The experimental investigations were conducted using galvanized steel ASTM A635CS with galvanneal A40 coating. The sheet thickness was chosen to fit within the thickness range usually used for automotive bodies (0.8 - 3.6 mm). Three different thicknesses are taken: gauges 11, 14 and 20, respectively 3.132 mm, 1.994 mm and 1.006 mm. The chemical compositions of all sheets are shown in Table 1. Specimens were cut to 30 mm × 50 mm using hydraulic shear. The edge did not receive further treatment. The small differences in chemical composition are neglected.

The welding equipment was a six-axis-robotic system FANUC M-710ic equipped with a fiber Laser source IPG YLS-3000-ST2 and a laser head HIGHYAG BIMO. The Nd:YAG laser source emits light with a wavelength of 1070 nm and a maximum power of 3 kW. The power is transferred through an optical fiber, with a diameter of 200 µm, to the laser head, with a focal length of 300 mm. A collimator allows the user to adjust the beam diameter between 340 µm and 520 µm. Figure 1 shows the laser welding setup with the laser and the clamping

system. The air gap was created using a 0.1 mm thick steel strip positioned between the samples.

After welding, the welded assemblies were cut perpendicular to the weld surface to obtain the desired cross-sections. The cut specimens were then prepared for microscopic observation using standard procedures including grinding, polishing, and etching.

An optical microscope, Clemex MMT Type A, with a CMT program was used for the evaluation of the weld dimensions. For each experiment, three dimensions were measured (penetration, width, and lack); each measurement was taken 3 times and then averaged to obtain the data used in the analysis.

Vickers micro-hardness testing was conducted with the same equipment using a load of 500 g and a dwell time of 15 s. The base material measurements were taken far from the fusion zone. All indentations were adequately spaced so that the strain fields from adjacent indents would not interfere with each other.

Laser power, welding speed, sheet thickness, beam diameter, and gap are the parameters used in this study. The thickness and laser beam diameter are chosen to fit with the available ranges: 1 mm, 2 mm and 3 mm for the thickness and 340 µm to 520 µm for the diameter. The results of preliminary studies on the para- meters limitations and the results of others similar studies [13] , were used to determine upper and lower limits for the parameters.

For the gap, three relevant information emerged from the literature review. The gap has to be less than 10% of the workpiece thickness and less than the diameter of the laser beam. A gap ranging from 0.1 mm to 0.2 mm is generally used and gives good results. Considering these constraints, a gap of 0.1 mm is needed, leaving only the welding speed and laser power to be determined. These two parameters are known to be very influential. Using the results of the preliminary tests, the upper and lower limits are taken to respect the weld penetration constraint. This means that the lowest speed with the highest power can result in full weld penetration for a 3 mm thickness specimen, and the highest speed with lowest power can result in partial weld for a 1 mm thickness specimen. Three levels are chosen to assess the nonlinear components of the parameter effects: speed between 3 and 9 m/min, and power between 2 and 3 kW. The factors and levels chosen can be seen in Table 2.

These factors and levels are used to generate an experiment design that will yield the maximum data with the minimum experimentation efforts. The Taguchi method is used to achieve an efficient statistical analysis. An explanation of this method can be found in [14] . An orthogonal array L54 is used allowing the combination of factors with 2 or 3 levels. In this case one factor with 2 levels and four factors with 3 levels are needed. Welding tests are carried out in a random order to avoid any systematic error in the experiment. In addition, repeatability tests are done using the median values from the experiment design (2500 W power, 2 mm thickness, 6 m/min speed and 450 µm diameter). Since the positioning of the gap experiment requires different handling, 12 repeatability tests are done with a gap and 12 without gap, for a total of 24 supplementary tests.

During the repeatability tests, the depth of penetration, bead width, and underfill are precisely measured. In order to evaluate the total measurement error resulting from non-studied parameters such as workpiece positioning, parameter stability and change in environment during the experiments, average of quality attributes, standard deviations, and the relative errors are estimated.

The results in Table 3 indicated the good precision for the measurement of the penetration and the width, showing that with a normal distribution, 68% of observations will be within one standard deviation and 95% will be within two standard deviations. This means that the penetration and the width can be estimated with less than 5% error and the underfilling can be estimated with less than 15% error. The underfill dimensions are relatively small generating higher error percentages. The standard deviations are similar for all parameters. The

standard deviation is slightly higher for the gap test. That means there is lower repeatability in the gap positioning.

An underfill of 10% of the thickness is present in the case of experimentation without gap. This must be caused by the edge morphology of the sheared edge. Flat or concave welds are produced using laser cutting or precision die blanking. A small gap exists when two sheared specimens are connected, as only the 2 burnished zone are in contact, the rest of the edge are spaced. The underfilling becomes about 20% of the thickness with the gap experiment. Figure 2 shows the morphology of the edge of the specimen used for the experiment. The dimensions of the sheared edge have been measured and are relatively close for the 3 mm thickness. The rollover accounts for about 10% of the thickness, the burnished zone between 40% and 45%, and the fractured zone for 50%. The fracture angle ranges from 12˚ to 15˚.

Figure 3 shows the distribution of micro-hardness across the laser welding zone for the repeatability test. According to the hardness test, the hardness of the substrate (150 Hv) was much lower than that of the welded region (330 Hv), which agreed with the results found for low carbon steel [8] . The increase in hardness is caused by grain refinement, formation of hard phases and generation of residual stress in the weld region due to rapid solidification [11] . The microhardness is comparable in all tests, with and without gap.

To obtain the effects of each factor on the weld dimensions, ANOVA were carried out based on the measured results as illustrated in Table 4 and Table 5. The p value is used to determine if the factor has an impact on a given parameter, and the percentage represents the risk of concluding that an influence exists when there is no actual influence. The percentage contribution is the variation caused by an individual factor over the total variation.

