Information Fusion Technology in Laser-TIG Hybrid Welding Monitoring

Literature Overview

This 2010 paper from Beihang University, published in the Journal of Shanghai Jiao Tong University, presents research on information fusion technology applied to real-time monitoring of laser-TIG hybrid welding processes. Laser-TIG hybrid welding combines the deep penetration of laser welding with the wider fusion and better bead shape of TIG welding, making it particularly suitable for thick-section welding and cladding applications. Real-time monitoring and control of this hybrid process requires the integration of multiple sensor signals, which is the focus of this study.

Core Technical Content

The laser-TIG hybrid welding process is characterized by complex interactions between the laser beam and the TIG arc, resulting in a dynamic weld pool with rapidly changing geometry and thermal conditions. Monitoring this process in real time requires sensors capable of capturing different aspects of the welding phenomenon, and information fusion techniques are used to combine these signals into a comprehensive assessment of weld quality.

Sensor Configuration and Signal Types

The study employs multiple sensing modalities to monitor the hybrid welding process:

Sensor Type	Signal Captured	Information Content	Sampling Rate
High-speed camera	Weld pool shape, arc morphology	Pool geometry, spatter, arc stability	1000–10000 fps
Optical fiber sensor	Arc intensity, UV/IR radiation	Arc power, shielding gas quality	10–100 kHz
Acoustic sensor	Welding sound, arc noise	Porosity, lack of fusion, arc instabilities	20–200 kHz
Electrical sensor	Arc voltage, welding current	Arc length, penetration depth	10–50 kHz
Infrared pyrometer	Surface temperature distribution	Heat input, cooling rate	1–10 kHz

Information Fusion Methodology

The information fusion approach combines data from multiple sensors using hierarchical fusion architectures:

Data-level fusion: Raw signals from different sensors are aligned in time and combined at the lowest level. This provides maximum information retention but requires precise time synchronization and is computationally intensive.
Feature-level fusion: Features are extracted from each sensor signal (e.g., pool width from optical, arc frequency from electrical, spectral peaks from acoustic), and these features are combined. This is the most common approach in practice.
Decision-level fusion: Each sensor independently assesses weld quality, and the decisions are combined using voting or Bayesian methods. This is robust to individual sensor failures but may lose some information.

Key Monitoring Parameters

For laser-TIG hybrid welding, the critical parameters monitored include:

Weld pool width and length: Indicators of heat input and penetration
Arc stability index: Derived from electrical signal variance
Spatter rate: From optical imaging
Porosity probability: From acoustic signal analysis
Weld bead convexity: From optical profilometry

Engineering Practice Applications

In the context of cladding and bimetal pressure vessel fabrication, laser-TIG hybrid welding is used for:

Multi-pass cladding of thick sections with nickel-based alloys
Repair welding of pressure vessel components
Joining of dissimilar metals with controlled dilution

The information fusion monitoring system described in this paper can be adapted for real-time quality control in these applications. Key benefits include:

Early defect detection: Acoustic signals can detect porosity formation in real time, allowing immediate process correction before the defect propagates.
Process stability assessment: Arc stability monitoring ensures consistent weld quality throughout the production run.
Documentation and traceability: Continuous monitoring provides a complete record of welding conditions for quality assurance documentation.

Study Insights and Reflections

The integration of multiple sensing modalities represents a significant advancement in welding process monitoring. For engineers in the pressure vessel and cladding industry, the practical challenge lies in implementing such systems on the shop floor. The computational requirements for real-time information fusion have decreased significantly since 2010, making these systems more accessible. However, the fundamental challenge of sensor calibration, signal processing, and defect classification remains. The study demonstrates that no single sensor can provide a complete picture of weld quality; only through information fusion can reliable real-time monitoring be achieved. This principle is directly applicable to automated cladding systems where process control is critical for maintaining consistent overlay quality.