Session: 1.2 - Machine Learning, Reduced Order Modeling in CFD and Design Optimization

Paper Number: 169365

169365 - Cfd-Based Optimization of Autonomous Underwater Vehicles Using Neural Networks 

Abstract: 

This research focuses on improving the design of underwater vehicle hulls to reduce drag and boost energy efficiency. Using Computational Fluid Dynamics (CFD), the study analyzes the hydrodynamic performance of the DARPA SUBOFF hull, validating CFD results with experimental data across various velocities. Once validated, the research explores how slight changes in hull shape affect drag at different flow speeds and orientations.

To make the design process faster and more efficient, machine learning is incorporated. A neural network is trained using CFD data that will be generated through simulation software, and this will help in predicting the best hull shape for minimizing drag, reducing the need for time-consuming simulations. In addition, an AI model is developed to estimate key hydrodynamic parameters based only on velocity and length, further streamlining the optimization process.

By leveraging CFD and AI, the study presents a new approach for efficiently designing underwater vehicles that minimize drag and energy consumption. The research focuses on improving the accuracy of CFD simulations and applying machine learning techniques to optimize hull shapes for performance across various operating conditions. This combination of methods allows for more efficient hull designs that could lead to reduced fuel consumption and enhanced performance in real-world applications.

The findings also highlight how AI can predict complex hydrodynamic parameters with minimal input, further simplifying the design process. This approach offers a more streamlined and data-driven path for developing energy-efficient underwater vehicles, potentially accelerating innovation in marine vehicle design and supporting advancements in autonomous underwater technology.

Presenting Author: Albara Salem Istanbul Technical University

Presenting Author Biography: