3.1 Narrow-gap weld geometry characteristics
Figure 4 illustrates cross-sectional form of the weld at various welding currents and speeds in the externally shielded mode. Arc intensity and single-pass weld metal filler increased with the increase in welding current due to unitary regulation of CMT welding process. As shown in Fig. 4(a), weld height and penetration increased as welding current increased, nevertheless sidewall penetration did not considerably increase. At a welding current of 100A, the heat input and filler metal were insufficient and the sidewalls and toes solidified without sufficient melting. The sidewalls and weld toes were well fused at currents above 120A, and the shape was concave semicircular with no undercut. At welding speeds less than 4 mm/s, the weld shape changed from semi-elliptical to semi-circular, with improved fusion of the sidewalls and weld toes in Fig. 4(b). The reduction in welding speed significantly increased the heat input and weld metal filler, but had no effect on sidewall penetration. The weld was distributed asymmetrically and infused holes appear on one side when the welding speed exceeds 5 mm/s. In the narrow gap welding process, the metal in the fusion pool flowed rapidly from the front to the tail under the force of gravity. Although the arc could stabilize the melting pool at a large welding speed, the welding heat input was reduced and the melting side wall was weak.
The increase in welding heat input and weld metal filler as the welding current increased or the welding speed decreased facilitated the improvement of sidewall fusion in unified pulse CMT ultra-narrow gap welding [26]. Also, when the melting pool cooled faster or the filler metal was thicker, the influence of the welding arc on the bottom and side walls of the melting pool was lessened. Due to the difficulties of the liquid metal at the bottom of the molten pool spreading fully under the surface tension of the molten pool and the high solidification rate, unfused weld toes were likely to occur. More liquid metal gathered below the arc with more filler metal in a single pass weld, preventing the arc from heating directly to the bottom of the molten pool and, in particular, limiting the toe from obtaining sufficient energy to melt. Likewise, the welding arc would contain a significant amount of liquid metal that may be wrapped onto the side wall, and the side wall depth of fusion would be limited to prevent the arc from directly heating the side wall. To avoid unfused defects on the weld toe and achieve a suitable sidewall depth of fusion, increasing the heat input and employing weld filler were insufficient.
The swing arc was crucial for controlling the stability of the weld pool in narrow gap welding by spreading arc energy, promoting heat dissipation, and increasing sidewall melting [27]. Consequently, the arc swing (frequency and amplitude) was used to control welding heat input and the dynamic behavior of the weld pool in order to optimize the welding formation. Figure 5(a) displays cross-sectional forming shape of the weld at various swing amplitudes. When the swing amplitude was 0 mm, the interface between the weld passes was well fused, but the arc being farther from the side wall caused a substantial lack of fusion on one side wall and a reduced arc energy acting on the sidewall. The arc was closer to the sidewall as the swing amplitude increased more heat was transmitted from the arc to the sidewall, and the depth of fusion of the sidewall increased. The asymmetric weld was pushed to one side of the groove with a 6 mm swing, while the opposite side was badly formed and fused.
Figure 5(b) presents cross-sectional shape of a weld at various swing frequencies. As swing frequency increased, the weld metal was deviated to one side of the groove, while the other side appeared to have unfused sidewalls. The sidewalls were well fused with small swing frequency, and the weld toes were prone to slight poor fusion. There was no significant difference in the depth of fusion and weld height of the sidewall at various swing frequencies. In addition, when the swing frequency was greater than 2Hz, the weld presented asymmetric shape and the groove side was poorly fused.