Any factor whose p-value is under 10% will be considered meaningful. According to the results for the BDW, every factor is found to be meaningful. The gap is the least meaningful factor, with a p-value of 7%. For the DOP, the range of variation in diameter is found to not be an impacting factor, while all the

other factors are meaningful. The percentage contribution for the BDW attributes 18% of variation to non-studied factors, with the two main factors being speed (63%) and diameter (13%). The major factors for the DOP are speed (24%) and thickness (39%), while non-studied factors represent 24% of variation.

Figure 4 shows the effects of each of the 5 factors. The direction and steepness of the slope are used to determine the impact. We can see that the use of a gap reduces the width and increases the penetration. The diameter, contrary to the gap, increases the width and reduces the penetration. Thickness and power have a positive impact on the width and the penetration, while speed has the opposite impact, reducing the penetration and the width. Every line is roughly straight, so the effect of the parameter does not change as it increases. The speed seems to have a stronger effect because it has a wide range of variation. In conclusion, three factors have a similar impact on penetration and width, while two factors have the opposite. The results for power, diameter and speed fit with [15] .

To visualize whether the effect of one factor depends on the level of another factor, we used an interaction plot. Parallel lines in an interaction plot indicate no interaction; otherwise, the greater the difference in slope between the lines, the higher the degree of interaction. Only one significant difference in slope is found for the BDW. As can be seen in Figure 5, the thickness tends to increase the width without gap, but when using gap, thickness variation has insignificant impact on the bead width. This means that the weld done using a gap would stay thinner for thicknesses greater than 1 mm.

Interaction seems to be found between the thickness and the other factors for the DOP. Figure 6 shows the graph for speed and thickness. The interaction found is that using a greater thickness, with the values of the other factors giving the best penetration, allows a deeper weld, which is a logical interaction. For example, welding a 1 mm thickness sheet at low speed gives a full penetration weld that is about 1 mm, but using a 3 mm sheet gives a 3 mm weld, and when welding at a high speed both thicknesses have the same low penetration. This explains the change in the slope steepness, and is the same for all of the factors.

The shape of the weld is mainly determined by the welding mode conduction or keyhole. As we can see in Assuncao study [16] [17] , there is not a sharp transition between conduction mode and keyhole mode. The mode is dependent on the interaction time and power density. In our experiments the interaction time varies between 2 ms and 10 ms, and the power density between 0.95 MW/cm² and 3.3 MW/cm². Given those values, the experiment’s range must be in the transition mode and the keyhole mode. The results give an aspect ratio between 0.5 and 2.2, which fits with these modes.

The weld shape is supposed to be from small and round to thin and deep, which is what we observe. Four weld shapes are considered: 1) round welds are

the welds almost in conduction mode (U weld), 2) wide V welds are the welds in transition mode (HV weld), 3) sharp V welds are those between transition mode and keyhole mode 4) weld), and 5) parallel edge welds are those with full penetration done in keyhole mode (H weld). Figure 7 gives a picture for each of the four morphologies.

Figure 8 represents the weld shape according to the DOP, the gap and the thickness. We notice that the round and wide V welds are only found in experiments without gap. The shape depends on the penetration of the weld, with the lowest penetration weld being round shaped, then wide V shaped, then sharp V shaped and finally parallel edge with the no-gap experiments. This order can be found for all three thicknesses. The penetration is directly linked to power density and interaction time. The experiments with gap only produce sharp V shaped welds for partial welds, and mainly parallel edge welds for full penetration. However, sometimes there can be a V shaped full penetration weld, which does not appear without gap.

The use of a gap tends to make only sharp V shaped welds in the case of partial penetration welds. The assumption can be made that the gap allows an easier penetration for the weld because the melted metal can flow into the gap. The deepest penetrations achieved with gap are deeper than penetrations without gap because of this easier penetration. The higher penetrations achieved with gap are shallower than those without gap, because there is more underfilling caused by the gap. We can also remark that all sharp V welds without gap have more than 50% penetration, which is where the burnished zone and the fracture zone begin forming an unintended gap between the pieces.

The experimental results reveal that tests with gap have more underfill than those with no gap. Figure 9 presents the distribution of the underfill versus the DOP in % of specimens’ thickness. Excluding the alignment on the right, the figure reveals the presence of an underfill of about 5% for the experiments without gap. The underfill tends to grow slightly after 50% penetration. This is certainly due to the fracture zone. An underfill between 10% and 15% is present when using gap. When looking at the alignment on the right, a trend appears with depth of penetration plus underfill equal 100%. Furthermore, the points on that line were for full penetration welds. Only the underfill on the upper face has been measured. That is why when underfill exists on the bottom face, the sums are slightly less than 100%. This suggests the question why all the full penetration welds do not have the same underfill.

Figure 10 presents the BDW versus the underfill for full penetration weld. This figure shows that the width decreases when the underfill increases. This observation fits with the refillable gap formula: g = A*B*ΔT*w proposed in [18] , where A is constant, B is the coefficient of thermal dilatation, ΔT is the temperature differential, and w is the width. The measurements in the bottom right corner reveal the weld with the least underfill. The use of gap moves this point upward by about 10%, but maintains the same slope, and the underfill is substantially greater for the same width. This means that when using a gap the underfill is bigger for the same bead width. The largest number of full penetration welds are achieved using 1 mm sheet thickness explains the larger number point

in the graph. The underfill also increases slightly with the thickness because of the sheared edge that creates more gap with increased thickness. In order to achieve good weld on thicker plate, wider weld have to be done if there’s no additional edge preparation.