Online Detection of Laser Welding Penetration Depth Based on Multi-Sensor Features.

Area of Science:

• Materials Science
• Manufacturing Engineering
• Optical Engineering

Background:

• Accurate online detection of laser welding penetration depth is crucial for industrial applications.
• TC4 titanium alloy welding presents specific challenges due to its material properties.

Purpose of the Study:

• To develop a multi-sensor monitoring system for real-time laser welding penetration depth detection.
• To investigate the correlation between process parameters, molten pool characteristics, plasma spectrum, and penetration depth.
• To establish high-precision predictive models for penetration depth.

Main Methods:

• Construction of a multi-sensor system capturing keyhole/molten pool images and laser-induced plasma spectra.
• Investigation of laser power effects on molten pool morphology and plasma characteristics.
• Extraction of image and spectral features using image processing and dimension-reduction techniques.
• Development of penetration depth prediction models using single-sensor and multi-sensor features.

Main Results:

• Significant correlations were identified among keyhole/molten pool variations, plasma spectrum, and penetration depth.
• A neural network model utilizing multi-sensor features achieved a mean square error of 0.0162.
• The multi-sensor model demonstrated superior prediction accuracy compared to single-sensor models.

Conclusions:

• The developed multi-sensor system effectively monitors laser welding penetration depth in TC4 titanium alloy.
• Multi-sensor data fusion significantly enhances the accuracy of penetration depth prediction.
• The high-precision model provides a foundation for real-time feedback control in laser welding processes.