Deep Learning to Analyze Sliding Drops

Area of Science:

• Materials Science
• Surface Science
• Fluid Dynamics

Background:

• Contact angle measurements are crucial for understanding surface properties.
• Analyzing sessile drops is straightforward due to symmetry and high resolution.
• Sliding drop analysis is challenging due to low resolution and asymmetry.

Purpose of the Study:

• To enhance the accuracy of contact angle measurements for sliding drops.
• To develop an automated method for analyzing asymmetric, low-resolution drop images.
• To improve dynamic surface tension analysis in fluid systems.

Main Methods:

• Developed a deep learning-based super-resolution model with an upscale ratio of 3.
• Implemented a 4-segment super-resolution optimized-fitting (4S-SROF) method.
• Conducted systematic experiments with synthetic images to optimize polynomial fitting parameters.

Main Results:

• The super-resolution model achieved an accurate image enlargement of 9 times (PSNR = 36.39).
• The 4S-SROF method improved contact angle accuracy by 21% for angles < 90°.
• The method improved accuracy by 33% for contact angles > 90°.

Conclusions:

• The 4S-SROF method significantly enhances the accuracy of dynamic contact angle measurements.
• This approach overcomes limitations of low resolution and asymmetry in sliding drop analysis.
• The developed technique offers a more reliable tool for surface characterization and fluid behavior studies.