Constrained Bayesian Optimization for Automatic Underwater Vehicle Hull Design

Overview

Developed an integrated CAD-CFD-AI toolchain for automated UUV hull optimization. The pipeline uses constrained Bayesian optimization with FreeCAD for parametric design and OpenFOAM for CFD simulation, respecting physical manufacturing constraints throughout the optimization.

Key Results

• Drag force optimized from 67N to 36N at 1.1 m/s (46% reduction)
• Bayesian optimization demonstrated as the most sample-efficient method for UUV design
• Constraints on manufacturing feasibility integrated directly into the optimization loop

Pipeline

1. FreeCAD — Parametric hull geometry generation
2. OpenFOAM — CFD simulation for drag evaluation
3. Constrained BO — Intelligent selection of next design candidates while respecting physical constraints