Figure 6 shows a weld cross-section in the built-in shielding mode. Due to the narrow groove width of narrow gap welding, the small diameter of the built-in shielding gas tube used resulted in poor blowing stability. The air was easily included in the pool in fusion with the shielding gas, resulting in porosity. When the welding speed was 3 mm/s and 4 mm/s, the weld developed large pores, as shown in Fig. 6(a). When the speed was 6 mm/s, no significant porosity was found. Figure 6(b) shows cross-section morphology of weld formation in various swivel amplitudes. The weld shaping was similar to that of external protection, and the sidewall depth of fusion was larger than that of external shield. Significant porosity appeared in the weld with swing amplitude of 4 mm. The narrow gap welding process window narrows and the welding stability suffered when the built-in shielding approach was applied. Despite the fact that the weld was prone to porosity, the sidewalls were properly fused and the sidewall depth of fusion was greater than that of the external shielding approach. The use of built-in shielding helped to keep the arc stiffness and the molten pool flowing.
The weld forming parameters were determined as the weld height (Hw) and the melting depth of the side walls (Hs) as a function of the weld geometry to assist the investigation of narrow gap weld shaping characteristics. Figure 7 shows the weld seam dimensions for various welding locations and swing parameters. As indicated in Fig. 7, increasing welding current and decreasing welding speed increased sidewall depth of fusion, but swing width and frequency had little effect. The heat acting on the side wall increased as welding speed and current increased, resulting in an increase in side wall depth of fusion. The melt depth in the sidewall was 0.9 mm at a swing width of 4 mm. The melt depth in the sidewall was 0.96 mm at a 1Hz swing frequency. It was not advantageous to increase the sidewall melt depth with lower or greater swing widths, whereas swing frequency had no effect. When the arc was close to the side wall, the arc wall climbing phenomenon caused higher heat loss from the arc and a decrease in the side wall depth of melt. The heat conduction from the molten pool to the sidewall became difficult as the swing frequency increased, making it challenging to melt the sidewall.
The melting sidewall depth is determined by the arc direct heating of the sidewall and heat conduction in the melt pool, whereas the swing angle, sidewall residence time, and their interactions have greater impact on the melting sidewall heat. As the swing angle and sidewall dwell time increased, the continuous heating duration of the arc to the sidewall increased as did the amount of arc heat supplied to the sidewall, both of which helped in improving the sidewall melted depth. Because of the tightly groove gap, the tilt angle of welding torch was limited, and the swing angle process window was small. Figure 8 illustrates the morphology of weld shaping with various swing angles and sidewall dwell periods. As seen in Fig. 8, the weld morphology changed from a U-shape to a V-shape as the swing angle increased. Despite an insignificant increase in sidewall fusion depth, an unusable hole emerged in the weld toes. The increased effect of the arc on the sidewall and the weakened thermal effect on the two were caused by the large swing angle. With the increase of swing frequency, the filler metal and heat input increase, resulting in better spreading of the molten pool, and the sidewall depth of fusion increased. The increase in weld thickness tended to lead to poor fusion near the weld toes as shown in Fig. 8(b).
According to the characterization of narrow gap weld formation in different process parameters, the cross-sectional morphology of the weld was divided into symmetric and asymmetric welds, and the weld defects were classified into lack of sidewall fusion (Ls), lack of weld toe fusion (Lt) and porosity defects. The arc pattern, melt pool flow, and droplet transition all had a significant impact on defect generation and weld morphology. Figure 9 exhibits a schematic diagram of weld shape and defect generation under various welding procedures. As shown in Fig. 9(a), when the welding current was low, the welding arc energy and the amount of filler metal were decreased. Due to the rapid cooling rate, it is difficult to make the molten metal flow to the two side walls and melt them by arc heating alone. As a result, it was impossible to keep the side walls from melting when the heat input was modest. The cooling rate of the molten pool was slower under high heat input, and the molten metal could flow more easily to both sidewalls. As seen in Fig. 9 (b), heat and mass transfer between the arc and the melt resulted in good wall fusing. When the arc swing is large, the arc is attracted by the sidewalls and deflected to one side of the groove sidewalls, while the other side of the groove sidewalls is not sufficiently heated, as shown in Fig. 9(c). The direction of plasma flowing force and electromagnetic force on the droplet were diverted when the arc was deflected to the sidewall side, and the current line pointed from the droplet to the sidewall. The molten metal accumulating on one side of the slope caused the weld asymmetry on both sides of the slope. A sloping weld was eventually developed, resulting in lack of fusion on the sidewall side. Thus, when the welding heat input and swing were appropriate, the arc burnt persistently and the groove bottom and sidewalls were properly heated and fused in synchronization.
