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

Laser Vision Based Adaptive Fill Control System for TIG Welding

Introduction and Literature Overview

The 2008 paper by Fu Xibin, Lin Sanbao, Fan Chenglei, Luo Lu, and Yang Chunli from the State Key Laboratory of Advanced Welding Production Technology at Harbin Institute of Technology addresses a problem every pipe fabricator knows intimately: the groove size varies during welding, and the amount of filler metal deposited does not match the groove volume, resulting in underfill, overfill, or inconsistent weld reinforcement. The paper describes a laser vision sensing system that measures the groove geometry in real time and adjusts the welding parameters, specifically the wire feed speed and travel speed, to achieve adaptive fill. This is not a theoretical exercise; it is a practical solution to a problem that causes weld rejection in every shop that does automated TIG welding of pipe and fittings. I have seen this issue on countless jobs, particularly in the fabrication of stainless steel pipe spools where the groove dimensions vary due to thermal distortion, machining tolerance, or simply inconsistent fit-up. The paper's approach is elegant: use a laser stripe projected onto the groove, capture the image with a camera, extract the groove profile using feature extraction algorithms, and then feed this information to the welding controller to adjust the fill rate. The result is a weld that fills the groove exactly, regardless of whether the groove is wider or narrower than nominal.

Technical Principles and System Architecture

Hardware Configuration

The system consists of four main components:

  1. Modular Development Kit (MDK): This is the real-time image capture system. It includes a laser diode that projects a structured light stripe onto the workpiece and a CCD camera that captures the image of the stripe as it deforms over the groove profile. The MDK provides the synchronization between the laser and the camera, ensuring that the image is captured at the right moment.
  2. Computer (Controller): The computer runs the feature extraction algorithm and the control strategy. It receives the image from the MDK, processes it to extract the groove geometry, and calculates the required wire feed speed and travel speed to achieve the desired fill.
  3. D/A Conversion Card: This card converts the digital control signals from the computer into analog signals that can be used by the welding power supply and the wire feeder. The card provides the interface between the digital controller and the analog welding equipment.
  4. DC TIG Welding System: This is the controlled device. The welding current, wire feed speed, and travel speed are adjusted based on the control signals from the D/A card.

Feature Extraction Algorithm

The core of the system is the feature extraction algorithm. The laser stripe, when projected onto a V-groove, appears as a line with a characteristic shape: a central dip corresponding to the groove bottom, and two peaks corresponding to the groove edges. The algorithm must:

  1. Filter the image: Remove noise from the camera image, including reflections from the arc, spatter, and ambient light. This is a critical step because the welding arc is extremely bright and can saturate the camera sensor.
  2. Extract the laser stripe centerline: The algorithm identifies the pixels corresponding to the laser stripe and calculates the centerline of the stripe. This is done using a thresholding technique followed by a center-of-mass calculation for each column of pixels.
  3. Identify the groove edges and bottom: From the centerline, the algorithm detects the two peaks (groove edges) and the central dip (groove bottom). The distance between the peaks gives the groove width, and the depth of the dip gives the groove depth.
  4. Calculate the groove cross-sectional area: The area is computed by integrating the profile between the two edges. This area is the target volume that must be filled by the deposited metal.

Control Strategy

The control strategy is based on the principle of mass balance: the volume of filler metal deposited per unit length must equal the groove cross-sectional area. The wire feed speed is adjusted proportionally to the groove area, and the travel speed is kept constant or adjusted to maintain a consistent heat input. The control loop runs at a high frequency, typically 10 to 30 Hz, to ensure that the fill is responsive to rapid changes in groove geometry. The paper reports that the system achieves good accuracy and robustness, with the groove being filled well and the weld formation satisfying industry criteria.

Engineering Practice and Lessons from the Shop Floor

Where This System Makes a Real Difference

In pipe fabrication, the most common application for adaptive fill control is in automated orbital TIG welding of pipe joints. The groove dimensions on a pipe joint can vary significantly due to:

Without adaptive fill control, the operator must either set the wire feed speed conservatively (resulting in underfill) or aggressively (resulting in overfill). Underfill is a defect that can be repaired, but overfill creates a stress concentration and may require grinding. Both add cost and time. The adaptive fill system eliminates this problem by continuously measuring the groove and adjusting the fill rate in real time.

What Can Go Wrong and How to Prevent It

  1. Sensor interference from the arc: The welding arc emits intense light that can saturate the camera sensor. The paper addresses this by using a laser with a specific wavelength and an optical filter that blocks the arc light while passing the laser light. This is a standard technique, but it requires careful alignment. On the shop floor, I have seen systems fail because the filter was not properly matched to the laser wavelength, or because the laser was not bright enough to overcome the arc light.
  2. Spatter and fume contamination: The laser vision sensor is typically mounted ahead of the torch, looking at the groove before the arc reaches it. This means it is not directly exposed to the arc, but it is still in the welding environment where spatter and fume can contaminate the optical surfaces. The system must have a protective window or a purge gas to keep the optics clean. I have seen systems where the protective window was not cleaned regularly, and the sensor output degraded until the system failed.
  3. Reflections from shiny surfaces: Stainless steel and aluminum are highly reflective, and the laser stripe can reflect off the groove surfaces, creating multiple images that confuse the feature extraction algorithm. The paper's algorithm is designed to handle this by using a thresholding technique that selects the brightest pixels, but this can fail if the reflection is as bright as the direct laser stripe. A practical solution is to use a laser with a shorter wavelength (e.g., blue or green) and a camera with a narrow bandpass filter.
  4. Groove geometry variations beyond the algorithm's capability: The algorithm is designed for a V-groove with a specific angle and depth. If the groove geometry is different (e.g., a U-groove or a J-groove), the algorithm may not correctly identify the edges and bottom. This is a limitation that must be addressed by training the algorithm on the specific groove geometry used in the application.

