Generation mechanism of lack of fusion in X70 steel welded joint by fully automatic welding under steep slope conditions based on numerical simulation of flow field

X70 pipeline steel is butt welded by fully automatic welding at 25° slope, and the characteristic information of welding process is collected. A 3D transient heat and mass transfer model of weld pool is established. This model was verified by means of the shape of the weld pool and the two-dimensional shape of the weld, and then, the flow field of the fully automatic welding process under steep slope conditions was simulated to clarify the mechanism of the lack of fusion defect. The results show that the welding arc is deflected when it stays on both side walls of the groove. The slope makes the liquid metal tend to gather at the lower groove. When gravity plays a leading role in the flow of the molten pool, some liquid metal begins to flow out from the lower groove to the front end, resulting in an overflow phenomenon. The generation of overflow phenomenon and arc deflection cause the formation of a “void zone” at the structure mutation position in front of the groove. In the welding process, the liquid metal tends to flow to the side walls for heat transfer under various thermal effects, and forms a large range of eddy current at the side walls. This has a positive effect on sidewall melting. In addition, the effect of surface tension at the lower groove is more obvious than that at the upper groove, and the trend of upward flow along the side wall is stronger, so it is difficult for the high-temperature liquid metal in the arc center to flow to the “void area” located at the bottom of the side wall of the lower groove. Therefore, the bottom of the lower groove is more likely to cause lack of fusion defect.


