Reconstructing the pressure field around swimming fish using a physics-informed neural network.

Area of Science:

• Fluid Dynamics
• Biophysics
• Computational Biology

Background:

• Fish utilize pressure signals detected by their lateral line system for sensing predators, flow, and social interactions.
• Lateral line ablation is a common but invasive technique to study pressure sensing roles.
• A non-invasive method is needed to accurately measure pressure fields experienced by fish.

Purpose of the Study:

• To propose and validate a non-invasive method for reconstructing the instantaneous pressure field sensed by a fish's lateral line system.
• To utilize two-dimensional particle image velocimetry (PIV) and physics-informed neural networks (PINN) for pressure field reconstruction.
• To assess the method's accuracy and robustness using simulations and experimental data.

Main Methods:

• Developed a physics-informed neural network (PINN) model.
• Integrated PINN with two-dimensional particle image velocimetry (PIV) measurements.
• Satisfied Navier-Stokes equations and PIV data constraints for pressure field optimization.
• Validated the method with direct numerical simulation of a swimming mackerel (Scomber scombrus).
• Applied the method to experimental data of a turning zebrafish (Danio rerio).

Main Results:

• The proposed method accurately reconstructs the pressure field on the fish's body.
• The technique demonstrated robustness, being relatively insensitive to PIV measurement resolution.
• Successful validation was achieved through both numerical simulations and experimental data analysis.
• The method provides a detailed pressure map experienced by the fish.

Conclusions:

• A novel, non-invasive method using PINN and PIV can reconstruct fish pressure sensing fields.
• This approach offers a valuable tool for studying fish sensory biology without invasive procedures.
• The method's accuracy and insensitivity to PIV resolution enhance its applicability in aquatic biomechanics research.