Figure 10 presents a schematic illustration of the creation process of unfused defects in the weld toes under high heat input. As shown in Fig. 10 (a), the high temperature residence time increases at high heat input and the effect of the arc on the sidewall increases. More arc and pool heat was delivered to the sidewall and weld toe location when filling the initial weld, and the sidewall was well fused. Moreover, after cooling, the weld toe molten metal poured into the weld pool by gravity, forming a noticeable nibbling defect (weld toe recessed into the sidewall), as shown in Fig. 10(b). Due to the enormous amount of filler metal used in the second weld, the arc heat could not be transferred to the toes, making it impossible for the molten metal to fill the toe and nibble the edge under the action of surface tension, resulting in an obvious infused weld toes.
Figure 11 shows cross-sectional shape morphology of narrow gap weld with different wire feed speed. As demonstrated in Fig. 11, increasing the amount of weld metal filler (wire feed speed) in a single pass resulted in a significant increase in weld height. The number of welding passes necessary at wire feed speed 140A was 9, but at wire feed speed 200A, only 4 passes are required to fill the groove gap. Besides, increasing the amount of filler metal for a single pass did not significantly increase the sidewall fusion depth, but rather caused it to fluctuate, especially when the amount of filler metal was too large, the single pass weld produced a significant undercut, resulting in an obvious poor fusion, as shown in Fig. 10(c).
3.2 Microstructure
Austenitic stainless steel welds can be classified into four types based on their cooling process solidification and solid-state phase transition mode: full austenite (A), austenite-ferrite (AF), ferrite-austenite (FA), and full ferrite (F). Austenite and ferrite morphology varies with solidification mode, with ferrite morphologies including eutectic, cytosolic, skeletal, lath, spherical, and worm-like. Figure 12 shows macroscopic and microscopic metallography of narrow gap welds. Figure 12(a) illustrates the observation position of micro-morphology in the welded joints. The metallographic examination of the lower part of the weld is depicted in Fig. 12(b) to (e). As illustrated in Fig. 12(b), the cross section of low heat input welded joint was divided into five regions: base metal (BM), heat affected zone (HAZ), fusion zone (FZ), weld (side and middle). The macroscopic morphology of the weld revealed apparent layering characteristics, which were primarily produced by changes in the microstructural morphology caused by the distinct temperature gradients. The microstructures of the base metal and HAZ did not differ appreciably, and both were composed of austenite and some amount of striated δ-ferrite. The non-convective mixing zone (NCMZ) near the fusion line is about 60µm wide, the ferrite is spherical and lath-like, and the weld grain shows epitaxial growth characteristics. The weld displayed predominantly typical columnar crystallites with a width of 5 µm, as shown in Fig. 12(d), generated by quick cooling of the bottom part of the weld. The fusion interface was visible in the middle of the weld, and grain development was parallel to the fusion line, but the morphology was substantially different, as seen in Fig. 21(e). The lower weld showed coarse columnar dendrite with spherical ferrite, whereas the upper weld presented single austenite with short rod grain shape and no δ-ferrite.