Introductions
In recent years, pipeline construction in mountainous areas has increased year by year in order to avoid urban planning and infrastructure construction in China. Thus, under the influence of terrain, environment, and transportation, pipeline welding in mountainous areas has put forward higher requirements for welding technology [1,2]. The fully automatic pipeline welding realizes the consistent control of welding process and reduces the influence of human factors through preset the welding process parameters [3,4]. At the same time, this welding process has low heat input, high efficiency, and can effectively ensure the stability of welding quality [5,6]. Now, Chinese pipeline enterprises have introduced the intelligent construction concept of "smart pipe network." Traditional manual welding and semi-automatic welding technologies cannot meet the corresponding construction requirements. Therefore, the automatic welding technology will become the most suitable welding technology for pipeline construction in the future.
Fully automatic welding process has the advantages of narrow gap groove, less filler metal, small welding heat input, and excellent mechanical properties. However, the fusion effect of the side wall is poor, resulting in a lack of fusion defect at the side wall. In fact, lack of fusion is the most typical welding defect in the automatic welding process of pipelines in mountainous areas. This not only reduces the bearing area of the structure but also tends to produce stress concentration at the edge of the lack of fusion, thus promoting the generation of cracks and eventually leading to the cracking of the entire weld.
The causes of the lack of fusion mainly include the following aspects: (1) Groove size. The groove shape commonly used in the project is shown in Fig. 1. The groove dimensions that significantly affect the lack of fusion defect include the height P of the blunt edge, the dimension H of the intersection point, and the width K of the groove opening. For the U-V composite groove, the size accuracy of the variable intersection point size H is an important size that affects the lack of fusion defect. (2) Cleanliness of the groove surface. The existence of oil stains, rust, and other oxides on the groove surface will more easily hinder the direct thermal action of the arc and lead to the generation of lack of fusion defects. Therefore, in the actual pipeline welding engineering construction process, the cleaning of the groove and the welding layer is essential. (3) Pipe assembly quality. In the process of pipe welding, there are often problems such as too small butt gap and excessive misalignment, which will increase the probability of lack of fusion defects. (4) Type and flow rate of shielding gas. The type and flow rate of shielding gas have significant effects on the weld width and penetration depth [6]. At present, 100% CO 2 gas and mixed gases of CO 2 and Ar in different proportions, such as 80% CO 2 + 20% Ar, are commonly used in fully automatic pipeline welding. Zhao et al. [7] indicated that oxygen rich in the shielding gas increases Marangoni convection force and molten pool flow, and drives the liquid metal to flow from the edge of the molten pool to the bottom of the molten pool. Thus, large penetration depth is easy to obtain. Cho et al. [8] found that when the gas flow rate is increased, the arc heat flux (arc pressure, current density) is increasingly focused to the center, and accordingly forms the finger shape penetration. (5) Welding parameters. The influence of welding parameters on lack of fusion is the most obvious and difficult to control. The causes of lack of fusion defect mainly include low heat input, not melting of the previous weld pass or side wall, narrow swing width, and mismatch of swing width, swing speed, and welding speed [9,10]. (6) Welding equipment. When there is a deviation in the installation accuracy of the welded track, it is easy to cause lack of fusion defect on one side. The stability of welding equipment and other factors will lead to poor wire feeding and the straightness of the arc, leading to the lack of fusion defect.
With the growing maturity of finite element and computational fluid dynamics theory, the numerical analysis model of welding process is established to study the stress state, temperature distribution, and flow behavior of the molten pool in the welding process. This is helpful to analyze the influencing factors and formation mechanism of lack of fusion defects. Cho et al. [11] studied the transient flow, heat transfer behavior, and flow patterns of the molten pool when the arc moves in a straight line at different welding positions. He et al. [12] carried out a series of welding tests to study the formation mechanism of lack of fusion defect in 5083 aluminum alloy. It was found that the molten pool surface tension, which is greatly influenced by temperature and oxide, is a main reason. Increasing the heat input to sidewalls by arc-swing, improving the gas protection of molten pool against oxidization, or reducing the heat dispersion of molten pool by preheating are all effective measures for eliminating the lack of sidewall fusion defect. Meng et al. [13] confirmed by experiments that lack of fusion depended on the cross-sectional shape of previous pass, which was classified into four types, namely, concave, convex, middle convex, and inclined shapes; only the concave shape was effective in suppressing the lack of fusion. The electromagnetic force and Marangoni convection can be increased to accelerate the liquid metal downward, thus increasing the concavity of weld surface. In addition, the wetting angle between the molten pool and the side-wall can be reduced by reducing the solid-liquid surface tension, thereby increasing the concavity. Cho et al. [14] established a three-dimensional transient analysis model of V-shaped groove, monitored the arc shape with high-speed camera, and established an asymmetric heat source model. This model considers the asymmetric arc action due to the existence of groove. Yang [15] established a finite element model of X-shaped groove and trapezoidal groove to analyze the causes of lack of defects from the heat and mass transfer behavior of the molten pool. It is found that the main flow direction of liquid metal is from the center of the molten pool, where the arc is located, to the root, and back to the surface of the molten pool through the fusion line. When the gap between grooves is small, the liquid bridge formed when the arc stays on both sidewalls blocks the flow of liquid metal to the root, resulting in lack of fusion defect. In a word, the research on lack of fusion defect mainly focuses on the fluid characteristics of the molten pool, such as the stress state of the molten pool, and the transient flow and heat transfer behavior of the internal liquid metal.
However, there is no report on the formation mechanism of lack of fusion in fully automatic welding under steep slope conditions. It is a fact that the overhead welding position is easy to produce lack of fusion defect, and two typical cross-sections weld, namely, medium convex weld and deflection weld, generate in our experiments. Thus, in this paper, a 3D transient heat and mass transfer model of weld pool is established. The FLUENT software is used to simulate the flow field of next weld pass based on the two typical cross sections at the overhead welding position during the fully automatic butting welding X70 pipeline steel under steep slope conditions, and the mechanism of lack of fusion is proposed. The research results provide an important theoretical guidance for the establishment of automatic pipeline welding parameters in mountainous areas.

Experimental materials
The experimental material is X70 pipeline steel with a pipe diameter of 813 mm and a wall thickness of 17 mm. The welding groove form is shown in Fig. 2. The relationship curves of thermophysical parameters of X70 steel such as specific heat capacity, thermal conductivity, density, and viscosity with temperature are shown in Fig. 3. The thermophysical parameters that do not vary significantly with temperature were set as constants, and the specific values are shown in Table 1.

