Laser milling (LM) is an emergent process for micro-fabrication, consisting of material removal by a laser beam through layer-by-layer ablation.
LM shows several advantages compared to traditional manufacturing processes, such as the ability to work hard to machine materials such as ceramics, graphite, and cemented carbides, the totally absence of tool wear, surface finish, aspect ratio, dimensional accuracy, and minimum feature size [1
Although several works can be found in literature on the effect of laser process parameters on quality of laser ablated parts, few of them focus on the analysis of the shape and dimensions of parts fabricated by laser milling.
Teixidor et al. [5
] studied the effects of laser milling process parameters on the geometrical and surface quality of micro-channels fabricated on AISI H13 steel. They showed that the laser milling process displays geometrical defects when grooves or micro-features are manufactured. These defects exhibit the complexity of the laser milling process and clarify the need to find methods and models to determine the best conditions and predict results to improve the productivity and quality of the process.
Darwish et al. [6
], investigated the fabrication of Inconel 718 micro-channels under dry and wet conditions, focusing on the analysis of the machined geometry in terms of width, depth, and wall shape angle. Geometries were measured with a scanning electron microscope (SEM) after cutting and polishing of samples.
Chen et al. [7
] analyzed the influence of several micro channel structure laser carving features (channel aperture, channel lines, and channel distance) on the efficiency and FF of polycrystalline silicon solar cells.
Karazi et al. [8
] developed models for the prediction of the width and depth dimensions of CO2
laser-formed micro-channels in glass. The width and depth dimensions of the micro-channels for each experiment were measured at three different locations along the produced channel. The measurement system used was an in-house built laser profilometer that had a 1.95 mm resolution in the x and y directions and a 0.5 mm resolution in the z-direction.
Schille et al. [9
] studied performance of laser micro processing of metals in terms of ablation depth, wall-angle, and surface roughness. They used a confocal point sensor CF 4 and a tactile roughness device (DEKTAK 3030) for measurement of parts.
Bulushev et al. [10
] measure micro-channels by a profilometer based on a confocal chromatic sensor and by a confocal microscope with higher lateral resolution, in order to measure the depth of micro channels using the first sensor and roughness and edge characteristics.
Rysava & Bruschi [11
] compared micro-milled channels on an Electron Beam Melting (EBM) and Direct Metal Laser Sintering (DMLS) work pieces. The micro-machinability is evaluated in terms of burr formation, surface integrity (surface topography and surface defects), tool damage, and microstructure alterations. Scanning Electron and Confocal Microscopes are the measuring instruments employed. Generally, the election measuring instrument for 3D characterization is the Confocal Microscope.
On the other hand, the use of optical and confocal microscopes is limited when high slope walls should be measured. Conoscopic holography (CH), a non-contact interferometric technique used for surface digitization, presents several advantages over other optical techniques such as laser triangulation [12
]. Among others, the ability for the reconstruction of high-sloped surfaces stands out, as does its lower dependence on surface optical properties [13
In the literature, one of the most widely investigated materials is Ti6Al4V, an alpha-beta titanium alloy with high corrosion resistance, applied where low density and corrosion resistance are critical factors, such as in the aerospace industry and in biomechanical applications. Recently, the interaction between lasers and this alloy has been investigated, with particular attention to superficial treatments such as texturing.
For example in [14
], a Yb-fiber laser is deployed to generate a quasi-uniform distribution of laser-induced periodic surface structures on titanium alloy substrates for adhesive bonding. In [15
], Ti6Al4V micro-dimple surfaces fabricated by a masked laser surface texturing technique within water were subjected to soft and hard contact laser shock peening. In [16
], the texturing was investigated of Ti6Al4V, by applying a pulsed Nd:YAG laser, aimed at surface characterization by microscopy.
Although many works can be found in the literature on analyses of laser surface treatments of Ti6Al4V titanium alloy, it seems there is not a systematic study showing the effects of main process laser parameters on shape geometry and roughness of parts fabricated by laser ablation. The aim of this paper was to investigate the shape geometry and roughness of Ti6Al4V features fabricated by laser milling using a nanosecond Nd:YAG laser source. The impact of the laser processing parameters on machining outcomes was studied in order to determine the optimal processing conditions, i.e., reducing geometrical defects and improving surface quality. In particular, the influence of average power, frequency and scan speed were investigated. Moreover, the laser machined 3D shape Ti6Al4V parts were measured and analyzed using CH technology. The Optimet Conoscan 4000 was employed with a spatial resolution of 12 μm × 12 μm and of 2.5 μm on the z axis.
In order to evaluate the influence of the considered process parameters on the error index calculated for single considered geometrical entities and on roughness, an Analysis of Variance (ANOVA) was performed [18
]. The evaluated response results were ER L_Top, ER L_Bottom, (ER%)α, Depth and Ram.
