Porous Media Sampling for Fin-Tube Heat Exchanger Design

This repository contains a complete workflow for generating and analyzing porous media parameters for fin-tube heat exchanger designs using Latin Hypercube Sampling (LHS) and CFD-based correlations.

📁 Dataset

The full dataset and additional resources are available on Google Drive: Dataset & Resources

📋 Overview

This project implements a systematic approach to:

1. Generate constrained design variable samples using LHS
2. Calculate porous media parameters from fin-tube geometries
3. Project samples onto existing datasets for diversity
4. Analyze sample spread and distribution quality
5. Visualize results in 3D parameter space

🔧 Code Files

1. 1. generate_lhs_100k.py

Generates constrained Latin Hypercube Samples for fin-tube design variables.

Features:

• Generates samples for S1=S2 (tube spacing), fin height, and fin spacing
• Enforces geometric feasibility constraints
• Uses tube diameter td=24mm with clearance margins
• Outputs: 1.constrained_LHS_100k.csv

Usage:

python "1. generate_lhs_100k.py" --n 100000 --seed 2025 --out constrained_LHS_100k.csv

Constraints:

• S1=S2: 45-200 mm
• Fin spacing (fs): 2-8 mm
• Fin height (fh): 6 mm ≤ fh ≤ 0.5*(s/√2 - td) - 0.4 mm

---

2. 2. porous_from_design_integrated.py

Batch processor that converts design variables to porous media parameters.

Features:

• Uses Nir(1991) pressure-drop correlation for staggered tube banks
• Fits Darcy-Forchheimer model via least-squares over velocity range
• Calculates viscous and inertial resistance coefficients
• Outputs: 2.porous_params_100k_full.csv

Physics:

• Computes porosity, hydraulic diameter, area ratios
• Evaluates pressure drop using validated correlations
• Extracts inverse permeability (1/K) and inertial coefficient (C2)

Usage:

python "2. porous_from_design_integrated.py" constrained_LHS_100k.csv --out porous_params_100k_full.csv

Key Parameters Calculated:

• Viscous_Resistance_1_m2 (1/K): Inverse permeability
• Inertial_Resistance_1_m (C2): Inertial resistance coefficient
• Porosity (ε): Volume fraction available for flow

---

3. 3. lhs_project_to_dataset.py

Projects LHS samples onto existing dataset to maximize diversity using optimal assignment.

Features:

• Log-transforms and normalizes porous parameters
• Uses KD-tree for efficient nearest-neighbor search
• Applies Hungarian algorithm for optimal one-to-one matching
• Minimizes total distance in parameter space
• Outputs: 3.porous_LHS_100_projected.csv

Usage:

python "3. lhs_project_to_dataset.py" \
    --lhs_csv porous_LHS_100.csv \
    --data_csv porous_params_100k_full.csv \
    --out porous_LHS_100_projected.csv

---

4. 4.spread_and_index_report.py

Analyzes distribution quality and provides sample selection diagnostics.

Features:

• Computes minimum nearest-neighbor distances
• Calculates spread metrics (mean, median, std)
• Identifies poorly covered regions
• Exports selected samples with original row indices
• Outputs: 4.selected_with_row_indices.csv + text report

Usage:

python "4.spread_and_index_report.py" \
    --csv porous_LHS_100_projected.csv \
    --out selected_with_row_indices.csv

Metrics Reported:

• Average/median minimum distance to neighbors
• Coefficient of variation
• Worst-case coverage gaps
• Quantile analysis of spread distribution

---

5. plot_porous_3d_and_hist.py

Visualization tool for parameter space exploration.

Features:

• 3D scatter plot in (Porosity, 1/K, C2) space
• Individual histograms for each parameter
• Supports log-scale transformation
• Configurable bins and output format

Usage:

python plot_porous_3d_and_hist.py

Outputs:

• porous_plots/porous_3d_scatter.png
• porous_plots/porous_histograms.png

---

🔄 Workflow

Step 1: Generate LHS samples
    └─> 1. generate_lhs_100k.py
         Output: constrained_LHS_100k.csv

Step 2: Calculate porous parameters
    └─> 2. porous_from_design_integrated.py
         Output: porous_params_100k_full.csv

Step 3: Project to optimal samples
    └─> 3. lhs_project_to_dataset.py
         Output: porous_LHS_100_projected.csv

Step 4: Analyze sample quality
    └─> 4.spread_and_index_report.py
         Output: selected_with_row_indices.csv + report

Step 5: Visualize results
    └─> plot_porous_3d_and_hist.py
         Output: PNG plots

📦 Dependencies

numpy
pandas
scipy
matplotlib

Install with:

pip install numpy pandas scipy matplotlib

📊 Data Files

• 1.constrained_LHS_100k.csv: Raw LHS samples in design space (S1, fh, fs)
• 2.porous_params_100k_full.csv: Full porous parameter dataset
• 3.porous_LHS_100_projected.csv: Projected optimal samples
• 4.selected_with_row_indices.csv: Final selection with metadata
• 4.porous_LHS_100_projected.csv: Alternative projected samples

Note: Full datasets are available on the Google Drive.

🔬 Scientific Background

This work applies optimal experimental design principles to CFD-based porous media modeling:

• Latin Hypercube Sampling: Ensures uniform coverage in design variable space
• Nir(1991) Correlation: Physics-based pressure drop for staggered fin-tube arrays
• Darcy-Forchheimer Model: Industry-standard porous media representation
• Optimal Assignment: Maximizes diversity in parameter space via Hungarian algorithm

📝 Reference

Nir, A. (1991). Heat Transfer and Friction Factor Correlations for Crossflow over Staggered Finned Tube Banks. Heat Transfer Engineering, 12(1), 43–58.

👥 Authors

AI-CO-SCIENTIST Project

📄 License

For research and educational purposes.

---

For questions or collaborations, please refer to the dataset folder for additional documentation.