CLADDING TECHNOLOGY SHANXI CO., LTDMIG Welding Molten Pool Image Processing and Digital Control System
Literature Overview
This foundational study by Wang Kezheng from Tsinghua University and Kenji Oshima from Saitama University, Japan, published in Journal of Tsinghua University (Science and Technology) in 1990, represents one of the pioneering works on real-time molten pool monitoring and digital control in MIG welding. The research establishes the theoretical and practical framework for using image processing techniques to monitor and control the MIG welding molten pool, laying the groundwork for modern intelligent welding systems.
Core Technical Framework
The system described in this paper consists of three integrated components: an image acquisition subsystem, a digital signal processing subsystem, and a real-time control subsystem. The image acquisition subsystem captures molten pool images using a high-speed camera with appropriate optical filters to isolate the visible and near-infrared spectral bands where the molten pool emits the most intense radiation.
The digital signal processing subsystem performs image segmentation to extract the molten pool contour from the background. The extracted contour is then characterized by geometric parameters such as pool width, pool length, pool area, and pool shape factor. These parameters are compared against reference values to determine whether the welding process is operating within the desired parameters.
| System Component | Function | Key Specification |
|---|---|---|
| Image acquisition | Capture molten pool images | Frame rate >100 fps, resolution adequate for pool measurement |
| Optical filter | Isolate molten pool radiation | Bandpass filter for visible/NIR spectrum |
| Signal processing | Extract pool geometric parameters | Real-time computation, sub-millisecond response |
| Control algorithm | Adjust welding parameters | PID or rule-based control |
| Actuator interface | Modify welding parameters | Current, voltage, travel speed, torch angle |
The digital control subsystem implements a feedback control loop that adjusts welding parameters based on the deviation between measured and reference molten pool characteristics. The control algorithm can be either a conventional PID controller or a rule-based system that applies predefined corrections for specific pool shape deviations.
Molten Pool Image Processing Methodology
The image processing pipeline begins with raw frame acquisition and proceeds through several stages: background subtraction to remove static features, thresholding to binarize the image based on the molten pool's luminance intensity, morphological operations to fill gaps and remove noise, and contour extraction to identify the pool boundary.
The molten pool in MIG welding exhibits characteristic features that are amenable to automated detection. The pool appears as a bright, roughly elliptical region with a well-defined boundary against the darker solidified weld metal. The pool width is typically 8-15 mm for standard MIG parameters, while the pool length ranges from 15-30 mm depending on the welding speed and current.
The shape factor, defined as the ratio of pool width to pool length, is a particularly useful diagnostic parameter. A stable, well-formed pool typically exhibits a shape factor in the range of 0.4-0.6. Deviations from this range indicate process instability: a lower shape factor suggests excessive penetration and potential burn-through, while a higher shape factor indicates insufficient penetration and possible lack of fusion.
Engineering Practice and Legacy
Although published in 1990, the principles established in this study remain highly relevant to modern welding automation. The fundamental approach of monitoring pool geometry through image processing and feeding the results back into the control loop is the basis for all modern welding process monitoring systems. Contemporary implementations use more advanced image processing algorithms, higher resolution cameras, and faster processors, but the underlying methodology is unchanged.
In the context of cladding and overlay welding, real-time pool monitoring is essential for maintaining consistent overlay layer geometry and bond quality. The overlay layer thickness, which directly affects corrosion resistance and service life, is strongly influenced by the molten pool dimensions and dynamics. By monitoring the pool in real time and adjusting parameters accordingly, engineers can maintain the overlay layer thickness within tight tolerances, typically ±0.1-0.2 mm, which is critical for pressure vessel applications where the clad layer thickness is specified to meet design requirements.
For the fabrication of bimetal pressure vessels using weld-overlay cladding, the integration of pool monitoring with the welding control system enables adaptive welding that compensates for variations in base metal thickness, joint fit-up, and ambient conditions. This adaptive capability is particularly important for large-scale fabrication where manual parameter adjustment for each weld is impractical.
Key Reflections
This 1990 study is a testament to the forward-looking vision of its authors, who recognized early on that real-time visual monitoring could transform welding from a process governed by empirical rules into one governed by quantitative feedback control. The digital control framework described here has evolved over the past three decades into sophisticated welding automation systems, but the core concept remains the same: observe the process, characterize the state, and adjust the inputs to maintain quality. Engineers working in modern welding automation should understand this foundational work as it provides the conceptual basis for all contemporary process monitoring approaches.