Basic assumptions and governing equations
The fully automatic butting welding of X70 steel under steep slope conditions involves complex physical phenomena, making it difficult to consider the effects of all factors arising in the numerical analysis process. Thus, based on the ability to accurately reflecting to the characteristics of the heat and mass transfer behavior of the liquid metal in the molten pool, in order to further improve the efficiency of the calculation, the following simplified assumptions are made for the numerical analysis model [16][17][18]: (1) The liquid metal in the molten pool is an incompressible Newtonian viscous fluid with laminar flow. (2) The specific heat capacity, thermal conductivity, and viscosity of X70 steel change with temperature, while the other thermophysical parameters are fixed values. (3) The VOF two-phase flow model are adopted as the numerical model, assuming the molten droplet and X70 steel as the same phase and using the same thermophysical parameters. (4) Only when calculating buoyancy, the density changes with temperature will be considered. (5) The energy and mass loss caused by evaporation and splashing of liquid metal in automatic welding process is ignored.
On the basis of the above assumptions, as a continuous medium, the heat transfer and flow behavior of liquid metal in molten pool under steep slope conditions can be described by the mass conservation equation, N-S equation, energy conservation equation, and VOF continuity equation.

Continuity equation
The liquid metal in molten pool, as a continuous, incompressible Newtonian fluid, follows the law of mass conservation and can be described by the continuity equation.
where V is the velocity vector, m s is the quality of the source term, and ρ is the density.

N-S equation
The flow process of the molten pool follows the law of momentum conservation, and its coupling relationship where ρ is the density, t is the time, and P is the pressure, F V is the momentum source, μ is the fluid viscosity, and K is the coefficient of velocity attenuation for the liquid metal in the paste region. The value of K is expressed in the Carman-Kozeny equation based on Darcy's theorem [19].
where C is a constant and σ is a very small positive number in order to avoid a denominator value of 0. F l is the volume fraction of the grid occupied by the liquid metal, which is related to the temperature expressed as where T s is the solidus temperature of the material and T l is the liquidus temperature of the material.

Energy conservation equation
The flow process of molten pool follows the law of energy conservation. The relationship between energy conservation equation and convection, heat conduction, and internal energy can be expressed as where q a and q d are the arc heat and molten drop heat, respectively; λ is the thermal conductivity of the material, h is the enthalpy of the material, C P is the specific heat capacity of the material, and L f is the latent heat of fusion of the base material.

Interface tracking VOF equation
The volume fraction of liquid metal as a function of time can be expressed as where ϕ is the volume fraction of fluid metal; when ϕ = 1, the grid is all liquid metal; when ϕ = 0, the grid is all gas phase; when the calculation result is 0 ≤ ϕ ≤ 1, there are both gas and liquid metals; that is, there is an interface.

Geometric models
Considering the large size of X70 pipe, it is difficult to realize the numerical calculation if the calculation domain is constructed according to the equal scale. Therefore, only the special welding location closer to the molten pool is selected as the calculation domain. As shown in 4, X is the welding direction, Y is the swing direction during welding, and the Z is the direction of the gravity component. It should be noted that the high side of the groove is defined as upper groove, and the other side is defined as lower groove. The computational domain size is 50 mm × 16 mm × 11 mm, using a uniform hexahedral mesh 7 size of 0.25 mm. An air layer is set on the inside of the groove and the bottom of the model, and the workpiece to be welded is placed in the middle. Then, the actual heat dissipation conditions are applied to this model, the droplet transfer process can be realized and the interface of the molten pool can be tracked in real-time.

Model of arc heat distribution
The molten pool is subjected to complex thermal and force interactions during fully automatic welding under steep slope conditions, as shown in Fig. 5. The gravity is applied using the module of the FLUENT software. But the other thermal forces are applied through UDF secondary program development.

