Session: 08-04-02 CFD Development II

Submission Number: 155973

Integration of Cfd and Lstm Methods for Predictive Analysis of Hydrodynamic Forces on a Neutrally Buoyant Autonomous Underwater Vehicle Under Wave Conditions 

This study simulates an Autonomous Underwater Vehicle (AUV) with a neutrally buoyant body using Computational Fluid Dynamics (CFD) to investigate force variations under wave conditions. Different wave conditions are generated using a numerical wave maker, allowing for the analysis of the AUV's response across various wave environments. The overset dynamic mesh technique, which is suitable for Six Degrees of Freedom (6-DOF) motion, accurately captures the AUV's dynamic behavior within a moving flow field. Additionally, the k-ε model simulates turbulence characteristics, enabling an effective description of the hydrodynamic forces acting on the body. Force data for the AUV is calculated under various wave environments through CFD simulations, which is crucial for understanding its dynamic response in complex fluid flows. This study integrates a Long Short-Term Memory (LSTM) deep learning model to enhance predictive capabilities regarding future force conditions, utilizing historical force data extracted from the CFD simulations for analysis. The LSTM model, known for its ability to capture long-term dependencies in time series data, effectively forecasts the force trends experienced by the AUV under different wave scenarios. The research provides key data for optimizing AUV design and performance, demonstrating the potential of combining CFD with deep learning. By predicting future force trends, this method improves AUV stability and efficiency in complex underwater environments and opens new pathways for advancing underwater vehicle design.

Presenting Author: Cheng-Chi Shih National Cheng Kung University

Presenting Author Biography: Mr. Shi is a now master student at National Cheng Kung University (NCKU).