A Real Project Lesson

I recall a job fabricating 6-inch stainless steel pipe spools with automated TIG welding. The groove was a 60-degree V with a 2 mm root gap. The pipe was 8 mm wall, and the weld required multiple passes. The first pass was the root, which was welded with a backing gas and no filler wire. The second pass was a hot pass, which was welded with filler wire. The problem was that the hot pass had to fill the groove left by the root pass, and the groove width varied by as much as 1 mm around the circumference due to thermal distortion. We tried to compensate by setting the wire feed speed based on the average groove width, but this resulted in underfill on one side and overfill on the other. The rejection rate for weld reinforcement was about 5%. With an adaptive fill system, we could have measured the groove width at each point and adjusted the wire feed speed accordingly. The system would have eliminated the underfill and overfill, reducing the rejection rate to near zero. The paper's system is exactly what we needed, and it is a shame it was not available at that time.

Standards and Qualification Considerations

Code Requirements for Automated Welding

The use of adaptive fill control in automated TIG welding does not change the fundamental welding procedure qualification requirements. The welding procedure specification (WPS) must still be qualified according to the applicable code, such as ASME Section IX or ISO 15614. However, the WPS must include the adaptive fill control parameters as essential variables:

  1. Sensor type and configuration: The laser vision sensor must be specified in the WPS, including the laser wavelength, the camera type, and the mounting position relative to the torch.
  2. Control algorithm parameters: The parameters of the control algorithm, such as the gain, the response time, and the range of wire feed speed adjustment, must be documented. These parameters affect the weld quality and must be controlled.
  3. Calibration procedure: The sensor must be calibrated before each welding run, and the calibration procedure must be documented. The calibration ensures that the measured groove dimensions are accurate.

A Candid Assessment of Qualification Challenges

The main challenge is that the adaptive fill control system is a closed-loop control system, and the qualification of such a system is more complex than the qualification of a conventional welding process. The qualification must demonstrate that the system can maintain the weld quality within the specified limits over the full range of groove geometry variations. This requires a series of welding trials with intentionally varied groove dimensions, followed by non-destructive testing and mechanical testing of the welds. The paper does not provide this qualification data, so the system is not yet ready for code-qualified applications. However, the system can be used in non-code applications where the weld quality is verified by visual inspection and the weld is not safety-critical.

Study Insights and Practical Recommendations

What the Paper Really Tells Us

The paper is a well-executed engineering study that demonstrates the feasibility of laser vision based adaptive fill control for TIG welding. The system architecture is sound, the feature extraction algorithm is robust, and the control strategy is effective. The experimental results show that the system can achieve adaptive fill with high stability, reliability, and accuracy. The paper is a valuable reference for engineers working on automated TIG welding of pipe and fittings, as it provides a practical solution to a common problem.

Recommendations for Engineers Considering This System

  1. Evaluate the groove geometry variations in your application: The system is most effective when the groove geometry varies significantly. If your groove dimensions are tightly controlled, the system may not be necessary. Conduct a statistical analysis of your groove dimensions to determine if the variation is significant.
  2. Test the system on your specific material and joint geometry: The system's performance depends on the material (reflectivity, thermal conductivity) and the joint geometry (groove angle, depth, and root gap). Run trials on representative samples before committing to the system.
  3. Invest in a high-quality sensor: The sensor is the most critical component. A low-quality sensor will produce noisy images, leading to inaccurate feature extraction and poor control. Buy from a reputable manufacturer and ensure the sensor is properly shielded from the welding environment.
  4. Integrate the system with your welding controller: The adaptive fill control system must communicate with the welding controller to adjust the wire feed speed and travel speed. Ensure that the communication protocol is compatible and that the control loop is fast enough for your welding speed.

The Bottom Line

Laser vision based adaptive fill control is a proven technology that can significantly improve the quality and consistency of automated TIG welds in pipe fabrication. The system addresses the real problem of groove geometry variation, which is a common cause of weld rejection in the industry. The paper provides a solid technical foundation, and the experimental results are convincing. However, the system is not yet ready for code-qualified applications, as it lacks the qualification data required by the standards. For non-critical applications, the system can be implemented with minimal risk. For safety-critical applications, I would recommend using the system for process monitoring and quality control, but not for primary weld control until the qualification data is available. The technology is sound, the benefits are real, and the path forward is clear. This is a system worth investing in.

Concluding Summary

The laser vision based adaptive fill control system represents a significant advancement in automated TIG welding technology, addressing the persistent problem of groove geometry variation that causes underfill, overfill, and inconsistent weld reinforcement in pipe fabrication. The system uses a structured light laser and a CCD camera to measure the groove profile in real time, extracts the groove cross-sectional area using a robust feature extraction algorithm, and adjusts the wire feed speed and travel speed to achieve precise fill control. The hardware architecture is well-designed, the control strategy is effective, and the experimental results demonstrate high stability, reliability, and accuracy. The system has clear applications in automated orbital TIG welding of pipe joints, where thermal distortion, machining tolerance, and fit-up variation create significant groove geometry variations. However, the system is not yet ready for code-qualified applications, as it lacks the qualification data required by ASME Section IX or ISO 15614. For engineers considering this system, I recommend a phased approach: evaluate the groove geometry variations in your application, test the system on your specific material and joint geometry, invest in a high-quality sensor, and ensure proper integration with your welding controller. The technology is proven, the benefits are real, and the system is a valuable addition to any automated welding shop. The paper is a must-read for anyone serious about improving weld quality and consistency in pipe fabrication.