A comprehensive CFD lifecycle dataset for marine vessel hydrodynamics.

Area of Science:

• Naval Architecture and Ocean Engineering
• Computational Fluid Dynamics (CFD)
• Artificial Intelligence (AI) in Engineering

Background:

• Industrial adoption of AI in computational simulation is hindered by a lack of comprehensive, real-world datasets.
• Existing CFD datasets often cover only specific simulation phases, limiting AI model development for the full workflow.

Purpose of the Study:

• To present novel, high-quality datasets spanning the complete CFD lifecycle from industrial workflows.
• To facilitate the development of AI models capable of learning across different simulation phases.
• To support research in CFD solver performance, data management, and flow visualization.

Main Methods:

• Collected data from operational industrial CFD workflows at the China Ship Scientific Research Center.
• Included data from three marine vessels (two tankers, one submarine) with complex stern flows.
• Structured datasets to encompass pre-processing (geometries), solving (linear systems), and post-processing (flow solutions, meshes).

Main Results:

• Developed three unique datasets covering the entire CFD simulation pipeline.
• The datasets capture complex separated stern flows characteristic of low-speed, high-block-coefficient hull forms.
• The data resource integrates interdependent stages, enabling holistic AI model training and analysis.

Conclusions:

• The presented datasets address the critical need for comprehensive CFD data to advance AI in engineering simulations.
• This resource enables AI model development across the full CFD lifecycle, improving robustness and applicability.
• The datasets also serve as valuable benchmarks for solver evaluation, data management research, and flow visualization techniques.