Application of AI technology in automatic welding of steel box girder members

Box girder steel structure is widely used in Bridges, buildings and large equipment, and its high strength and durability put forward high requirements for welding quality.

Its cross section is box-shaped and usually consists of a top plate, a bottom plate, a web (vertical supports on both sides) and an internal partition to form a closed hollow section. Box girder is widely used in bridge engineering because of its structural characteristics.

With the development of industrial automation, traditional manual welding is gradually replaced by intelligent welding solutions. Welding of large components is not only a key part of the manufacturing process, but also a challenge of technical difficulty and complexity. Although it has made remarkable progress in some areas, it still faces many unique difficulties and challenges in the field of large-scale component welding.

1 Welding process adaptability and flexibility challenges

Large components often have complex structures, diverse materials and special performance requirements, which makes it difficult to directly apply the traditional welding process.

2. Process flexibility

Unlike the batch manufacturing model, large components in the field of mechanical manufacturing and building steel structures are usually small-batch and non-standardized. This feature requires a high degree of flexibility in the welding process, which can quickly adapt to components of different shapes, sizes and materials.

3. Welding automation and intelligent problems

In the welding of large components, because the workpiece is large, heavy and difficult to move flexibly, the welding equipment often needs to work around the workpiece. This mode of operation puts forward higher requirements for the flexibility, stability and accuracy of welding equipment. At the same time, it also increases the difficulty of equipment layout, debugging and maintenance.

Most welds are irregular curves (such as variable cross-section and arc), and it is difficult for robots to accurately track complex trajectories, which are prone to defects such as welding bias and non-fusion.

- Large box girder size error, resulting in a preset path and the actual weld deviation.

4. Welding quality control

The welding quality of large components is directly related to the structural strength, durability and safety of the product. Welding quality control is particularly complex.

 Welding safety and environmental protection issues

1. Welding safety:

High temperature, high pressure, harmful gas and other dangerous factors are involved in the welding process of large components, which poses a threat to the safety of operators.

2. Environmental issues:

The smoke, harmful gas and noise produced in the welding process pollute the environment.

To sum up, multiple problems. All affect the efficiency of the enterprise, product quality, and cost control.

A steel structure head enterprise, through the creation of digital smart factory, break the industrial upgrading, to create new quality productivity.

Adopt Minyue technology based on AI technology without teaching adaptive welding scheme.

High precision perception + intelligent decision + flexible execution. With self-developed RobotSmart- intelligent decision system, SmartVision- binocular structured light vision system, SmartEye- laser welding seam tracking system. Digital-analog flexible solution, no digital-analog intelligent solution.

1. Support the welding task order issued by MES, and send the drawings and plans directly to the welding workstation.

2. RobotSmart - Use digital twin technology to build twin welding workstation in the computer, analyze 3D model data, and automatically extract welds. From the movement trajectory of each robot to the parameters and production beats of each processing process, it can be accurately simulated.

Support multiple robots, multiple external axes, positioner, trajectory planning and task allocation, narrow space welding torch collision and other welding problems.

3. No need to import the model, scan the point cloud, reverse modeling, the workpiece is placed at will, the camera is coarse positioning, and the workpiece position information is quickly obtained.

No need for precise positioning of welding fixtures.

4.SmartEye- Extremely accurate positioning, support zero spacing weld tracking and locating. Laser tracking +AI compensation uses laser vision sensors to scan welds in real time, combined with deep learning algorithms to predict trajectory deviations and dynamically correct the position of the welding torch. Intelligent process library, support butt, lap, fillet weld, multi-layer multi-channel, welding adaptive, simple parameter input, everyone is a master of the process.

Through the technical closed loop of high-precision perception + intelligent decision + flexible execution, robot box girder welding is evolving from "automation" to "autonomy", but it needs to further break through the balance point between cost and reliability.