diameter approaches the processable structural wavelength of the surface specimen. At d_B = 1 mm, the beam diameter is too small for the rough surface structure, while the beam diameter of d_B = 3 mm results in an extended melt pool dimension. Hence, a local minimum of the surface roughness Sa at d_B = 2 mm is reached. This explanation is confirmed by an excerpt of a line profile of each specimen (see Fig. 8). The structural wavelength of the surface specimen is λ ≈ 1.8 mm so that the increase of the beam diameter results in a decrease of the surface roughness by Sa ≈ 2 μm. Surface structures with a wavelength of λ ≈ 4 mm can be identified on the wall surface. Due to this, the change of the beam diameter from d_B = 1 to d_B= 2 mm has a greater influence on the resulting surface roughness, Sa. The surface waviness, Sa,w, indicates no local minimum for the coating surface. A possible explanation is the already low waviness value of the as-built surface so that the remelted track dimension becomes the dominant waviness structure.

The variation of the feed rates resulted in an increasing surface roughness and waviness with an increasing applied feed rate (see Fig. 9). For the wall specimen, only a feed rate until v_f = 5 m/min could be tested as higher feed rates resulted in not remelted areas.

Within the feed rate v_f variation, the lowest surface roughness and surface waviness were achieved with a feed rate of v_f = 2 m/min. A possible explanation is also provided by the state of the art of conventional remelting processes, in which an undercut is formed at higher feed rates. According to the previous studies, the depth of the undercuts increases with an increasing feed rate.¹³ Another mechanism for an increased surface roughness can be a higher solidification rate of the melt pool at higher feed rates so that the time for the smoothing mechanism by surface tension is not sufficiently provided. Still, both mechanisms need to be confirmed by subsequent studies, which involves, e.g., simulation and further evaluation of single melt track cross sections.

Due to the lower initial surface roughness of the coating surface, a lower beam power in the range of few 100 W is required to fully remelt the surface. The surface roughness also tends to increase at areas of too high and too low beam power, while the surface waviness stays comparable except after exceeding a beam power threshold [see Fig. 10(a)]. To fully remelt the wall surface, a higher regime of beam power is required to homogenize the initial rougher surface. In the case of the rough wall surface, the surface roughness further decreases with an increasing beam power [see Fig. 10(b)].

A possible mechanism for the decrease of the surface roughness with the applied beam power is that the temperature of the melt pool is also increasing, resulting in an improved surface smoothing effect by a lower viscosity. After the minimum, the surface roughness increases further, as a decreased viscosity results in an increased melt pool dynamic. Still, this mechanism needs to be verified in subsequent investigations of the microstructure. In the case of the wall surface, the surface roughness tends to further decrease with the increase of the beam power. The higher initial surface roughness, which is characterized by powder attachments and overspray, seems to be more robust against higher beam power.

In scope of the remelting parameter study, the following parameter sets achieved the lowest surface roughness and waviness values (Table VI).

A qualitative evaluation of the microstructure by metallographic cross sections indicates no major change of the resulting microstructure.

According to the metallographic cross section, the remelting depth for the coating specimen is ∼300 μm, while for the wall specimen, it is ∼1.5 mm. The remelted depths match to the applied beam power and the associated energy input of the remelting parameter. Still, the remelting depths are at least three times higher than the usual remelting depths of conventional laser macroremelting (Fig. 11).

In scope of this work, a remelting parameter study was conducted to reduce the surface roughness and surface waviness of coatings as well as bulk specimens deposited by EHLA3D. Following dependencies between the applied remelt parameters and resulting surface properties can be summarized:

• – An increasing overlap of a remelt track generally results in a smaller surface roughness and waviness, as the surface is evenly smoothed by more remelting passes. This outcome matches with the literature in which remelting experiments were conducted on surfaces deposited by DED-LB.

• – Within the tested beam diameters, the surface roughness first decreases toward a minimum and increases again with an increasing beam diameter. This matches to the literature of the macroremelting processes in which the beam diameter should be at least half the length of the structural wavelength λ of the surface. In case of this study, the applied beam diameter should be in the range between 1 < d_B < 3 mm.

• – The surface roughness increases with increasing feed rates. For the initially rough wall surface, the applied feed rate could not exceed 5 m/min as the energy input was not sufficient to fully remelt the rough surface area. Despite a resulting higher surface roughness, higher feed rates still can be considered for the remelting process as a feed rate of up to 30 m/min majorly increases the productivity of a remelt process.

• – Similar to the beam diameter, the surface roughness decreases toward a local minimum with an increasing beam power for the coating specimen. After the minimum, the surface roughness increases again with an increasing beam power. For wall surfaces, an increase of the surface roughness with further increasing beam power could not be observed within this study.

The parameter study validates that different sets of remelt process parameters are required for the remelting of different degrees of surface roughness. Due to this, two remelting parameter sets were, respectively, defined for EHLA3D coating and wall surfaces. For the coating surface, an improvement of the surface roughness by 76.6% could be achieved. The initially low surface waviness of the coating specimen could not be further improved by remelting. Due to a high initial surface roughness, the surface roughness was improved by 93.97% for the wall surface specimen. Furthermore, the surface waviness could be decreased by 63.85%.

The authors have no conflicts to disclose.

Min-Uh Ko: Conceptualization (lead); Data curation (lead); Formal analysis (lead); Funding acquisition (lead); Investigation (equal); Methodology (lead); Project administration (lead); Supervision (lead); Validation (lead); Visualization (lead); Writing – original draft (lead); Writing – review & editing (lead). Mikael Bueno Da Silveira: Conceptualization (supporting); Data curation (equal); Formal analysis (equal); Investigation (lead); Methodology (equal); Visualization (equal). Andres Gasser: Project administration (supporting); Supervision (supporting). Erwin Teichmann: Supervision (supporting). Thomas Schopphoven: Supervision (supporting). Constantin L. Häfner: Supervision (supporting).