Modelling of fibre laser cutting via deep learning

Area of Science:

• Materials Science
• Computational Engineering
• Artificial Intelligence

Background:

• Laser cutting is a key industrial and academic materials processing technique.
• Defects like striations can compromise the quality of laser-cut surfaces.
• The non-linear light-matter interactions in laser machining are challenging to model mathematically.

Purpose of the Study:

• To develop novel simulation methods for laser machining.
• To apply deep learning for analyzing laser-cut surfaces.
• To predict laser cutting parameters and visualize cut surface characteristics.

Main Methods:

• Utilized deep learning, specifically neural networks, for image-based analysis.
• Trained a neural network to determine scanning speed directly from microscope images of laser-cut surfaces.
• Developed a predictive visualization tool based on the trained neural network.

Main Results:

• Successfully determined laser cutting scanning speed from surface images using deep learning.
• Demonstrated the capability of a trained neural network to generate realistic predictions of laser cut surface appearance.
• Validated the use of the neural network as a predictive visualization tool for laser machining.

Conclusions:

• Deep learning offers a data-driven approach to modeling complex laser-matter interactions.
• The developed deep learning model can accurately predict scanning speed and visualize cut surface quality.
• This technique provides a novel tool for understanding and optimizing laser cutting processes.