Figure 12 (f)-(i) shows macroscopic and microscopic morphology in the middle of weld. As demonstrated in Fig. 12(f), the grains in the HAZ grew and the fusion line was not evident. Figure 12(g) illustrates the microstructure near the fusing line. The width of NCMZ was approximately 30 µm, and the ferrite shape was predominantly spherical and wormlike. The weld was mostly made up of columnar dendrites with visible secondary dendrites, the ferrite morphology was skeletal, and the solidification transformation mode was AF. The microstructure of the center of the weld was composed of coarse dendritic crystallites, but no primary dendrites were seen. Figure 12 (j)-(m) shows microstructure of the upper part of the weld. As shown in Fig. 12(k), the fusion line and NCMZ were not evident and more ferrite was retained. Microstructure of weld was dominated by coarse dendrites which were interconnected and showed a reticulated structure, consisting of initial ferrite dendrites and austenite between the dendrites, and the direction of grain boundaries and grain growth was not obvious. Microstructures in the middle of the two welds had small differences, both being austenite and irregularly banded ferrite. The grain growth in the middle of weld was not significant, and there was a tendency to transform into an equiaxial crystal, as shown in Fig. 12 (m).
The relationship between austenitic stainless steel weld crystallization morphology and temperature gradient is depicted in Fig. 13 (a). The temperature gradient G in the liquid phase at the solid/liquid interface during weld pool solidification and the growth rate V at the solidification front determine the morphology and size of the grains as well as the internal substructure [28]. As illustrated in Fig. 13, the lower the G/V, the easier it is to construct an equiaxial crystal structure, and conversely for a columnar crystal structure. The temperature gradient and growth rate inside the weld pool are not the same at different locations. Usually the temperature gradient at the bottom of the melt pool is large, the growth rate is small, easy to form columnar crystals. The temperature gradient at the top of the molten pool is small, the growth rate is large, easy to form equiaxial crystals.
Narrow gap welding using a single layer-by-layer deposition, often the last weld surface will be remelted (remelted zone), so the top of the weld equiaxial crystals usually do not exist, and the bottom of the molten pool of the columnar crystals through the epitaxial growth of upward by way of the extension of the way. The area below the fusion line between the two layers is affected by the heat of the next pass (weld heat-affected zone), and the microstructure shows a clear gradient. As can be seen in Figs. 12 and 13(b), there are significant differences in grain morphology and size within each weld, and the phenomenon of "layering" occurs internally. The remelted zone is directly fused with the metal of the next weld, and the cooling rate is the fastest due to heat transfer, so it consists of fine columnar crystals (dashed ellipse), and the central zone consists of dendrites and cytosolic crystals (dashed circle).The growth direction of the grains in each region of each weld is basically the same, mainly along the longitudinal direction, which is due to the narrow gap welding process due to the axial direction of the arc is along the longitudinal direction, so the longitudinal direction is the direction of the temperature gradient is the largest, but also the most conducive to the growth of the grains in the direction. In addition, the narrow gap welding molten pool has an extremely fast cooling rate during solidification, which limits the formation and growth of secondary dendrites. Therefore, the solid/liquid interface of 316L stainless steel molten pool during narrow gap welding usually adopts the cellular growth method to advance forward.
The macro and micro morphology of narrow gap welds under high heat input is shown in Fig. 14. Despite the fact that the filler metal of the single-pass weld was greater under high heat input, the molten pool had a long residence time at high temperature, the temperature gradient was small, and the single-pass weld showed no noticeable delamination occurrence. Figure 14(b)-(e) indicate the macroscopic and microscopic morphology of the lower part of the weld. As shown in Fig. 14(b)(c), the fusion lines and grain boundaries were not visible, the weld exhibited associative crystallization properties, and the δ-ferrite was worm-like. As indicated in Fig. 14(d), the weld microstructure was primarily columnar grain, the remaining high temperature ferrite was less, and the width of the columnar crystals was much more than the low heat input. The middle weld channel had a lamellar dendritic crystal morphology, with the lamella extending down the thickness direction of the base metal and being more neatly ordered, with a greater dendritic crystal spacing, whereas the upper weld channel had smaller-sized rod austenite grains.
