Zhuojin Pipe Fitting Co., LtdDetection of Molten Pool Geometric Parameters in TIG Welding via Image Processing
Literature Overview
The paper by Duan Ruixia, Ke Yu, and Xu Jianyu (2015), published in Hot Working Technology (Vol. 44, No. 9, pp. 236-239), addresses a fundamental challenge in TIG welding process monitoring: the real-time extraction of molten pool geometric parameters using visible-light and infrared imaging. The work originates from Ningbo University's School of Information Science and Engineering in collaboration with Ningbo Xinle Living Appliance Co., Ltd., reflecting an industry-academia partnership focused on practical welding quality assurance.
Core Technical Approach
The central insight of this study lies in the spectral characteristics of TIG welding light sources. The authors exploit the fact that the arc's continuous spectrum intensity decreases with increasing wavelength, while the molten pool's own thermal radiation increases with wavelength. This complementary behavior enables the design of a narrow-band composite filter system that suppresses arc glare while preserving the molten pool signal.
Spectral Filtering Strategy
The filter design is critical to image quality. The approach involves:
- Selecting filter bands in the near-infrared region (approximately 700-1000 nm) where the arc emission is attenuated but the molten pool thermal radiation remains strong.
- Combining multiple narrow-band filters to create a composite optical window that optimizes the signal-to-noise ratio between the molten pool and the surrounding arc.
- Using a conventional industrial camera rather than a specialized high-cost imaging device, which significantly reduces system cost.
Molten Pool Extraction Algorithm
Given the high welding speeds typical in production environments, the molten pool is in a state of continuous flow, and the grayscale values in the pool region fluctuate dynamically. The authors adopt a frame-difference method to isolate the molten pool region:
- Capture consecutive video frames at a sufficient frame rate.
- Compute the pixel-wise difference between adjacent frames.
- Apply thresholding to the difference image to segment the moving molten pool region.
- Perform morphological operations to clean up noise and fill gaps.
- Extract geometric parameters from the segmented region.
Extracted Geometric Parameters
The system outputs the following key molten pool parameters:
| Parameter | Definition | Significance |
|---|---|---|
| Maximum width | Widest transverse dimension of the pool | Indicator of heat input and bead geometry |
| Maximum length | Longitudinal extent of the pool | Related to welding speed and thermal profile |
| Pool area | Total pixel count within the pool boundary | Composite indicator of thermal accumulation |
Engineering Practice Implications
For steel pipe and fitting manufacturers, molten pool monitoring represents a critical step toward closed-loop welding process control. In the context of ERW, HFW, and TIG welding of pipe joints, real-time molten pool dimensions can serve as process health indicators:
- Width deviation may signal incorrect current or travel speed settings, leading to excessive penetration or insufficient fusion.
- Length variation can indicate changes in thermal input, which directly affects the heat-affected zone (HAZ) width and residual stress distribution.
- Area fluctuation may reveal instability in the welding process, such as arc drift or shielding gas disruption.
Connection to Pipe Welding Quality
In longitudinal submerged-arc welded (LSAW) and spiral welded pipe production, the weld pool geometry directly governs:
- Penetration depth and fusion quality
- Weld bead profile and reinforcement height
- HAZ grain growth and microstructural integrity
- Residual stress magnitude and distortion
The frame-difference method described in this paper, while relatively straightforward, provides a cost-effective foundation for in-line monitoring systems. However, for high-speed pipe welding applications (travel speeds exceeding 1000 mm/min), the frame rate and processing speed become limiting factors, necessitating hardware-accelerated image processing or higher-frame-rate cameras.
Key Questions and Reflections
The paper raises several important considerations for engineering implementation:
- Frame rate adequacy: The frame-difference method requires consecutive frames to be temporally close enough that the molten pool does not change dramatically between frames. At high welding speeds, this constraint becomes severe.
- Environmental robustness: The optical system must withstand welding spatter, smoke, and ambient light variations on the production floor.
- Parameter correlation: The paper establishes that geometric parameters can be extracted, but does not fully explore the quantitative relationship between pool dimensions and final weld quality metrics such as tensile strength or fatigue life.
The study serves as a valuable proof-of-concept for optical molten pool monitoring. Its practical value lies in demonstrating that a relatively simple imaging setup, combined with basic image processing algorithms, can provide meaningful process information without requiring expensive infrared or high-speed camera systems.
Study Insights
The fundamental lesson from this work is that effective process monitoring does not always require sophisticated hardware. By understanding the physics of light emission and absorption during welding, engineers can design cost-effective optical solutions that extract actionable process data. The frame-difference approach, while limited in temporal resolution, is particularly suitable for medium-speed TIG welding applications such as pipe fitting repair, small-diameter pipe welding, and in-situ field welding where high-speed imaging is impractical.