Research and Optimization of Surface Roughness in Milling of SLM Semi-Finished Parts Manufactured by Using the Different Laser Scanning Speed

The paper studies the potential to improve the surface roughness in parts manufactured in the Selective Laser Melting (SLM) process by using additional milling. The studied process was machining of samples made of the AlSi10Mg alloy powder. The simultaneous impacts of the laser scanning speed of the SLM process and the machining parameters of the milling process (such as the feed rate and milling width) on the surface roughness were analyzed. A mathematical model was created as a basis for optimizing the parameters of the studied processes and for selecting the sets of optimum solutions. As a result of the research, surface with low roughness (Ra = 0.14 μm, Rz = 1.1 μm) was obtained after the face milling. The performed milling allowed to reduce more than 20-fold the roughness of the SLM sample surfaces. The feed rate and the cutting width increase resulted in the surface roughness deterioration. Some milled surfaces were damaged by the chip adjoining to the rake face of the cutting tool back tooth.


Introduction
The Additive Manufacturing (AM) technology, commonly known as 3D printing, allows to layer-by-layer parts manufacture. One of the AM variants is Selective Laser Melting (SLM) which can be an alternative for casting of metals. It involves using a laser beam for selective melting of metal powders of which a part is made. This technology gives new possibilities to manufacture machine parts, but is not flawless. The diagnosed flaws include porosity, incomplete powder melting, insufficient dimensional, and shape accuracy and high surface roughness [1][2][3][4][5].
The process parameters have a principal impact on the quality and properties of parts manufactured using the SLM method. The basic parameter that characterizes the SLM process is the energy density of laser beam. The laser beam energy density increases with the laser power increase, decrease of the scanning speed, decrease of the powder layer thickness, or decrease of the hatching space [6,7]. As the laser beam energy density increases, the density of parts made in the SLM process initially grows and then is reduced [7,8]. The increase of the laser beam energy density causes the temperature increase which results in the low viscosity and a large amount of liquid metal [1,9]. The temperature increase prolongs the residence time of the molten pool as a result of which the metal flows easily and fills the pores [10]. When the laser energy input is too low, large-sized irregular pores form due to the small amount of liquid and the high viscosity as well as the short lifetime of the molten pool. Excessively high laser beam energy density causes vaporization of the material, creating spherical pores [11]. The surface roughness of the SLM-made parts, in correlation to the porosity, initially decreases and then grows as the laser beam energy density increases [9,12]. Section 1 includes an analysis of the current knowledge. Section 2 gives the characteristics of the AlSi10Mg alloy and describes the research methodology. Section 3 analyses the experimental results, creates the mathematical model and selects the sets of optimum solutions. Section 4 includes the summary and conclusions.

Materials and Methods
The research includes the milling of semi-finished parts manufactured using the SLM technology. The SLM process was performed on a TruPrint 1000 machine (Trumpf, Ditzingen, Germany) with a 200 W fiber laser. The SLM parts were made of the powders of the AlSi10Mg alloy which is widely used in the automotive and aerospace industries. The powder grain sizes are given in Table 1 and the chemical composition of powder in Table 2. The SLM process parameters were chosen based on the literature review. The laser scanning speed v was variable in the 600-1400 mm/s range which allowed obtaining the laser beam energy density E d of 31-73 J/mm 3 . The remaining, constant SLM process parameters are presented in Table 3. The SLM process was used to manufacture semi-finished parts of the following dimensions: 45 mm in axis x, 5 mm in axis y, and 15 mm in axis z. Based on the analysis results of authors [21], the SLM semi-finished parts were subjected to milling on the x-z plane ( Figure 1).
The machining was performed on a CNC MiniMill2 machine tool (Haas, Oxnard, CA, USA) using a double-edged end mill cutter with 3 mm diameter and catalogue number 0068030A (Datron, Mühltal, Germany). The cutting tools with small diameter are more universal. These cutting tools give the possibility to perform pocket bottom with walls characterized small inner radius. The tool is characterized in Table 4. The machining zone was lubricated with an oil mist.
The machining parameters were chosen based on the recommendations of the cutting tool manufacturer. The feed rate f was variable in the 835-2045 mm/min range, and the machining width a e , varied from 0.829 to 1.67 mm. The machining depth was constant and equal to a p = 0.1 mm and the machining speed was v c = 848 m/min. A pneumatic spindle type 602CAT40 (Air Turbine Technology) with rotating speed of n = 90,000 rpm was used in the milling process. Up milling was analyzed.
Chamber temperature (°C) 26.8 (27% RF) Oxygen level (%) 0.11 The SLM process was used to manufacture semi-finished parts of the following dimensions: 45 mm in axis x, 5 mm in axis y, and 15 mm in axis z. Based on the analysis results of authors [21], the SLM semi-finished parts were subjected to milling on the x-z plane ( Figure 1). The machining was performed on a CNC MiniMill2 machine tool (Haas, Oxnard, CA, USA) using a double-edged end mill cutter with 3 mm diameter and catalogue number 0068030A (Datron, Mühltal, Germany). The cutting tools with small diameter are more universal. These cutting tools give the possibility to perform pocket bottom with walls characterized small inner radius. The tool is characterized in Table 4. The machining zone was lubricated with an oil mist.  The machining parameters were chosen based on the recommendations of the cutting to manufacturer. The feed rate f was variable in the 835-2045 mm/min range, and the machining wid ae, varied from 0.829 to 1.67 mm. The machining depth was constant and equal to ap = 0.1 mm and th machining speed was vc = 848 m/min. A pneumatic spindle type 602CAT40 (Air Turbine Technolog with rotating speed of n = 90,000 rpm was used in the milling process. Up milling was analyzed. The roughness was measured on a Talysurf Intra 50 profilometer (Taylor Hobson, Leiceste UK). To perform the surface roughness measurements, a measuring tip with a rounding radius of μm was used. The measurements were performed in the transverse direction to the machining mark (parallel to the cutting tool feed). The Ra and Rz parameters were determined in accordance with IS 4287. After surface roughness measurements, non-periodic primary profiles were obtained. The filter value was selected based on the values recommended for non-periodic profiles. Based on th obtained ranges of the surface roughness parameters values, the cut-off values λc = 0.8 mm and λs 2.5 μm were selected. The measurements were repeated five-fold. The microscopic observations the surface were performed by using the VHX-600 digital microscope (Keyence, Osaka, Japan) at 500 magnification.
The studies were conducted based on a central-compositional design of experiment. Th variables were analyzed on five levels. The measurement analysis results were examined on th Statistica software (13.1, StatSoft, Tulsa, OK, USA) for the α = 0.05 significance level.

Impact of the Laser Scanning Speed and Machining Parameters on the Surface Roughness
The surfaces of manufactured semi-finished parts were subjected to roughness measurement The best surface roughness (Ra = 3.55 ± 0.32 μm, Rz = 18.3±2.3 μm) was obtained for a semi-finishe part made with the lowest laser scanning speed v = 600 mm/s which corresponds to the laser bea energy density of Ed = 73 J/mm 3 . The surface roughness parameter increases with increasing the las scanning speed. For highest laser scanning speed v = 1400 mm/s (Ed = 31 J/mm 3 ) the values roughness parameters were Ra = 6.91 ± 0.65 μm and Rz = 38.3 ± 3.9 μm, respectively. The high standard deviations of the surface roughness parameters are observed for the higher laser scannin speeds, which indicates the lower homogeneity of the surfaces.
After milling of the SLM semi-finished parts, microscopic observations of the obtained surfac were performed. Microscopic photographs of the milled surfaces of SLM semi-finished par produced by using the various values of laser scanning speed v and, the next, milled with using th various values of feed rate f were presented in Figure 2a,b.
The photographs present surface fragments made during the last pass of the cutting too Therefore, only the traces made during the last tool pass are visible, not erased by the traces fro successive passes, which occurs during the face milling. The roughness was measured on a Talysurf Intra 50 profilometer (Taylor Hobson, Leicester, UK). To perform the surface roughness measurements, a measuring tip with a rounding radius of 2 µm was used. The measurements were performed in the transverse direction to the machining marks (parallel to the cutting tool feed). The Ra and Rz parameters were determined in accordance with ISO 4287. After surface roughness measurements, non-periodic primary profiles were obtained. The λ c filter value was selected based on the values recommended for non-periodic profiles. Based on the obtained ranges of the surface roughness parameters values, the cut-off values λ c = 0.8 mm and λ s = 2.5 µm were selected. The measurements were repeated five-fold. The microscopic observations of the surface were performed by using the VHX-600 digital microscope (Keyence, Osaka, Japan) at 500× magnification.
The studies were conducted based on a central-compositional design of experiment. The variables were analyzed on five levels. The measurement analysis results were examined on the Statistica software (13.1, StatSoft, Tulsa, OK, USA) for the α = 0.05 significance level.

