Skip to content

DTMB 5415 Hull Dataset

Overview

Benchmark dataset for the evaluation and comparison of design-space dimensionality reduction methods in hydrodynamic shape optimization.

The dataset is based on the DTMB 5415 hull and provides a physics-informed representation combining geometry, design variables, and hydrodynamic quantities.

DOI: https://zenodo.org/records/19222008

Dataset Description

  • Total configurations: 9,001
  • Baseline configuration: 1 (DTMB 5415 reference hull)
  • Sampled configurations: 9,000
  • Invalid simulations: 0

Configurations are generated through Monte Carlo sampling of a nonlinear parametric model based on global 3D shape modifications.

Each configuration is described by 27 normalized design variables in the range [-1, 1].

Data Structure

Each configuration includes:

Geometry (D)

  • Hull surface discretized with 2,250 points
  • Flattened coordinate vectors:
  • x, y, z

Design Variables (U)

  • 27 parameters defining geometric perturbations
  • Normalized in [-1, 1]

Physics Fields

  • Pressure distribution on hull (2,250 points)
  • Wave elevation field (1,500-point structured grid)

Integrated Quantities

  • Wave resistance coefficient (Cw)
  • RMS vertical acceleration at the bridge (az)
  • RMS pitch motion (ξ₅)

Simulation Setup

Hydrodynamic quantities are computed using low-fidelity solvers:

  • WARP (linear potential flow) → calm-water resistance
  • SMP (linear strip theory) → seakeeping

Conditions

  • Head waves
  • Sea state 5
  • Bretschneider spectrum
  • Froude numbers: 0.25, 0.33, 0.41

Scale Consistency

The dataset combines model-scale and full-scale representations:

  • Geometry provided at full scale (Lpp = 142 m)
  • Seakeeping quantities evaluated at full scale
  • Calm-water quantities computed at model scale (Lpp = 5.720 m)

Data Quality

  • All configurations correspond to physically valid simulations
  • Invalid samples were removed during dataset generation
  • Final dataset contains only consistent hydrodynamic responses

Use Cases

  • Dimensionality reduction (PME, PI-PME, PD-PME)
  • Multi-physics embedding
  • Surrogate-based optimization
  • Benchmarking DR methods in naval hydrodynamics

Scientific Context

The dataset follows a physics-informed representation of the design space, where:

  • geometry
  • design parameters
  • physical responses

are consistently combined within a unified framework.