OpenSeeSimE-Fluid: Engineering Simulation Visual Question Answering Benchmark

Dataset Summary

OpenSeeSimE-Fluid is a large-scale benchmark dataset for evaluating vision-language models on computational fluid dynamics (CFD) simulation interpretation tasks. It contains approximately 98,000 question-answer pairs across parametrically-varied fluid simulations including turbulent flow, heat transfer, and complex flow patterns.

Purpose

While vision-language models (VLMs) have shown promise in general visual reasoning, their effectiveness for specialized engineering simulation interpretation remains unknown. This benchmark enables:

• Statistically robust evaluation of VLM performance on CFD visualizations
• Assessment across multiple reasoning capabilities (captioning, reasoning, grounding, relationship understanding)
• Evaluation using different question formats (binary classification, multiple-choice, spatial grounding)

Dataset Composition

Statistics

• Total instances: ~98,000 question-answer pairs
• Simulation types: 5 fluid models (Bent Pipe, Converging Nozzle, Mixing Pipe, Heat Sink, Heat Exchanger)
• Parametric variations: 1,024 unique instances per base model (4^5 parameter combinations)
• Question categories: Captioning, Reasoning, Grounding, Relationship Understanding
• Question types: Binary, Multiple-choice, Spatial grounding
• Media formats: Both static images (1920×1440 PNG) and videos (Originally Extracted at: 200 frames, 40 fps, 5 seconds (some exceptions apply for longer fluid flow development))

Simulation Parameters

Each base model varies across 5 parameters with 4 values each:

Bent Pipe: Bend Angle, Turn Radius, Pipe Diameter, Fluid Viscosity, Fluid Velocity
Converging Nozzle: Pipe Diameter, Front Chamfer Length, Back Chamfer Length, Inner Fillet Radius, Fluid Velocity
Mixing Pipe: Pipe 1 Diameter, Pipe 2 Diameter, Fillet Radius, Fluid 1 Velocity, Fluid 2 Velocity
Heat Sink: Fin Thickness, Sink Length, Fin Spacing, Fin Number, Fluid Velocity
Heat Exchanger: Tube Diameter, Fin Diameter, Fin Thickness, Fin Spacing, Fluid Velocity

Question Distribution

• Binary Classification: 40% (yes/no questions about dead zones, symmetry, flow direction, etc.)
• Multiple-Choice: 30% (4-option questions about flow regime, axis of symmetry, magnitude ranges, etc.)
• Spatial Grounding: 30% (location-based questions with labeled regions A/B/C/D)

Data Collection Process

Simulation Generation

1. Base models sourced from Ansys Fluent tutorial files
2. Parametric automation via PyFluent and PyGeometry interfaces
3. Systematic variation across 5 parameters with 4 linearly-spaced values
4. All simulations solved with validated turbulence models and convergence criteria

Ground Truth Extraction

Automated extraction eliminates human annotation costs and ensures consistency:

• Statistical Analysis: Direct queries on velocity, pressure, and temperature fields
• Distribution Analysis: Dead zone detection via velocity magnitude thresholds (1% of maximum)
• Physics-Based Classification: Mach number calculations for flow regime classification
• Spatial Localization: Color-based region generation with computer vision algorithms

All ground truth derived from numerical simulation results rather than visual interpretation.

Preprocessing and Data Format

Image Processing

• Resolution: 1920×1440 pixels
• Format: PNG with lossless compression
• Standardized viewing orientations: front, back, left, right, top, bottom, isometric
• Consistent color mapping: rainbow gradients (red=maximum, blue=minimum)
• Visualization types: contour plots (pressure, velocity magnitude) and vector plots (velocity vectors, streamlines)

Video Processing

• 200 frames at 40 fps (5 seconds duration); some exceptions apply for longer fluid flow development
• Pathlines showing steady-state flow solution
• H.264 compression at 1920×1440 resolution

