A Novel Method for Hydrogen Permeation Detection Using Reflective Microscopy and Machine Learning.

Area of Science:

• Materials Science
• Analytical Chemistry
• Surface Science

Background:

• Localized detection of hydrogen permeation in steel membranes is critical for industrial applications.
• Current methods for analyzing hydrogen permeation are often limited in scope or accessibility.
• Understanding hydrogen's interaction with steel is vital for preventing material degradation.

Purpose of the Study:

• To develop and validate a novel framework for the localized detection and quantification of hydrogen permeation in steel membranes.
• To enhance the sensitivity of optical methods for detecting subtle changes caused by hydrogen.
• To provide an accessible and effective alternative to existing hydrogen analysis techniques.

Main Methods:

• Utilized reflective microscopy (RM) to capture changes in steel surfaces.
• Applied machine learning (ML)-driven image analysis, including UMAP and K-means clustering, for contrast enhancement and feature detection.
• Employed time-series analysis to amplify subtle optical signals related to hydrogen presence.
• Validated findings using Scanning Kelvin Probe (SKP) measurements.

Main Results:

• Successfully detected subtle changes in reflected light intensity caused by altered Fe(II)/Fe(III) ratios due to hydrogen permeation.
• The RM-ML framework effectively enhanced contrast, making near-optical-limit signals detectable.
• Identified specific hydrogen-affected zones in steel membranes with high accuracy.
• Corroborated RM-ML findings with independent SKP measurements.

Conclusions:

• Established a novel Reflective Microscopy-Machine Learning (RM-ML) framework for localized hydrogen detection in metallic membranes.
• Demonstrated the potential of this accessible technique as a powerful tool for analyzing hydrogen permeation.
• This approach offers a promising alternative for quantitative, localized analysis of hydrogen in materials.