Impact of the Laser Scanning Speed and Machining Parameters on the Surface Roughness
The surfaces of manufactured semi-finished parts were subjected to roughness measurements. The best surface roughness (Ra = 3.55 ± 0.32 µm, Rz = 18.3 ± 2.3 µm) was obtained for a semi-finished part made with the lowest laser scanning speed v = 600 mm/s which corresponds to the laser beam energy density of E d = 73 J/mm 3 . The surface roughness parameter increases with increasing the laser scanning speed. For highest laser scanning speed v = 1400 mm/s (E d = 31 J/mm 3 ) the values of roughness parameters were Ra = 6.91 ± 0.65 µm and Rz = 38.3 ± 3.9 µm, respectively. The higher standard deviations of the surface roughness parameters are observed for the higher laser scanning speeds, which indicates the lower homogeneity of the surfaces.
After milling of the SLM semi-finished parts, microscopic observations of the obtained surfaces were performed. Microscopic photographs of the milled surfaces of SLM semi-finished parts produced by using the various values of laser scanning speed v and, the next, milled with using the various values of feed rate f were presented in Figure 2a On same samples, defects are observed in the internal microstructure of the SLM semi-finished parts. The amount of defects increases with increasing the laser scanning speed. Analogous results were also observed in the previous papers [9,12,13,24].
The surface damages due to the cutting tool back tooth were observed on the surfaces. The probable mechanism of this phenomenon takes place in result adjoining chip to the rake face of cutting tool after the machining cycle is completed. Then, the chip moves under the cutting tooth and damages the surface. Such phenomena are observed also during the machining of other Al alloys. The analysis of the samples indicates that lower values of feed rate and lower values of laser scanning speed minimize the surface damage resulting from the cutting tool back tooth. The differences between the surfaces made using various milling widths are insignificant.
Based on the measurements of surface roughness parameters, the ANOVA was performed and the regression equations were determined. The ANOVA was performed in order to determine the significance of the impact of the analyzed process parameters on the surface roughness. The results of the ANOVA analysis for Ra and Rz roughness parameters are given in Tables 5 and 6, respectively. The process parameters which affect the surface roughness at the applied significance level of α = 0.05 are marked in bold typeface.  The photographs present surface fragments made during the last pass of the cutting tool. Therefore, only the traces made during the last tool pass are visible, not erased by the traces from successive passes, which occurs during the face milling.
On same samples, defects are observed in the internal microstructure of the SLM semi-finished parts. The amount of defects increases with increasing the laser scanning speed. Analogous results were also observed in the previous papers [9,12,13,24].
The surface damages due to the cutting tool back tooth were observed on the surfaces. The probable mechanism of this phenomenon takes place in result adjoining chip to the rake face of cutting tool after the machining cycle is completed. Then, the chip moves under the cutting tooth and damages the surface. Such phenomena are observed also during the machining of other Al alloys. The analysis of the samples indicates that lower values of feed rate and lower values of laser scanning speed minimize the surface damage resulting from the cutting tool back tooth. The differences between the surfaces made using various milling widths are insignificant.
Based on the measurements of surface roughness parameters, the ANOVA was performed and the regression equations were determined. The ANOVA was performed in order to determine the significance of the impact of the analyzed process parameters on the surface roughness. The results of the ANOVA analysis for Ra and Rz roughness parameters are given in Tables 5 and 6, respectively. The process parameters which affect the surface roughness at the applied significance level of α = 0.05 are marked in bold typeface. The regression Equations (1) and (2) The obtained regression equations indicate a good fit between the calculated and the measured values. The R 2 is 0.87 for parameter Ra and 0.78 for parameter Rz, respectively.  The regression Equations (1) and (2) The obtained regression equations indicate a good fit between the calculated and the measured values. The R 2 is 0.87 for parameter Ra and 0.78 for parameter Rz, respectively.     The regression Equations (1) and (2) The obtained regression equations indicate a good fit between the calculated and the measured values. The R 2 is 0.87 for parameter Ra and 0.78 for parameter Rz, respectively.   The increase of the laser scanning speed and the feed rate deteriorates the surface roughness of the milled surfaces. The increased milling width also deteriorates the surface roughness, but the trend is less significant in this case. The observed influences of the machining parameters are typical for the face milling.

