ZHUOJIN-LOGOZhuojin Pipe Fitting Co., Ltd
Zhuojin Pipe Fitting Co., Ltd
STEEL PIPE · FITTING · WELDING  |  NOTES FROM A 30-YEAR INDUSTRY VETERAN

Twin-Wire Pulsed MIG Welding of Aluminum: Synchronized Molten Pool Vision Sensing and Image Processing

Literature Overview

This paper from the Journal of Mechanical Engineering (2014, Vol. 50, No. 12) tackles a problem I have seen burn more than one production line: getting a clean, usable image of the molten pool during aluminum twin-wire pulsed MIG welding. The authors, Yang Jiajia and colleagues from Nanjing University of Science and Technology, set out to build a synchronized dual-direction vision sensing system that could capture both the tail and side views of the pool simultaneously, and then process those images to extract reliable geometric parameters.

The core difficulty is well known to anyone who has stood behind a welding screen watching aluminum being welded. The double arc from twin wires produces an intense, broad-spectrum light that overwhelms any ordinary camera. On top of that, aluminum's high reflectivity means the arc light bounces off the pool surface and creates a glare that washes out the pool boundary. The authors attack this on two fronts: hardware filtering using near-infrared technology and precise control of the CCD shutter exposure time, and software processing using a Butterworth low-pass filter to remove the high-frequency horizontal striping interference that plagues these images.

Core Technical Points

The Sensing System Architecture

The authors describe a system that captures two views simultaneously. This is not a trivial achievement. In twin-wire welding, the two wires are typically arranged in tandem, with a leading wire and a trailing wire. The pool shape is asymmetric and dynamic, and a single view cannot capture the full geometry. The synchronized dual-view approach allows the operator to see the pool from the tail direction and from the side at the same instant, which is essential for understanding the actual pool behavior.

The Image Processing Chain

The processing chain is where the real engineering value lies. The authors use a Butterworth low-pass filter for enhanced filtering to remove the high-frequency horizontal striping. This striping is a characteristic artifact of the interaction between the pulsed arc and the CCD sensor's scanning. After filtering, they apply gray-level stretching, threshold segmentation, and edge extraction to obtain the pool contour. The result is a clean, binary representation of the pool boundary that can be measured.

Geometric Feature Parameterization

The most valuable contribution from a production standpoint is the geometric parameterization. The authors found that the front-middle portion of the pool is elliptical and the tail portion is parabolic. They then define and calculate parameters including pool width, total length, area, perimeter, and a tail coefficient. These parameters give the welding engineer quantitative metrics that can be correlated with process stability and weld quality.

Integration with Engineering Practice

Why This Matters on the Shop Floor

In my experience, the single biggest cause of aluminum weld defects in production is not the welding parameters themselves but the inability to see what is happening at the pool. Operators tune the parameters based on sound and arc appearance, which is subjective and inconsistent between shifts. A system that can provide real-time geometric feedback allows the process to be controlled objectively. The pool width and length directly correlate with penetration and fusion, and the tail coefficient is an indicator of the pool's ability to wet out and avoid lack of fusion.

The Practical Challenge of Implementation

I have seen vision systems installed on production lines that failed because the engineers did not account for the real environment. The welding cell is full of smoke, spatter, and reflected light from the workpiece and fixtures. The near-infrared filtering helps, but the system must also be protected from spatter and the lens must be kept clean. In my experience, a purged enclosure with a replaceable protective window is essential for any production deployment.

Common Defects and Countermeasures

Defect Root Cause Countermeasure
Lack of fusion Pool too narrow, insufficient wetting Increase pulse current, adjust wire angle, increase pool width
Porosity Pool turbulence, gas entrapment Reduce welding speed, improve shielding, check gas flow
Burn-through Pool too long, excessive heat input Reduce current, increase speed, adjust tail coefficient
Undercut Pool too wide, edge wetting poor Reduce current, adjust torch angle, improve edge preparation

Study Insights and Reflections

The parameterization approach is sound, but I would push back on one point. The authors define the pool shape as elliptical in the front and parabolic in the tail, which is a simplification. In real production, especially with twin-wire welding at high speed, the pool shape is influenced by the weld gap, the fit-up, and the presence of tack welds. The parameters are useful for process development and monitoring, but they should not be used as a substitute for direct inspection of the finished weld.

Another practical consideration is the frame rate. The authors do not specify the exact frame rate, but for pulsed welding with a frequency of 50-100 Hz, the camera must capture at least 200 frames per second to resolve the pool dynamics within each pulse cycle. If the system captures at a lower rate, the geometric parameters will be averaged and the correlation with process stability will be weakened.

Reference Value and Outlook

This paper is a solid foundation for anyone looking to implement vision-based process monitoring in aluminum welding. The combination of hardware filtering and software processing is a practical approach that can be adapted to other welding processes. The geometric parameters defined here could be extended to include more sophisticated metrics such as pool oscillation frequency and amplitude, which are known to correlate with penetration depth.

The next step in this line of work should be closed-loop control. Once the pool geometry can be measured reliably, the welding parameters can be adjusted in real time to maintain a target pool shape. This would be a significant step toward fully automated aluminum welding, which is currently limited by the difficulty of maintaining consistent quality in mass production.