Unraveling Oil-Microbubble Interactions via a Ternary Phase-Field Model and Machine Learning for Enhanced Surface...

Area of Science:

• Fluid dynamics
• Surface science
• Computational modeling

Background:

• Oily surface cleaning often involves microbubble (MB) interactions with oil droplets adhered to surfaces.
• Understanding these dynamics is crucial for optimizing cleaning efficiency.
• Existing models struggle with the complex, nonlinear interactions involved.

Purpose of the Study:

• To develop and validate a predictive model for microbubble-oil droplet collision dynamics.
• To assess the effectiveness of traditional methods versus machine learning approaches for classifying interaction outcomes.
• To propose a practical model for engineering applications in oily surface cleaning.

Main Methods:

• A high-resolution numerical model using a ternary phase-field approach was employed.
• A computational fluid dynamics database of 182 samples was generated, varying parameters like contact angle, interfacial tensions, and droplet/bubble sizes.
• Support Vector Machine (SVM) and hybrid SVM-spreading coefficient models were developed and tested.

Main Results:

• The traditional spreading coefficient method showed limitations in accurately classifying MB-oil droplet detachment.
• SVM-based classification significantly improved prediction accuracy for MB-oil interactions.
• A hybrid model integrating the spreading coefficient and SVM achieved 100% accuracy across various capillary numbers.

Conclusions:

• Machine learning, particularly SVM, offers superior accuracy in predicting MB-oil droplet interactions compared to traditional methods.
• A hybrid SVM approach provides a robust and highly accurate method for oily surface cleaning applications.
• A simplified correlation model is proposed for practical engineering use.