Autonomous analysis to identify bijels from two-dimensional images.

Area of Science:

• Materials Science
• Chemical Engineering
• Data Science

Background:

• Bicontinuous interfacially jammed emulsion gels (bijels) are advanced composite materials.
• Manufacturing bijels presents significant challenges, hindering scalable production.

Purpose of the Study:

• To develop an automated machine learning tool for classifying bijel fabrication attempts.
• To improve the efficiency and reliability of bijel production through data-driven insights.

Main Methods:

• Confocal microscopy images of successful and unsuccessful bijel fabrications were used for training and testing.
• Image processing techniques, including autocorrelation functions and structure factors, were employed.
• Supervised machine learning models and decision trees were utilized for image classification.
• Adaptive design was used to optimize image pre-processing steps.

Main Results:

• Two distinct machine learning approaches achieved classification accuracies of 85.4% and 87.5%.
• Optimal performance was achieved by integrating data from both liquid and particle channels.
• A characteristic length scale was identified as a crucial factor for successful bijel classification.
• Direct classification of liquid domain shapes using shape descriptors also proved effective.

Conclusions:

• Machine learning offers a viable pathway for automating the classification of bijel fabrication.
• The characteristic length scale of bijel samples is a key determinant of successful fabrication.
• Further development of these methods could lead to automated bijel production.