Machine Learning Classification of Self-Organized Surface Structures in Ultrashort-Pulse Laser Processing Based on...

Area of Science:

• Materials Science
• Laser Technology
• Machine Learning

Background:

• Ultrashort-pulsed laser processing enables surface modification but lacks automated quality assurance.
• Machine learning (ML) offers potential for automated routines but is underutilized in this field.
• Current methods require complex parameter studies and lack robust monitoring systems.

Purpose of the Study:

• To develop a machine learning methodology for classifying self-organized surface structures from light microscopic images.
• To enable automated quality assurance in laser-based surface modification.
• To establish a foundation for automated process parameter recommendation and inline control.

Main Methods:

• Fabrication of three types of self-organized surface structures on hot working tool steel and stainless steel using a 300 fs laser.
• Training a classification algorithm using Google's Teachable Machine with optical images of the steel substrates.
• Testing the algorithm's accuracy in distinguishing surface types across different substrates and magnifications.

Main Results:

• The ML classification algorithm achieved very high accuracy in distinguishing surface structure types.
• The trained algorithm demonstrated high accuracy on both the learned hot working steel substrate and the stainless-steel substrate.
• The algorithm maintained high accuracy when classifying images of the same structure type captured at different optical magnifications.

Conclusions:

• The proposed methodology provides a simple, robust, and automated approach for classifying laser-induced surface structures.
• This ML-based classification can serve as a basis for developing advanced quality assurance systems.
• The approach supports future advancements in automated laser process parameter recommendation and inline control.