Virtual draw of microstructured optical fiber based on physics-informed neural networks

Area of Science:

• Materials Science and Engineering
• Computational Physics
• Optical Fiber Technology

Background:

• Fabricating microstructured optical fibers (MOFs) with advanced microstructures and high performance is hindered by complex manufacturing processes.
• Existing simulation methods struggle to accurately model the intricate physics governing fiber drawing, particularly free boundary conditions.

Purpose of the Study:

• To introduce a novel computational approach using physics-informed neural networks (PINNs) for simulating the microstructured optical fiber drawing process.
• To develop a virtual fabrication model that accurately captures the evolution of fiber dimensions and internal fluid dynamics.

Main Methods:

• Integrated complex partial differential equations and boundary conditions of the fiber drawing process into a neural network's loss function.
• Developed a secondary neural network to specifically address and solve the free boundary conditions related to the fiber's inner and outer diameters.
• Utilized PINNs to model molten glass flow, predicting velocity and pressure distributions.

Main Results:

• The PINN model successfully predicted the evolution of the fiber's inner and outer diameters during the drawing process.
• The model provided insights into the velocity and pressure distributions within the molten glass, crucial for analyzing capillary collapse.
• Simulation results demonstrated good agreement with experimental observations regarding the influence of temperature, feed speed, and draw speed.

Conclusions:

• Physics-informed neural networks offer a powerful and novel method for simulating the complex physics of microstructured optical fiber drawing.
• This virtual fabrication approach validates the feasibility of using PINNs for MOF design and optimization.
• The methodology holds significant promise for advancing the practical applications and manufacturing of microstructured optical fibers.