Data Fields

{
    'file_name': str,           # Unique identifier
    'source_file': str,         # Base simulation model
    'question': str,            # Question text
    'question_type': str,       # 'Binary', 'Multiple Choice', or 'Spatial'
    'question_id': int,         # Question identifier (1-20)
    'answer': str,              # Ground truth answer
    'answer_choices': List[str], # Available options
    'correct_choice_idx': int,  # Index of correct answer
    'image': Image,             # PIL Image object (1920×1440)
    'video': Video,             # Video frames
    'media_type': str          # 'image' or 'video'
}

Labels

All labels are automatically generated from simulation numerical results:

• Binary questions: "Yes" or "No"
• Multiple-choice: Single letter (A/B/C/D) or descriptive option
• Spatial grounding: Region label (A/B/C/D) corresponding to labeled visualization locations

Dataset Splits

• Test split only: ~98K instances
• No train/validation splits provided (evaluation benchmark, not for model training)
• Representative sampling across all simulation types and question categories

Intended Use

Primary Use Cases

1. Benchmark evaluation of vision-language models on CFD simulation interpretation
2. Capability assessment across visual reasoning dimensions (captioning, spatial grounding, relationship understanding)
3. Transfer learning analysis from general-domain to specialized technical visual reasoning

Out-of-Scope Use

• Real-time engineering decision-making without expert validation
• Safety-critical applications without human oversight
• Generalization to simulation types beyond fluid dynamics

Limitations

Technical Limitations

• Objective tasks only: Excludes subjective engineering judgments requiring domain expertise
• Single physics domain: Fluid dynamics only (see OpenSeeSimE-Structural for structural mechanics)
• Ansys-specific: Visualizations generated using Ansys Fluent rendering conventions
• Steady-state focus: Videos show pathlines of steady-state solutions, not transient phenomena
• 2D visualizations: All inputs are 2D projections of 3D simulations (or 2D Cross Sectional Planes)

Known Biases

• Color scheme dependency: Questions exploit default color gradient conventions
• Geometry bias: Selected simulation types may not represent full diversity of CFD applications
• Flow regime bias: Limited supersonic cases due to parameter range constraints
• View orientation bias: Standardized camera positions may not capture all critical flow features

Ethical Considerations

Responsible Use

• Models evaluated on this benchmark should NOT be deployed for safety-critical engineering decisions without expert validation
• Automated interpretation should augment, not replace, human engineering expertise
• Users should verify that benchmark performance translates to their specific simulation contexts

Data Privacy

• All simulations have no proprietary or confidential engineering data
• No personal information collected
• Publicly available tutorial files used as base models

Environmental Impact

• Dataset generation required significant CPU computational resources with parallel processing
• Consider environmental cost of large-scale model evaluation on this benchmark
• CFD simulations are computationally intensive (hours per case on multi-core workstations)

License

MIT License - Free for academic and commercial use with attribution

Citation

If you use this dataset, please cite:

@article{ezemba2024opensesime,
  title={OpenSeeSimE: A Large-Scale Benchmark to Assess Vision-Language Model Question Answering Capabilities in Engineering Simulations},
  author={Ezemba, Jessica and Pohl, Jason and Tucker, Conrad and McComb, Christopher},
  year={2025}
}

AI Usage Disclosure

Dataset Generation

• Simulation automation: Python scripts with Ansys PyFluent interface
• Ground truth extraction: Automated computational protocols (no AI involvement)
• Quality validation: Expert oversight of automated extraction procedures
• No generative AI used in dataset creation, labeling, or curation

Visualization Generation

• Ansys Fluent rendering engine (deterministic, physics-based)
• Standardized color mapping and camera controls
• No AI-based image generation or enhancement

Contact

Authors: Jessica Ezemba ([email protected]), Jason Pohl, Conrad Tucker, Christopher McComb
Institution: Department of Mechanical Engineering, Carnegie Mellon University

Acknowledgments

• Ansys for providing simulation software and tutorial files
• Carnegie Mellon University for computational resources
• Reviewers and domain experts who validated the automated extraction protocols

---

Version: 1.0
Last Updated: December 2025
Status: Complete and stable