Insights into Flexible Bioinspired Fins for Unmanned Underwater Vehicle Systems through Deep Learning

Area of Science:

• Robotics and Mechanical Engineering
• Biomimetic Design
• Fluid Dynamics

Background:

• Bio-inspired unmanned underwater vehicles (UUVs) offer enhanced maneuverability over traditional designs.
• Existing research utilizes neural networks to predict UUV thrust and power based on design and fin kinematics.
• Material properties of fins are critical for optimizing performance but are not fully integrated into predictive models.

Purpose of the Study:

• To develop novel forward neural network models that incorporate fin material stiffness.
• To predict kinematic performance, thrust, and power across a spectrum of fin gaits and materials.
• To enhance the selection of materials and kinematics for improved UUV efficiency and range.

Main Methods:

• Created new forward neural network models to simulate fin material stiffness effects.
• Validated models using holdout data to predict thrust and power for unseen gaits.
• Utilized a non-dimensional figure of merit (FOM) to evaluate propulsive efficiency and compare designs.

Main Results:

• Developed models accurately predict thrust and power with high resolution for various fin parameters.
• Demonstrated significant improvements: 83.89% increase in thrust and 137.58% increase in efficiency.
• Identified optimal fin stiffness and kinematics for maximizing thrust and FOM.

Conclusions:

• Neural network models effectively capture the relationship between fin material, kinematics, and UUV performance.
• Optimized material selection and kinematic design can drastically enhance UUV thrust and efficiency.
• This research provides a pathway for improved bio-inspired fin design for extended underwater operations.