Microstructure of the center of the weld is shown in Fig. 14 (f)-(i). The partially melted zone had a width of about 50µm, as illustrated in Fig. 14(f). The weld microstructure exhibited coarse dendritic crystals but no sign of a dendrite. The crystallization temperature interval for austenitic stainless steel was rather small, and it crystallized to the middle of the weld in the form of dendritic growth. The equiaxial crystal was the primary representation of the middle of the weld. Figure 14 (j)-(m) shows the microstructure of the upper part of the welded joint. The weld was mostly made up of coarse dendritic crystals with clear direction, while the cellular crystals were barely visible. The dendrites in the center of the weld were exactly aligned along the base metal thickness direction. In comparison to the low heat input, the number of lamella dendrites reduced, lamella width increased, and dendrite spacing increased without crossing.
As shown in Figs. 14(j) and 15(a), significant longitudinal cracks were found in the upper part of the welded joints, mainly in the central region of the weld, and the cracks extended and expanded along the subgranular boundaries of the dendrites in different orientations. The center of the weld presents solidified subgrain boundaries that display small-angle grain boundaries in crystallography, as seen in Fig. 14(j). During solidification, the grains of weld metal crystallized along the preferred crystallization direction or easy growth direction, resulting in noticeable variances in solidification crystallization orientation in the middle of the weld. Crystallization cracks can be easily formed by the significant area of solidification sub-grain boundaries for elemental segregation. Figure 15(b) shows EDS surface scan of the cracked surface. Although there was no noticeable segregation of Fe, Ni, or Cr elements on the broken surface, the segregation of Si elements was visible. The low melting point eutectic elements of Si were polarized at the interface of grains with different orientations in the center of weld, forming a low melting point eutectic liquid film during solidification and causing along-crystal cracking under the influence of shrinkage tensile stress.
Grain boundaries (SGB) or sub-grain boundaries (SSGB) intersect the grain interface of austenitic stainless steel weld. The SGB is formed during weld solidification by grains generated by competitive growth along the edge of the molten pool, and the grain boundaries always have a large angular orientation deviation [29]. The majority of the SSGB structure was an interface made up of clusters of honeycomb crystals, dendritic crystals, or isometric crystals in the center of the weld. It should be noted that SGB was difficult to identify from the metallographic appearance of Figs. 13 and 14 alone, and it was also challenging to examine the grain morphology in the HAZ of the welded joints. Thus, EBSD analysis was performed to further explore the grain growth characteristics, and results are given in Fig. 16. HAZ has fine grains with a randomly oriented color dispersion, with an average grain size of roughly 50µm. At low heat input, the weld along the fusion line exhibited coarse columnar dendrites, with fine dendrites and a few equiaxed crystals forming in the center of the weld. The stress distribution of welded joints was similar to KAM, according to Ramazani A., 2014, and the stress distribution of welded joints was examined by the kernel average misorientation (KAM). Figure 16(b) and (e) displayed the KAM results for the weld, HAZ, and fusion zone. While the stress at the weld edge was higher than that in the middle of the weld, the HAZ stress was higher than that in the fusion zone. Between the HAZ grain boundaries and the columnar dendritic grain boundaries, the weld generated reticular stress distribution and linear stress distribution, respectively. The weaving of the weld was more apparent than the fusion zone and the HAZ, as seen by the polar plot (Pf) in Fig. 16(c)(f).
The grains grew more preferentially along the direction perpendicular to the melt pool boundary, or 100>, which is the direction in which face-centered cubic crystals of austenitic stainless steels tend to grow [30]. This was due to the narrow gap welding lateral for the maximum temperature gradient and heat transfer. The grain size distributions of the HAZ, fusion zone, and weld seam were illustrated in Fig. 16(h)(g) based on the EBSD results. Because the heat affected zone had small grains and the fusion zone had large grains, the dispersion of grain size near the fusion zone increased with increasing grain size. More than 60% of the grains had a diameter of less than 100 µm and were mostly found in the heat-affected zone. The weld grain size was substantially greater than in the HAZ and fusion zone, with around 80% of the grains being 100 µm − 550 µm.