Motion trajectory of arc center
The motion of the arc center is a periodic process, as shown in Fig. 6. The arc swing cycle (marked as T) consists of five stages, namely, stage (1): the arc firstly swings towards the lower groove; stage (2): the arc stays at the lower groove); stage (3): the arc swings towards the upper groove; stage (4): the arc stays at the upper groove; and stage (5): the arc secondly swings towards the lower groove. X and Y represent welding direction and swing direction, respectively. Meanwhile, Vx and Vy represent welding speed and swing speed, respectively. L is the half of swing width, T is the arc swing cycle, T1 is the dwell time of the arc at both lower and upper grooves, and T2 is the half time of swing width. In addition, x ′ represents the coordinate change in the welding The relationship between the swing parameters can be described as With the help of the remainder operation relationship between the time step function CURRENT_TIME and the swing cycle T in the FLUENT software, the different motion stages of the arc center during the welding process are described. The remainder operation is expressed by the fomd function in C language, and the calculation relationship is Remainder = fmod (TIME, T). The specific arc center motion relationship is as follows.
When remainder ≤ T1, the arc is in stage (1). At this time, the coordinate change in Y direction can be described as When T1 < remainder ≤ T1 + T2, the arc is in stage (2). At this time, the coordinate change in Y direction can be described as When T1 + T2 < remainder ≤ 3•T1 + T2, the arc is in stage (3). At this time, the coordinate change in Y direction can be described as When 3•T1 + T2 < remainder ≤ 3•T1 + 2•T2, the arc in stage (4). At this time, the coordinate change in Y direction can be described as When 3•T1 + 2•T2 < remainder ≤ T, the arc is in stage (5). At this time, the coordinate change in Y direction can be described as The coordinate change in the X-direction for each stage is The above arithmetic relationship can not only describe the motion state of the arc center at different times, but also parameterize the specific motion trajectory of the arc center. In the automatic welding process, the thermal and mechanical effects of the arc on the molten pool are added on the basis of the above motion coordinates. Figure 7 shows the welding arc shapes of automatic welding under steep slope conditions. It is found that when the arc is in the center of the weld pass (Fig. 7a), the arc is elliptical, and the shapes of both sides of the arc center are basically (13)

Distribution model of electric arc heat source distribution model
symmetrical. When the arc moves from the center of the weld pass to the left side of the weld pass ( Fig. 7b corresponds to stage (1) in Fig. 6), the arc tends to deflect to the left side wall, but the original characteristics are still retained overall. When the arc stays on the left side of the weld pass ( Fig. 7c corresponds to stage (2) in Fig. 6), the arc will deflect due to the influence of the surrounding magnetic field. This makes the one sidewall susceptible to more heat input than the other side [20,21]. After staying on the left side of the weld pass, the arc begins to move to the right (Fig. 7 d, e, and f correspond to stage (3) in Fig. 6). When the arc has just left the left side of the weld pass and is about to reach the right side of the weld pass, because the arc center is close to the side wall, the arc will still be slightly deflected by the side wall. When the arc center moves to the center of the weld pass, the arc returns to elliptical shape. When the arc stays on the right side of the weld pass ( Fig. 7g corresponds to stage (4) in Fig. 6), the arc shape is similar to that of stage (2). The arc burns between the welding wire and the side wall of the groove, resulting in a deflection to the right. When the arc moves to the center of the weld pass from the right stop position ( Fig. 7 h and i correspond to stage (5) in Fig. 6), the arc changes from deflection form to symmetrical elliptic form again. It can be seen from the above conclusions that the arc shape changes regularly with the swing of the arc center.
To realize the impact of different arc shapes generated during the welding process, while taking into account the FLUENT's treatment for the VOF free interface, the heat  16) and (17). When the arc stays at the lower groove or the upper groove, the deflected double elliptical heat source model [14] is adopted, expressed by Eqs. (18) and (19).
where q f and q r are the heat flow density distribution before and after the arc center, respectively. Q l and Q r are the heat flow distribution of the arc staying at lower groove and upper groove, respectively. a f and a r are the long semi-axis of the front semiellipse and back semiellipse, respectively, while b h is the other semi-axis. b l and b r are the long semiaxes of the left semiellipse and right semiellipse, respectively, and the other semi-axis is l a . I is the welding current, U is the welding voltage, and η is the welding efficiency.

Distribution model of arc pressure
Arc pressure is an important surface force, and its distribution form is closely related to the current density. Therefore, similar to the heat source model, different arc pressure models are used in the arc swing stage and the arc stay stage on both sidewalls. The arc pressure model for the arc swing stage is shown in Eqs. (20) and (21). For the arc stay stage, the deflected arc pressure model is expressed by Eqs. (22) and (23).
where I is the welding current and μ is the vacuum magnetic permeability. a f , a r , and b h are the distribution radius of arc pressure in swing stage. p l , p r , and p a are the distribution radius of arc pressure the arc stay stage on both sidewalls.

Distribution model of surface tension
At present, the influence of surface tension on the flow of molten pool is generally described by the surface tension model of Fe-S binary alloy studied by Sahoo and Debroy [22]. The surface tension coefficient can be expressed by Eq. (24).
where γ 0 is the value of surface tension at the melting temperature of the pure metal, A r is the surface tension constant, T is the temperature, T s is the solidus temperature of the material, R g is gas constant, Γ s is the supersaturation degree, k 1 is a constant, α s is the elemental sulfur content, and ∆H 0 is the standard absorption enthalpy.

