Visual Detection of TIG Welding Molten Pool Shape Parameters

Literature Overview

This study by Gao Jinqiang and Wu Chuansong (2000, Acta Metallurgica Sinica, Vol. 36, No. 12, pp. 1284-1288), funded by the National Natural Science Foundation of China (Grant No. 59875053), presents a systematic approach to real-time visual monitoring of TIG welding molten pool geometry. The work originates from the Institute of Joining Technology at Shandong University of Technology and addresses a fundamental challenge in welding quality control: the inability to directly observe the molten pool during welding and the difficulty of correlating visible surface features with subsurface weld quality.

Core Technical Approach

The research team developed a composite filter lens system based on arc spectral analysis, which serves as the optical front-end of a CCD-based visual sensor. The key innovation lies in the spectral filtering strategy: TIG arcs emit intense radiation across a broad wavelength range, with the peak emission concentrated in the ultraviolet and visible regions. A conventional optical filter would either attenuate the molten pool signal excessively or fail to suppress arc background adequately. By analyzing the spectral distribution of the TIG arc, the authors designed a composite filter that selectively transmits wavelengths where the molten pool emits strongly while rejecting the dominant arc radiation bands.

The composite filter was paired with a standard CCD camera to form a compact visual sensor, observing the molten pool from the front (welding) side. A critical engineering challenge was maintaining adequate image brightness across varying welding currents. The authors implemented a software-based brightness adjustment algorithm that modulates exposure parameters in response to real-time current readings, ensuring consistent image quality regardless of welding parameter fluctuations.

Image Processing Algorithm and Performance

The authors developed a proprietary image processing algorithm capable of extracting front-face molten pool shape parameters including molten pool width, length, and aspect ratio. The algorithm processes a single frame in no more than 70 ms, satisfying the real-time requirement for closed-loop welding control systems. This processing speed is significant because typical TIG welding travel speeds range from 20 to 100 mm/min, meaning the molten pool geometry changes relatively slowly compared to the frame acquisition rate, making 70 ms processing time more than adequate for feedback control.

Technical Parameters and Process Window

Engineering Practice Implications

From a manufacturing quality control perspective, this visual detection system represents a significant advancement over conventional post-weld inspection methods such as radiographic testing (RT) and ultrasonic testing (UT). While RT and UT provide definitive assessment of internal weld defects, they are destructive in terms of throughput and cannot provide real-time feedback to the welding operator or automated control system. The visual monitoring approach enables:

1. Immediate detection of welding parameter drift during production runs
2. Early identification of incomplete fusion or excessive penetration before the defect propagates
3. Quantitative documentation of welding process stability for quality audits
4. Integration with adaptive control systems for automatic parameter correction

Key Observations and Reflections

The composite filter design is particularly noteworthy because it demonstrates that sophisticated optical hardware is not always necessary for effective molten pool monitoring. The spectral analysis approach shows that understanding the physics of arc emission and molten pool radiation is more important than simply using expensive imaging equipment. This insight has direct relevance to modern welding monitoring systems, where cost-effective solutions are often preferred over high-end optical setups.

The 70 ms processing time, while adequate for TIG welding, raises questions about applicability to faster welding processes such as GMAW or laser welding, where molten pool dynamics are significantly faster. However, the algorithmic framework and the adaptive brightness adjustment methodology can be transferred to faster processes with appropriate modifications to the temporal resolution requirements.

The study also implicitly addresses the challenge of visual sensor durability in welding environments. Arc radiation, spatter, and high temperatures can degrade optical components over time. The composite filter design must therefore balance spectral performance with environmental robustness—a consideration that becomes even more critical in industrial deployment scenarios.

Summary

This foundational work establishes that real-time visual monitoring of TIG molten pool geometry is technically feasible with commercially available CCD cameras and custom optical filtering. The combination of spectral analysis-driven filter design, adaptive brightness control, and efficient image processing algorithms creates a practical system that can support both manual operator guidance and automated welding quality assurance. The methodology described here forms the basis for more advanced molten pool monitoring systems that have since been developed for various welding applications, including orbital welding of pipe joints and automated TIG welding of thin-wall stainless steel components.