Application of the Response Surface Methodology for the Optimization of the Analysed Process Parameters
The Response Surface Methodology (RSM) was used to optimize the parameters of the studied processes. The RSM is widely used in the optimization of technological processes. The mathematical model was created for significant factors determined based on the ANOVA. Figure 6 presents the impact of the laser scanning speed v and the feed rate f on the Ra and Rz parameters for the milling width of ae = 1.25 mm. Based on the mathematical model (Equations (1) and (2)), the least surface roughness was calculated for the lowest process values of v = 600 mm/s, f = 835 mm/min, and ae = 0.829 mm. The calculated values of surface roughness parameters in this case are Ra = 0.142 ± 0.013 μm and Rz = 1.043 ± 0.094 μm. Surface roughness measurements were made in order to verify the model, and the measured mean values of the surface roughness parameters values were Ra = 0.144 ± 0.008 μm and Rz = 1.104 ± 0.089 μm.
A surface of the lowest possible roughness is not always desirable. The specification of machine parts includes the limit values of surface roughness. The Ra and Rz values were specified for further The increase of the laser scanning speed and the feed rate deteriorates the surface roughness of the milled surfaces. The increased milling width also deteriorates the surface roughness, but the trend is less significant in this case. The observed influences of the machining parameters are typical for the face milling.

Application of the Response Surface Methodology for the Optimization of the Analysed Process Parameters
The Response Surface Methodology (RSM) was used to optimize the parameters of the studied processes. The RSM is widely used in the optimization of technological processes. The mathematical model was created for significant factors determined based on the ANOVA. Figure 6 presents the impact of the laser scanning speed v and the feed rate f on the Ra and Rz parameters for the milling width of a e = 1.25 mm. The increase of the laser scanning speed and the feed rate deteriorates the surface roughness of the milled surfaces. The increased milling width also deteriorates the surface roughness, but the trend is less significant in this case. The observed influences of the machining parameters are typical for the face milling.

Application of the Response Surface Methodology for the Optimization of the Analysed Process Parameters
The Response Surface Methodology (RSM) was used to optimize the parameters of the studied processes. The RSM is widely used in the optimization of technological processes. The mathematical model was created for significant factors determined based on the ANOVA. Figure 6 presents the impact of the laser scanning speed v and the feed rate f on the Ra and Rz parameters for the milling width of ae = 1.25 mm.  Based on the mathematical model (Equations (1) and (2)), the least surface roughness was calculated for the lowest process values of v = 600 mm/s, f = 835 mm/min, and a e = 0.829 mm.
The calculated values of surface roughness parameters in this case are Ra = 0.142 ± 0.013 µm and Rz = 1.043 ± 0.094 µm. Surface roughness measurements were made in order to verify the model, and the measured mean values of the surface roughness parameters values were Ra = 0.144 ± 0.008 µm and Rz = 1.104 ± 0.089 µm.
A surface of the lowest possible roughness is not always desirable. The specification of machine parts includes the limit values of surface roughness. The Ra and Rz values were specified for further optimization of the surface roughness parameters, as follows: for case I Ra ≤ 0.2 µm and Rz ≤ 1.4 µm, and for case II Ra ≤ 0.28 µm and Rz ≤ 1.8 µm. As the analysis covers the parameters of two different processes, the optimum solutions can be obtained in many ways. The diagram for the selecting optimum laser scanning speed and the feed rate at a constant cutting width a e = 1.25 mm are presented in Figure 7. The ranges of variable values of the analyzed process parameters were marked with blue area for case I and orange area for case II. optimization of the surface roughness parameters, as follows: for case I Ra ≤ 0.2 μm and Rz ≤ 1.4 μm, and for case II Ra ≤ 0.28 μm and Rz ≤ 1.8 μm. As the analysis covers the parameters of two different processes, the optimum solutions can be obtained in many ways. The diagram for the selecting optimum laser scanning speed and the feed rate at a constant cutting width ae = 1.25 mm are presented in Figure 7. The ranges of variable values of the analyzed process parameters were marked with blue area for case I and orange area for case II. Based on Figure 7, it is possible to choose one optimum solution, but the features of the manufactured part must be known. The choice of the optimum solution depends, inter alia, on the volume of the material which is added while the additive manufacturing and removed while the subtractive manufacturing. Without any additional assumptions, at this stage it is not possible to choose one optimum solution. Thus, only the relationship between the optimum laser scanning speed and the feed rate was chosen.