Distribution model of electromagnetic force
The electromagnetic force has less influence on the morphology and flow of the molten pool compared to other thermal forces, so the simplified model proposed by Wu [23] was used for the electromagnetic force, as shown in Eqs. where F x , F y , and F z are the components of electromagnetic force in X, Y, and Z directions, respectively. u m is the magnetic permeability of the material, r is the radial distance from the center of the arc heat source, δ j is the welding current distribution parameter, and H is the effective thickness of the workpiece.

Buoyancy distribution model
Due to the existence of both the temperature gradient and composition gradient in the molten pool, the density of liquid metal inside the molten pool changes, resulting in the buoyancy generation. The effect of buoyancy on the molten pool in this numerical model adopts Boussinesq model [24] which can be expressed as (24) where ρ is the density of the liquid metal, G is the gravity acceleration, β is the expansion coefficient of the material, and T ref is the liquidus temperature.

Gravity distribution model
The influence of the gravity on molten pool is related to the welding position, as shown in Fig. 8. The influence of the gravity on a molten pool at different welding positions is determined by the component force in each direction in the rectangular coordinate system. Then, the components in all directions are added in the FLUENT gravity module with the following specific numerical expressions.
The effect of gravity on swing process at any welding position G b can be expressed as The effect of gravity on welding direction at each welding position G τ can be expressed as The effect of gravity at each welding position on the direction of melt depth G r can be expressed as where θ is the angle between the current welding position and the 0° position of the pipe.

Molten drop transition
The physical processes involved in the droplet transition process are complex, as shown in Fig. 9. The precise numerical model of the molten drop transition will result in high computational costs, so the molten drop transition is somewhat simplified based on the ability to accurately reflect the effect of molten drop on the energy, mass, and momentum of the molten pool.
To realize the molten drop transition process in FLUENT, the UDF secondary program is compiled to control the speed inlet to be opened and closed in a specific cycle. The opening time is the molten drop generation time. At the same time, the UDF secondary program is used to control the volume fraction of steel phase at the velocity inlet to be 1 and the temperature to be 2400 K. After the above settings, when the speed inlet is opened, a molten drop with a certain speed and temperature will be formed, thus realizing the numerical simulation of the molten drop transition.

Free surface boundary conditions
The energy boundary conditions of the molten pool mainly include both the heat input part and heat dissipation part. The input energy is mainly in the form of both the arc heat and the molten drop heat, and the heat dissipation is mainly in the form of radiation and convection heat dissipation. The mathematical expression of the temperature boundary condition is shown in Eq. (34).
where λ is the heat transfer coefficient, q s is the energy source, q cov is the convective heat dissipation, α is the convection heat dissipation coefficient, q rad is the radiation heat dissipation, and ε r is the radiation coefficient. δ is the Boltzmann constant, and T ∞ is the ambient temperature.
The normal and tangential momentum boundary conditions of the upper surface of the molten pool can be expressed by Eqs. (37) and (38).
where P is the pressure in the direction normal to the upper surface, and → V n is the velocity vector in the normal direction; P A and P d are the arc pressure and the molten drop impact force, respectively; γ is the surface tension coefficient; R 1 and R 2 are the radius of curvature for the molten pool surface; → V t is the tangential velocity vector; and T is the surface tension temperature gradient.

Model boundary conditions
The air-layer side and the workpiece surface in the established numerical model are solid boundaries, and the speed is 0. The heat dissipation conditions are radiation and convection. The air layer along the welding direction is the pressure outlet, and the temperature is 300 K. The specific boundary conditions are shown in Table 2. Figure 10 shows the numerical simulation results compared with experiments. It can be seen that the weld cross-section and molten pool size obtained by numerical simulation agree well with those obtained by the experiment. This indicates that the numerical model established has a high degree of accuracy and can be used to analyze the transient heat and mass transfer behavior of the molten pool in the automatic welding process under steep slope conditions.

