⚡ electrical engineering

Deep-penetration keyhole behaviour and weld pool morphology evolution mechanismsin K-TIG welding of low-carbon steel under variable inclination angles

This study investigates the destabilization mechanisms of K-TIG deep-penetration keyholes in low-carbon steel under variable inclination angles, revealing that the tangential gravity component drives directional backfilling and flow pattern shifts that lead to premature keyhole closure and incomplete back penetration, thereby establishing a theoretical basis for regulating all-position deep-penetration welding.

Original authors: Haihua Liu, Kangda Yue, Hao Zhang, Ruibin Duan

Published 2026-10-06

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Haihua Liu, Kangda Yue, Hao Zhang, Ruibin Duan

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine trying to pour a thick, glowing river of molten metal onto a surface that is slowly tilting away from you. In the world of heavy industry, where massive steel plates and long-distance pipelines are joined together, this is a daily reality. The process of welding these thick sections often relies on a technique called Keyhole Tungsten Inert Gas welding, or K-TIG. This method uses an incredibly powerful electric arc to punch a deep, narrow tunnel through the metal, allowing a single pass of welding to fuse plates that are eight millimeters thick. In a perfect, flat world, this tunnel stays open, and the molten metal flows evenly around it, creating a strong, seamless joint on both sides. However, the real world is rarely flat. When workers must weld vertical walls, curved tanks, or the underside of structures, the angle of the workpiece changes. The question researchers have long faced is whether this powerful welding method can survive the pull of gravity when the metal is no longer sitting level.

A team of scientists at Tiangong University in China set out to solve this puzzle by studying exactly what happens when the welding angle shifts. They focused on a specific type of low-carbon steel, eight millimeters thick, which is common in construction and pipelines. To understand the invisible forces at play, they combined real-world experiments with a sophisticated computer simulation. This digital model acted as a high-speed, microscopic camera, tracking how the molten metal moved, how heat spread, and how the deep tunnel behaved as the angle of the plate was tilted from flat up to thirty degrees. They wanted to see if simply turning up the heat or slowing down the welding speed could fix the problems caused by tilting, or if the physics of the situation made the task fundamentally different.

The researchers discovered that tilting the plate does not just make the weld look a little messy; it fundamentally breaks the delicate balance that keeps the deep tunnel open. In a flat position, the force of the electric arc pushing down, the surface tension of the liquid metal holding its shape, and the weight of the metal pushing down all work together in harmony. But as soon as the plate is tilted, a new force appears: the component of gravity that pulls the molten metal sideways, sliding it down the slope. The team found that as the angle increased from ten to thirty degrees, this sideways pull began to overwhelm the other forces. The deep tunnel, which is essential for the welding method to work, started to shrink, then wobble, and finally collapse.

At a thirty-degree angle, the situation became critical. The simulation showed that the tunnel could open for a brief moment, penetrating the full thickness of the steel, but it could not stay open. The sideways pull of gravity acted like a slow-moving landslide, dragging the liquid metal away from the bottom of the tunnel and filling it back in from the rear. This happened so quickly that the tunnel closed up just a few millimeters above the bottom of the plate. Once the tunnel closed, the intense heat from the electric arc could no longer reach the bottom of the steel directly. Instead of the powerful, focused heat of the open tunnel, the heat had to travel through the metal by slow conduction, which was not enough to melt the bottom layer completely. The result was a weld that looked fine on top but failed to join the metal all the way through.

Perhaps the most surprising finding was that simply adding more power to the weld could not fix this problem. The researchers tested whether increasing the electric current or slowing down the welding speed could keep the tunnel open against the pull of gravity. The answer was no. The force pulling the metal sideways is a directional force, acting along the surface of the plate, while the electric arc pushes straight down. No matter how hard the arc pushes down, it cannot stop the metal from sliding sideways. In fact, making the pool of molten metal larger by adding more heat only gave gravity more material to pull, making the collapse even worse. The study revealed that the flow of the metal changed from a gentle, circular swirl to a violent, high-speed stream that rushed down the slope, scouring the bottom of the tunnel and preventing it from staying open.

The team concluded that the failure of the weld in these tilted positions is not a matter of insufficient heat, but a matter of the wrong kind of force. The sideways pull of gravity is the only force that appears when the angle changes, and it disrupts the entire system. Because this force cannot be canceled out by the arc itself, the researchers suggest that fixing the problem will require new approaches. Instead of just trying to add more heat, future solutions may need to introduce external forces, such as magnetic fields or forced cooling, to counteract the slide of the molten metal. This work provides a clear map of why deep-penetration welding fails on slopes and points the way toward the specific tools needed to make it work again, ensuring that the massive structures of the future can be built safely, even when they are not standing flat on the ground.