Conclusions
The mathematical model created with the use of RSM allows a prediction of the surface roughness parameters based on the values of the studied SLM and milling processes. As a result of using the model, the process parameters were chosen which enabled to obtain a surface with the roughness of Ra = 0.14 μm and Rz = 1.1 μm, respectively. The sets of optimum process parameters were also chosen, as a result of which it is possible to obtain the surface roughness of Ra ≤ 0.2 μm and Rz ≤ 1.4 μm or Ra ≤ 0.28 μm and Rz ≤ 1.8 μm, respectively (Figure 7). It addition, it was found that: 1. The milling allows the surface roughness of the SLM-manufactured semi-finish parts to be reduced more than 20-fold. 2. The impact of the laser scanning speed on the surface roughness of the SLM-manufactured semifinish parts was observed. The increase of the laser scanning speed in the analyzed range causes a deterioration of the surface roughness of SLM-manufactured semi-finished parts. 3. Defects in the internal microstructure of the SLM semi-finished parts manufactured by using high laser scanning speed were observed. 4. The impact of the laser scanning speed used in the SLM process on the obtained by milling surface was observed. The milled SLM semi-finish parts made at higher laser scanning speeds have higher surface roughness. It was also observed that the chip adjoining to the rake face of back cutting tooth causes the milled surface damages. This phenomenon intensifies as the laser scanning speed used in the SLM process increases. Based on Figure 7, it is possible to choose one optimum solution, but the features of the manufactured part must be known. The choice of the optimum solution depends, inter alia, on the volume of the material which is added while the additive manufacturing and removed while the subtractive manufacturing. Without any additional assumptions, at this stage it is not possible to choose one optimum solution. Thus, only the relationship between the optimum laser scanning speed and the feed rate was chosen.

Conclusions
The mathematical model created with the use of RSM allows a prediction of the surface roughness parameters based on the values of the studied SLM and milling processes. As a result of using the model, the process parameters were chosen which enabled to obtain a surface with the roughness of Ra = 0.14 µm and Rz = 1.1 µm, respectively. The sets of optimum process parameters were also chosen, as a result of which it is possible to obtain the surface roughness of Ra ≤ 0.2 µm and Rz ≤ 1.4 µm or Ra ≤ 0.28 µm and Rz ≤ 1.8 µm, respectively (Figure 7). It addition, it was found that: 1.
The milling allows the surface roughness of the SLM-manufactured semi-finish parts to be reduced more than 20-fold.

2.
The impact of the laser scanning speed on the surface roughness of the SLM-manufactured semi-finish parts was observed. The increase of the laser scanning speed in the analyzed range causes a deterioration of the surface roughness of SLM-manufactured semi-finished parts.

3.
Defects in the internal microstructure of the SLM semi-finished parts manufactured by using high laser scanning speed were observed. 4.
The impact of the laser scanning speed used in the SLM process on the obtained by milling surface was observed. The milled SLM semi-finish parts made at higher laser scanning speeds have higher surface roughness. It was also observed that the chip adjoining to the rake face of back cutting tooth causes the milled surface damages. This phenomenon intensifies as the laser scanning speed used in the SLM process increases. 5.
The impact of the studied milling parameters on the milled surface roughness was observed. The surface roughness increases as the feed rate grows. The increase of cutting width also increases the surface roughness, but to a lesser degree than the feed rate increases.
Funding: This research received no external funding.