Construction of computational domain of medium convex weld
At the overhead welding position, the molten pool far from the arc center will remain in fluid state for a long time under the effect of gravity stretching, which has a great impact on the solidification of the molten pool. As shown in Fig. 11, the middle convex weld is formed at the overhead welding position, and there are sharp corners at the bottom of both the upper and lower groove. Accordingly, the VOF calculation domain is designed closer to the molten pool, which is used for the numerical simulation of next weld pass to study the relationship between lack of fusion defect and the medium convex appearance of the weld. Figure 12 shows the temperature field and flow field of 3D molten pool when the arc is in the dwell stage of the lower groove. The red area above the liquidus temperature (1790 K) represents the molten pool, the white arrow represents the arc center at this moment, the green solid line represents the lower groove position, and the red solid line represents the upper groove position. When the arc is in both the dwell stage of the upper groove and lower groove, the arc is deflected. When the arc firstly stays at the lower groove, the molten pool is formed and the overflow phenomenon occurs on the surface of the molten pool. When the arc secondly stays at the lower groove, the molten pool has a certain volume, and the width parameter of the molten pool does not change greatly, but the overflow phenomenon on the surface of the molten pool occurs again. When the arc thirdly stays at the lower groove, the molten pool reaches a relatively stable state, and overflow phenomenon still occurs on the surface of the molten pool. Figure 13 shows the temperature field and flow field of 3D molten pool when the arc is in the dwell stage of the upper groove. Similar to staying at the lower groove, the arc will also deflect. The difference is that there is no overflow at the front end of the molten pool. It is obvious that when the arc swings to the upper or lower groove in the fully automatic welding process under large slope, the surface shapes and flow behaviors of the molten pool are different, and the overflow is more likely to occur at the lower groove. Figure 14 shows the temperature field and flow field of 2D arc center cross-section with medium convex weld. The red area above the liquidus temperature (1790 K) represents the molten pool, the black arrow represents the arc center, and the white curved with arrows represents the flow trend of the liquid metal, the red arrow indicates the component of gravity on this cross section, and the

Air surface along welding direction
Pressure outlet 300 K It can be seen from Fig. 14 that after forming a stable molten pool, although the arc shape changes continuously in a swing cycle, the transient flow behavior of liquid metal in the molten pool is similar. No matter which stage the molten drop is in, the molten pool is diverted from the arc center to the upper and lower grooves on the whole, forming a wide range of eddy current at the side walls. However, it is difficult for liquid metal to flow to the bottom of the side walls, as shown by the yellow circle. With the transition of the molten drop to the molten pool, the flow trend of the liquid metal does not change greatly, and the eddy current effect is enhanced. This flow mode makes it easier for the high temperature liquid metal in the arc center to flow to the side walls for heat transfer, but it is difficult to flow to the "void area" at the bottom of the side wall for heat transfer. As the arc is in stay stage at the lower groove each time, the overflow will be generated, forming a "void area" with tension effect resulting from the existence of gas-liquid interface. In contrast, it is more difficult for liquid metal to transfer heat to the bottom of the lower groove. The numerical simulation results show that the fusion effect at the bottom of the upper groove is better than that at the bottom of the lower groove, so the lack of fusion defect occurs in the void area of the bottom of the lower groove.

