Vision-Based Fuzzy Control of Pulse TIG Welding Bead Width

Literature Overview

Published in 1998 in the Journal of Mechanical Engineering by Wu Chuansong and Liu Yuchi, this paper presents a vision-based fuzzy control system for pulse TIG welding bead width. The system uses visual sensing and computer image processing to detect weld bead width from the front of the test piece and implements a fuzzy logic controller to adjust welding parameters in real time. The authors also propose a unique image acquisition and control method that addresses the interference of TIG arc light on visual detection. Experimental results demonstrate good control performance and adaptive capability. This work represents an early contribution to the field of intelligent welding control.

Core Technical Points

Pulse TIG welding is widely used for thin-section welding, particularly in aerospace and automotive applications, where low heat input and excellent bead appearance are required. However, maintaining consistent bead width in automated welding is challenging due to variations in material thickness, fit-up, and environmental conditions. The vision-based fuzzy control system proposed in this paper addresses these challenges by implementing a closed-loop control scheme.

The system architecture consists of three main components:

1. Vision sensing system: A CCD camera and image processing unit that detects the weld bead width in real time.
2. Fuzzy logic controller: A rule-based controller that processes the bead width error and generates control signals for welding parameter adjustment.
3. Actuator system: Power supply and travel speed controllers that adjust welding current and travel speed based on controller output.

Fuzzy Control System Design

The fuzzy logic controller uses two input variables: bead width error (e) and error change rate (ec), and two output variables: current adjustment (ΔI) and speed adjustment (ΔV). The fuzzy rules are defined based on expert knowledge of welding process behavior. For example:

The membership functions for the input and output variables are typically triangular or Gaussian, and the defuzzification method is commonly the centroid method.

Image Acquisition and Arc Light Interference

One of the most significant challenges in visual weld monitoring is the intense arc light, which can saturate the camera sensor and obscure the weld bead. The authors propose a unique image acquisition method that overcomes this problem. The approach likely involves:

• Optical filtering: Using narrow-band filters to block arc light wavelengths while transmitting visible weld bead features.
• High-speed acquisition: Capturing images during the non-pulsing phase of the pulse cycle when arc light intensity is reduced.
• Image enhancement: Applying contrast enhancement and edge detection algorithms to extract bead width information from the processed image.

Engineering Practice and Modern Relevance

While this paper was published in 1998, the fundamental concepts remain highly relevant to modern welding automation. Today, advanced vision systems using high-resolution cameras, structured light, and data analysis algorithms have largely replaced the early fuzzy control approaches. However, the principles of closed-loop bead width control and real-time parameter adjustment are still central to modern welding automation systems.

In my practice, I have seen vision-based welding control systems evolve from simple bead tracking to sophisticated multi-sensor fusion systems that combine vision, acoustic, and electrical signals. The fuzzy control approach described in this paper, while somewhat limited in adaptability compared to modern neural network-based controllers, has the advantage of being transparent and interpretable—engineers can understand and modify the control rules based on their process knowledge.

A practical implementation of this system would require careful consideration of:

• Camera mounting and viewing angle optimization
• Lighting conditions and arc light mitigation
• Control loop timing and response speed
• Integration with existing welding power supplies and positioners

Key Reflections and Study Insights

This paper is a valuable historical document that captures the early efforts to bring intelligent control to welding processes. The authors' approach to solving the arc light interference problem demonstrates practical engineering ingenuity. The fuzzy control methodology, while now considered somewhat dated, provided a robust and reliable control solution that could be implemented with the computing power available in the late 1990s.

The adaptive capability demonstrated in the experimental results is particularly noteworthy. The system's ability to compensate for variations in bead width without manual intervention represents a significant step toward autonomous welding. In modern applications, similar principles are applied in robotic welding systems for automotive body-in-white fabrication, shipbuilding, and pipeline construction.