The General Linear Model was used to perform ANOVA. Factors for this model are discrete variables; therefore, ANOVA examines whether the variance of the factor is zero. The p-value determines whether the effect for that term is significant. If the effect of a discrete factor is significant, then the variance of the factor is not zero.
The probability value α represents the factor coefficient; the smaller the value, the more significantly it represents the factor. In the present analysis, a threshold value of 0.05 was chosen. If the p-value is greater than the chosen level, the null hypothesis is accepted and the coefficient is considered not to be significant. Analysis of variance results are summarized in Table 2
. The symbol (S) refers to the influence of each part parameter on the measured geometrical entities; the symbol (NS) means that the considered process parameter does not affect the result.
(ER%)α, Depth, and Ram result are affected by the variation of all the considered process parameters, while (ER% L_Top) is influenced only by (Fp) and (ER% L_bottom) by both (v) and (Fp).
a–c show main effects plot for (ER%)α, Depth and Ram vs. the three input parameters Average Laser Power, Scanning Speed and Frequency. Some observations can be made:
The slope of the curves confirms results of ANOVA on significance of input process parameters on analyzed results;
The lowest value of (ER%)α can be obtained setting P = 20 W, v = 300 mm/min, Fp = 40 kHz.
The lowest value of Ram can be obtained for P = 10 W, v = 300 mm/min and Fp = 30 kHz;
A peak of Depth is present for P = 20 W, v = 300 mm/min, Fp = 40 kHz.
Moreover, it is possible to compare the relative magnitude of the factor effects, comparing the slopes of the lines on the plots. Plots for Ram show that by setting the Average Laser Power to within a range of 10–15 W, average roughness does not change significantly; the same consideration can be made for the variation of Frequency in the range 30–40 kHz.
Values of process parameters that minimize wall shape angle are the same that maximize Depth. This is due to the same definition of the wall shape angle, as assessed by Darwish et al. [6
]. The wall shape angle depends on the difference of the channel’s width at top and bottom, in direct proportion, and two times the depth of channel, in inverse proportion. So, it can be said that wall shape angle is mainly controlled by Depth.
After ANOVA, the response surface methodology (RSM) was used for modeling and analyzing the response of the single outputs with the considered factors. The contour plot methodology was used to visualize the response surface. Figure 9
and Figure 10
show contour plots of (ER%)α and Depth versus v and Fp
at constant Average Laser Power of respectively 10 W and 20 W. Since the purpose of this work is also to analyze the reduction of the error, and at the same time to get the best surface quality, laser power was set to these two levels (the minimum and the maximum of the considered range). Indeed, results of the previous analysis on main effects showed that the minimum values of Ram and (ER%)α are obtained respectively for P = 10 W and P = 20 W.
The contour Plots of (ER%)α for the two different P values confirm that values of (ER%)α lower than 25% can be obtained for P = 20 W, Fp between 10 kHz and 35 kHz and v equal or slightly higher than 300 mm/min.
Comparing these values to those of Ram (Figure 11
), it can be observed that for the same average laser power and scanning speed values, values of Ram lower than 2 μm can be found in a frequency range of 20–37 kHz.
Results for roughness can be related to the degree of overlap, as demonstrated in Figure 7
, in which can be detected a range of O% between 64% and 79% in which Ram shows values lower than 2 μm. However, Ram it is also affected by the optical energy delivered per unit area (fluence). Figure 12
b shows, the scatterplot of Ram vs. Fluence (J/cm2
) with panel variable scanning speed. At lower values of fluence (<20 J/cm2
), roughness is very low and it seems not to be much affected by energy; at higher values of fluence (>20 J/cm2
) roughness increases slowly for v = 300 mm/s e faster at v = 500 and 700 mm/s. Similar results were found by Campanelli et al. [4
] on laser ablation of 5754 aluminum alloy.
a,b confirm that maximum values of Depth are found per the minimum scanning speed v = 300 mm/s. This result is validated by Figure 12
b, which shows scatterplot of Depth (μm) vs. Fluence (J/cm2
) with panel variable scanning speed. For the highest scanning speed (v = 700 mm/s) depth seems to have only a slight variation with laser fluence. This is due to the lower interaction time between the laser beam and the machined surface [4
]. For lower values of speeds (300 and 400 mm/s), depth increases quickly for lower values, and slowly for higher values of fluence.
ER%α shows a slight reduction with the degree of Overlap, but there is large data dispersion; thus, the variation of ER%α cannot be attributed entirely to O%, but is also due to lower machining depths, as also assessed by Darwish et al. [6
In conclusion, the analyzed results show that by setting appropriate process parameter values, it is possible to reduce the error on vertical walls, and at the same time, increase the surface quality.