Technical Summary: Deep-penetration keyhole behaviour and weld pool morphology evolution mechanisms in K-TIG welding of low-carbon steel under variable inclination angles

Problem Statement
Keyhole Tungsten Inert Gas (K-TIG) welding is a high-efficiency process capable of single-side welding with double-side formation for medium-thickness plates (e.g., 8 mm low-carbon steel) without groove preparation. While the process is well-understood in flat positions, practical engineering applications—such as girth welding of long-distance pipelines and large storage tanks—require welding at variable inclination angles. In these spatial positions, the gravity vector changes, disrupting the heat and mass transfer balance of the weld pool. Previous studies on conventional arc and laser hybrid welding have highlighted gravity-induced defects like directional flow and incomplete fusion. However, K-TIG presents a unique challenge: its deep-penetration mechanism relies on a delicate dynamic balance between intense arc pressure and surface tension to maintain a stable keyhole. The specific mechanisms by which variable angles degrade this balance, particularly the transition from stable penetration to keyhole closure, and the relative contributions of driving forces (arc pressure, surface tension, gravity components) in this degradation, remain insufficiently understood.

Methodology
This study investigates 8 mm Q235 low-carbon steel using a combined experimental and numerical approach:

Experimental Setup: K-TIG welding experiments were conducted at a flat position (0°) and variable inclination angles (10°, 20°, and 30°) using a 460 A welding current and 210 mm/min travel speed. The setup included a WSM-1000 inverter power source, a water-cooled K-TIG torch, and 99.99% pure argon shielding gas.
Numerical Simulation: A three-dimensional transient multiphysics model was developed using Fluent with the Volume of Fluid (VOF) method to track the gas-liquid interface. The model coupled:
- Governing Equations: Mass, momentum, and energy conservation equations for incompressible Newtonian fluid flow.
- Force Models: Gaussian arc pressure, surface tension (including Marangoni shear stress driven by temperature gradients), electromagnetic force, thermal buoyancy (Boussinesq approximation), and spatial gravity decomposition.
- Heat Source: A combined model featuring a double-ellipsoidal surface heat source and a dynamic conical heat source within the keyhole channel.
- Validation: The model was validated against 0° flat-position experiments, showing relative errors of 2.5% for top bead width and 4.8% for back bead width.
Simulation Strategy: Using the validated 0° model, simulations were extrapolated to 10°, 20°, and 30° by altering the gravity vector components ( $g \cdot \sin\theta$ tangential and $-g \cdot \cos\theta$ normal) while keeping all other process parameters constant.

Key Results

Keyhole Degradation Stages: As the inclination angle increased from 0° to 30°, the keyhole underwent a four-stage degradation process: stable penetration $\rightarrow$ size shrinkage $\rightarrow$ near closure $\rightarrow$ complete closure. At 30°, the keyhole penetrated briefly (0.6 s) but closed prematurely (2.5–3.2 mm from the bottom) due to directional backfilling of liquid metal.
Weld Formation Evolution:
- Back Bead Width: Decreased monotonically from 2.2 mm (0°) to 0 mm (30°), indicating a total loss of back penetration capability.
- Reinforcement: Increased non-linearly from 0.5 mm (0°) to 3.8 mm (30°), as liquid metal accumulated on the top surface rather than flowing through the plate.
- Morphology: The cross-section evolved from a symmetric "goblet" shape to a shallow-dish asymmetric profile.
Flow Field Transformation:
- 0°–10°: The pool maintained a symmetric circulatory convection pattern with a maximum velocity of $\approx$ 0.11 m/s.
- 20°–30°: The flow pattern mutated into directional scouring opposite the welding direction. At 30°, the quasi-steady maximum velocity exceeded 0.5 m/s, with instantaneous peaks reaching 0.81 m/s (approx. 6 times the flat-position peak).
- Mechanism: The tangential gravity component drove high-speed liquid metal toward the pool tail, thinning the front keyhole wall and scouring the rear wall, leading to premature closure.
Force Balance Analysis:
- The tangential component of gravity (approx. 275 Pa at 30°) is the only force that appears with inclination. Although its magnitude is smaller than the peak arc pressure, it acts continuously on all liquid metal elements.
- This component becomes comparable to the surface tension restoring force in shallow-curvature regions.
- The Bond number ( $Bo$ ) increased from 0 (flat) to $\approx$ 4.6 (30°), indicating gravity replaced surface tension as the dominant factor.
- Crucially, increasing heat input (current) cannot restore the balance because the arc pressure acts perpendicular to the plate, while the destabilizing gravity component acts tangentially; they are orthogonal and cannot offset each other.

Significance and Claims
The paper claims to reveal the full mechanistic chain linking keyhole behaviour, flow patterns, and heat transfer paths in variable-angle K-TIG welding. The primary significance lies in identifying that the root cause of formation deterioration is not insufficient heat input, but the "inability to maintain" the keyhole due to the tangential gravity component breaking the force balance.

The study concludes that for large inclination angles (e.g., 30°), the problem represents a "qualitative difference" rather than a "quantitative deficiency." Consequently, optimizing process parameters (current or speed) alone cannot overcome the formation deterioration. The authors propose that future regulation of all-position deep-penetration welding must shift from "supplementing heat" to "supplementing force," suggesting the necessity of external field interventions (such as external magnetic fields, pulsed current, or forced cooling) to rebuild the pool force balance.

More like this