Fluid Properties Extraction in Confined Nanochannels with Molecular Dynamics and Symbolic Regression Methods

Area of Science:

• Fluid dynamics
• Computational physics
• Machine learning

Background:

• Calculating fluid transport coefficients and slip length in nano-conduits is computationally intensive.
• These properties depend on complex dynamic, thermal, and geometrical factors.
• Existing methods often require extensive simulations.

Purpose of the Study:

• To propose an alternative, computationally efficient method for determining fluid transport coefficients and slip length.
• To derive interpretable, data-driven mathematical expressions using genetic programming.
• To ensure derived equations adhere to physical principles and existing domain knowledge.

Main Methods:

• Utilized genetic programming-based symbolic regression.
• Applied the method to molecular dynamics simulation data.
• Focused on deriving physically interpretable mathematical expressions.

Main Results:

• Developed a set of accurate and less complex mathematical equations.
• The derived equations effectively represent fluid transport properties.
• Demonstrated the potential for bypassing time-consuming simulations.

Conclusions:

• Symbolic regression offers a powerful alternative for calculating fluid properties.
• The derived equations facilitate efficient fluid property interpolation and extrapolation.
• This approach enhances the understanding and application of fluid behavior in nano-conduits.