Figure 17 shows the crystallization of an narrow gap weld under various heat inputs. At low heat input, the weld near to the fusion line was columnar crystal with more ferrite. The weld microstructure was dominated by narrow columnar dendrites and undeveloped secondary dendrites. Small rod-like crystals and columnar dendrites formed along the thickness direction in the center of the weld. Under high heat input, the weld was dominated by columnar crystals and dendritic crystals with apparent coarsening characteristics, and secondary dendrites were well formed, and the ferrite had a network or skeleton shape. The melt pool boundary temperature gradient (G) decreased as welding heat input increased, as did the G/R and G-R values. Increased supercooling of components during solidification and crystallization of the molten pool aided in the formation of equiaxed grains and the coarsening of the weld crystallization scale. Nonetheless, pulsed CMT narrow gap welding was a quick heating and cooling operation, and the temperature differential inside the molten pool was quite significant, inhibiting equiaxial crystal nucleation and development. As a result, there was no evident equiaxial crystal shape in the CMT narrow gap weld, and the middle of the weld primarily displayed columnar dendritic crystals.
In addition, the chemical inhomogeneity of 316L stainless steel welds is mainly characterised by micro-segregation and regional segregation. Microscopic segregation occurred mostly at the weld columnar dendritic grain border, while regional segregation happened primarily in the middle of the weld where dendritic clusters converged at the solidification sub-grain boundary. The appearance of bigger solidification subgranular barriers in the middle of the weld was primarily responsible for the formation of longitudinal solidification cracks in the weld under high heat input.
3.3 The mechanical properties of welded joints
Figure 18 shows microhardness distribution of welded joints under various heat inputs. Microhardness of the welded joint was higher than that of the base metal at low heat input, as shown in Fig. 18(a), with a difference of around 60 HV between the highest and minimum values, and microhardness at the bottom of the weld was higher than that at the top. The heat-affected zone and weld had higher microhardness than the base metal due to minimal grain coarsening and low ferrite content in these two locations. Microhardness values in the weld metal varied significantly due to changes in the form and composition of austenite grains and ferrite. The grain size had a large influence on microhardness, according to Hall Petch's formula [31], since grain boundaries are effective potential barriers for dislocation motion. The microhardness of the high-heat input weld was slightly lower than that of the low-heat input weld, owing to coarser columnar grains and more ferrite production, and the distribution pattern of the weld, heat-affected zone, and base metal was similar to that of the low-heat input, with the weld hardness value remaining higher than the base metal. The microhardness distribution indicated that the mechanical qualities of welded connections were satisfactory.
Experimental research was done on the tensile characteristics of welded connections at various locations and under various heat inputs. The stress-strain curves and tensile fracture locations of narrow gap welded joints were depicted in Fig. 19. As shown in Fig. 19, the order of the tensile strength was low heat input > high heat input, with average tensile strengths of 430 MPa, and 420 MPa, respectively. The elongation was also not significantly different. Defect-free tensile specimens with low heat input were all fractured in the base material away from the weld (middle position), and there was obvious necking at the fracture position, obtaining a welded joint with excellent static mechanical properties. Additionally, when unfused defects were present in the welded joint, the specimens tended to fracture at the defects. For instance, the high heat input specimen-3 tensile stress-strain curve (dashed line) broke close to the fusion line. The tension fracture morphology of welded joints is shown in Fig. 20. The low-heat input tensile specimen's fracture surface was composed of tiny equiaxial dimples, as shown in Fig. 20(a)-(c), and the fracture displayed typical ductile features. As seen in Fig. 20 (c), there was a significant variation in the size of dimples on the fracture surface, which was explained by the existence of coarse cellular crystals and elemental segregation inside the dendritic grains. Figure 20(d)(e) shows the fracture surface morphology of tensile specimens of welded joints with unfused defects at high heat input. The dark area was the unfused position with a flat fracture surface, while the other areas were dominated by dimples. Additionally, due to the coarse grain size of high heat input weld and the presence of δ−ferrite at the grain boundaries, the fracture morphology exhibited large-size dimples and tear ribs with quasi-cleavage features.