Construction of computational domain of deflection weld
In the process of fully automatic welding under steep slope conditions, both the gravity and the deviation of welding wire have great influence on the solidification of molten pool. As shown in Fig. 15, the deflection weld occurs at the overhead welding position, namely, a sharp corner is located only at the bottom of the lower groove. Accordingly, the VOF calculation domain is designed closer to the molten pool, which is used for the next numerical simulation to study the relationship between lack of fusion defect and the deflection weld. Figure 16 shows the temperature field and flow field of 3D molten pool with deflection weld in formation stage. It is found that when the arc is in the dwell stage at lower groove, the arc deflects, and the overflow phenomenon of the molten pool occurs. When the arc swings rapidly towards to the upper groove after the dwell stage at lower groove, the arc deflection disappears, and the overflow at the front end of  the molten pool stays. However, when the arc is in the dwell stage at upper groove, the arc deflects, but no overflow phenomenon occurs at the front end of the molten pool which is different from the arc staying at lower groove. Figure 17 shows the temperature and flow fields of 3D molten pool with deflection weld in stable stage. After the deflected molten pool is stabilized, when the arc is again in the dwell stage at lower groove, the arc is deflected again, and the overflow phenomenon is also generated again at the front end of the molten pool. When the arc swings towards the upper groove or is in the dwell stage at the upper groove, the overflow phenomenon stops again, similar to the formation stage of the molten pool. Thus, it is obvious that when the arc stays at the lower groove each time, the arc deflects, resulting in overflow at the lower groove. Figure 18 shows the temperature field and flow field of 2D arc center cross-section with deflection weld. The red area above the liquidus temperature (1790 K) represents the molten pool, the black arrow represents the arc center, and the white curved with arrows represents the flow trend of the liquid metal; on this cross-section, the red arrow indicates the component of gravity, and the white arrow indicates the movement direction of the arc in the swing stage or the movement direction of the arc in the next stage after the dwell stage. It is noted that the transient flow behavior of liquid metal in the molten pool is similar after forming a stable molten pool. The arc shape changes constantly during a  large-scale eddy current. This flow mode makes it easier for the high temperature liquid metal located in the arc center to flow to the side walls for heat transfer, resulting in a good fusion effect. However, it is difficult for the liquid metal to flow and heat transfer to the void area at the lower groove. At the same time, when the arc is in the dwell stage at the lower groove, overflow will occur at this position, and a void area with the tension due to the existence of gas-liquid interface will be formed in the groove. It is more difficult for the liquid metal to transfer heat to the bottom of the lower groove due to the influence of the gas-liquid interface in the void area. Finally, the fusion effect at the bottom of the lower groove is poor, and the lack of fusion defect occurs.

Mechanism of lack of fusion defect in fully automatic welding under steep slope conditions
At the overhead welding position, the flow field of the next weld pass on the deflection and middle convex weld is analyzed. It is found that the heat transfer and transient flow behavior of liquid metal in the molten pool are complex and changeable under the influence of steep slope. The transient flow behavior of liquid metal is obviously different when it flows near the upper groove and lower groove. This causes the heat transfer effect of liquid metal to differ at these two positions, resulting in different melting effects and effects on the lack of fusion defect. The generation process of lack of fusion defect in fully automatic welding under steep slope conditions is shown in Fig. 19. As can be seen from Fig. 19 a and b, during the next welding on the convex and deflection weld beads, the arc thermal action center is deflected due to the groove influence when the arc stays in dwell stage on both sides of the groove. The liquid metal flows more to the side wall under the action of the deflecting arc to transfer heat. This makes the sidewall of upper and lower grooves more susceptible to stronger thermal action. However, it is not suitable to directly heat the sharp corner at the bottom of both upper and lower grooves. This will lead to the melting difficulty of the metal located at the sharp corner. At the same time, a different flow behavior will occur in the molten pool. Under the influence of surface tension, the molten pool flows upward along the side wall of the groove. When the arc stays at lower groove each time, if there is a sharp corner at the bottom of lower groove, the flow speed at the front end of the molten pool is very high, which will cause the liquid metal to easily break away from the molten pool and overflow to the front end of the lower groove. At the same time, at the overhead welding position, due to the slope of the pipeline, the liquid metal tends to gather in the lower groove. The force of the arc pressure and droplet impact force on the weld pool is upward, while the force of gravity on the weld pool is downward. When gravity plays a leading role in the flow of the weld pool, some liquid metal begins to flow out from the front end of the downward groove where the liquid metal gathers more. Thus, the molten fluid state at the front end of the molten pool is maintained for a long time. In the form of convection, radiation, and other forms of heat dissipation, the overflowing liquid metal maintains a certain shape at the sharp corner at the bottom of lower groove, forming a void area with a gas-liquid interface. The tension of the gas-liquid interface affects the direct thermal action of the arc on the void area to a certain extent. However, when the arc stays near the upper groove, this phenomenon does not exist on the surface of the molten pool. Figure 19 c shows the flow of molten pool in dwell stage at the lower groove. After the stable molten pool is formed, when the arc stays at upper and lower grooves, the liquid metal shunted to this position forms a large range of eddy current at the side walls. Therefore, the side walls of both upper and lower grooves in the welding process always bear a greater thermal effect, resulting in a good fusion effect. On the contrary, under the influence of side-wall eddy current at the void area located at the bottom of lower groove, more high-temperature liquid metal located in the arc center flows to the side wall for heat transfer. This to some extent hinders the flow and heat transfer of high-temperature liquid metal to the bottom of the side wall. In addition, there is also the tension of gas-liquid interface at the void area. Both of them cause melting difficulty in the void area and produce lack of fusion defect. Figure 19 d shows the flow of molten pool in dwell stage at the upper groove. The arc is also deflected when the arc stays at upper groove. The high-temperature liquid metal diverted to the upper and lower grooves also produces a wide range of vortex effects at the side wall. This makes the side wall produce a good fusion. Due to the combined action of gas-liquid interfacial tension and side-wall eddy current at the bottom of lower groove, it is difficult for liquid metal to flow to this position for heat transfer, resulting in melting difficulty. However, for the side-wall bottom of upper groove, its melting effect is significantly better than that of lower Fig. 19 Generation mechanism of lack of fusion defect in fully automatic welding under steep slope conditions. a arc deflection; b formation of molten pool; c flow of molten pool in dwell stage at lower groove; d flow of molten pool in dwell stage at upper groove groove. There are two reasons for this phenomenon. On the one hand, there is no void area with gas-liquid interface at the bottom of upper groove. On the other hand, the local eddy current will also be generated at the bottom of upper groove during droplet transfer, which facilitates the flow of the molten pool. In a word, the eddy current generated at the lower groove is small and has a large range of action, while the eddy current at the upper groove is small and dense, so the effect of surface tension at the lower groove is more obvious, and the upward flow trend along the side wall is stronger, the bottom of upper groove can be melted by the flow and heat transfer of liquid metal to a certain extent, while the bottom of lower groove is more likely to cause lack of fusion defect.

