README.md

SAWP-20 Benchmark Dataset

This dataset contains 20 structural analysis problems with their corresponding ground truth schematics, simplified problem descriptions, and reference code solutions.

Dataset Structure

• ground_truth/: Contains 20 PNG images (schematic_01.png to schematic_20.png) showing the ground truth structural schematics
• problem_descriptions/: Contains 20 text files (problem_1.txt to problem_20.txt) with simplified problem descriptions containing only the problem context
• code/: Contains 20 Python files (1.py to 20.py) with ground truth reference solutions implemented using OpenSeesPy

Problem Description Format

Each problem description file contains only the essential problem context, including:

• Structural geometry and dimensions
• Material properties (Young's modulus, cross-sectional areas, moments of inertia)
• Loading conditions
• Support conditions
• Expected analysis outputs

Ground Truth Code Solutions

The code/ directory contains reference implementations for all 20 problems. These solutions are written in Python using:

• OpenSeesPy: For structural analysis and finite element modeling
• opsvis: For visualization of structural models, loads, deformations, and internal forces
• matplotlib: For plotting and visualization

Each code file follows a consistent structure:

1. Model initialization and parameter configuration
2. Node and element definition
3. Boundary conditions (supports)
4. Load application (point loads and/or distributed loads)
5. Analysis setup and execution
6. Visualization of results (model, loads, deformations, and internal force diagrams)

These reference solutions represent the expected output that code generation frameworks should produce for each problem.

Usage

This dataset is designed for benchmarking structural analysis code generation systems. Each problem provides:

• Clear structural problem definition
• All necessary parameters for analysis
• Ground truth reference code solution
• Expected outputs (deformations and internal forces)

File Naming Convention

• Ground truth images: schematic_XX.png where XX is the problem number (01-20)
• Problem descriptions: problem_X.txt where X is the problem number (1-20)
• Ground truth code: X.py where X is the problem number (1-20)