Conclusions
In view of the fact that the overhead welding position is easy to produce lack of fusion defect, the FLUENT software is used to establish the numerical model of fully automatic welding X70 pipeline steel under steep slope conditions. The influence of the medium convex weld and deflection weld on the temperature field and flow field of the next weld pass is analyzed, and the reasons of producing lack of fusion defect at a certain gradient are studied. Some main conclusions are as follows: (1) Considering the characteristics of the fully automatic welding process under steep slope conditions, a corresponding 3D transient heat and mass transfer model of molten pool was established. In addition, the model fully considers the influence of arc trajectory, arc shape, transfer of molten drop, and surface and volume forces on the molten pool flow. The accuracy of this numerical model was verified by the cross section of the weld and the surface profile of the weld pool. (2) Numerical analysis of the next welding pass on both deflection and middle convex welds shows that under the influence of surface tension, there is always a large range of eddy current at both side walls of the groove, and the eddy current transfers the heat of the molten pool to the side wall, which makes good fusion effect at both side walls. The slope makes the liquid metal tend to gather at the lower groove, the force of the arc pressure and droplet impact force on the weld pool is upward, while the force of gravity on the weld pool is downward. When gravity plays a leading role in the flow of the molten pool, some liquid metal begins to flow out from the lower groove to the front end, resulting in an overflow phenomenon. The generation of overflow phenomenon and arc deflection cause the formation of a "void zone" at the structure mutation position in front of the groove. (3) In the automatic welding process under steep slope conditions, the generation of lack of fusion defect and the quality of previous welding pass are closely related. When the arc stays on both sides of the groove, the arc will be deflected, and the overflow will occur only at the lower groove, forming a "void area" with gas-liquid interface. The effect of surface tension at the lower groove is more obvious, and the trend of upward flow along the side wall is stronger. The bottom of the lower groove is more likely to cause lack of fusion defects due to the combined influence of interface tension and side wall eddy current. (4) The model ignores the influence of the shielding gas and the evaporation heat of the molten pool metal. The simulated molten depth and width have errors with the actual measurement results, and the influence of the periodic changes of current and voltage on the molten pool flow during the actual arc swing process is not considered in the simulation. Therefore, only qualitative analysis can be carried out for the lack of fusion defects. The model still needs to be improved. The above problems will be the focus of the next research. Data availability The data supporting the article content is presented in the article.

Declarations
Ethics approval The authors confirm that they have abided by the publication ethics and state that this work is original and has not been used for publication anywhere before.

Consent to participate
The authors are willing to participate in journal promotions and updates.

Consent for publication
The authors give consent to the journal regarding the publication of this work.

Conflict of interest
The authors declare no competing interests.