A Study of Designing Heat Exchanger

This repository supports the workflow:

1. Generate design samples (LHS)
2. Convert design samples to porous parameters (Darcy–Forchheimer: 1/K, C2, etc.)
3. Train GP surrogate models (Q'' and ΔP) and optionally run GA optimization / prediction
4. Validate the Value with CFD Data

I. Step-by-Step

0) Setup (recommended)

From the repository root (HeatExchanger/):

# Go to repo root
cd /path/to/HeatExchanger

# (Recommended) create & activate a venv
python3 -m venv .venv
source .venv/bin/activate

# install dependencies
pip install -U pip
pip install numpy pandas scipy scikit-learn pygad

0.1 (Optional) Conda environment

cd /path/to/HeatExchanger

conda create -n wind-hx python=3.10 -y
conda activate wind-hx

# install dependencies
pip install -U pip
pip install -r requirements.txt

Pick one environment manager: venv or conda (you don't need both).

---

1) Create / Prepare the Sampling Dataset

You need a dataset with (at minimum) these design variables:

• S1_mm
• fin_height_fh_mm
• fin_spacing_fs_mm

There are two options.

1.1 Option A — Download a prepared dataset

Download porousdata.xlsx (or an equivalent prepared dataset) and place it into:

cd /path/to/HeatExchanger
mkdir -p data
# put your downloaded file into HeatExchanger/data/
ls -lh data

Expected examples:

• data/porousdata.xlsx
• data/total_2D_Data.xlsx

1.2 Option B — Generate the dataset locally (Sampling)

Run the sampling script (LHS sampling + porous parameter calculation):

• Script: HeatExchanger/scripts/surrogate/sampling.py
• Output directory: HeatExchanger/data/ (recommended convention)

cd /path/to/HeatExchanger

python3 scripts/surrogate/sampling.py

# Check generated file(s)
ls -lh data | tail -n 20

Expected output example:

• data/LHS_Porous_Params_Result.csv

1.3 (Optional) Select representative points (similarity / downsampling)

If you select representative points after sampling:

• Script: HeatExchanger/scripts/finding_similar.py
• Output directory: HeatExchanger/data/ (recommended)

cd /path/to/HeatExchanger

python3 scripts/finding_similar.py

# Check generated file(s)
ls -lh data | tail -n 20

---

2) Convert Design CSV → Porous CSV (porous_converter.py)

If you already have a design-only CSV (S1/FH/FS) and want to convert it into a porous-parameter CSV:

• Script: HeatExchanger/scripts/porous_calc/porous_converter.py
• Input directory: HeatExchanger/data/
• Output directory: HeatExchanger/data/

cd /path/to/HeatExchanger

# Example:
#   input  : data/LHS_design_samples.csv
#   output : data/porous_from_design.csv
python3 scripts/porous_calc/porous_converter.py   --in data/LHS_design_samples.csv   --out data/porous_from_design.csv   --T 14.8   --v 2.019   --Dc 24.0   --delta_f 0.5   --pitch_ratio 1.0   --N 4   --v_min 0.5   --v_max 3.5   --n_points 50   --check_constraint

# Verify output
ls -lh data/porous_from_design.csv

Column name requirements (input CSV)
The input CSV must contain these columns (case-insensitive matching is supported):

• S1: S1_mm / S1 / s1_mm / s1
• fin height: fin_height_fh_mm / fh_mm / fin_height / fh / hf_mm / hf
• fin spacing: fin_spacing_fs_mm / fs_mm / fin_spacing / Fs_mm / Fs / fs

---

3) Train GP Surrogate Models (Q'' and ΔP)

This stage trains Gaussian Process models for:

• Q'' (heat flux, W/m²)
• ΔP (pressure drop, Pa)

Your surrogate training script(s) may differ by filename. Use the commands below as a template.

3.1 Train surrogate

cd /path/to/HeatExchanger

# Example: run a surrogate training script
python3 scripts/surrogate/train_surrogate.py --data data/total_2D_Data.xlsx

If your script is a plain Python file without CLI args, just run it directly:

cd /path/to/HeatExchanger
python3 scripts/surrogate/train_surrogate.py

3.2 Run GA optimization (optional)

cd /path/to/HeatExchanger

python3 scripts/surrogate/train_surrogate.py --mode ga --data data/total_2D_Data.xlsx

3.3 Predict using a user-defined design (optional)

cd /path/to/HeatExchanger

python3 scripts/surrogate/train_surrogate.py --mode predict   --data data/total_2D_Data.xlsx   --s1 181.0394 --fh 28.9923 --fs 2.6038

---

4) Validate the value with CFD data

Compare:

• Surrogate-predicted Q'' and ΔP vs.
• CFD-evaluated Q'' and ΔP

4.1 Prepare CFD validation dataset

Put your CFD results into HeatExchanger/data/ as a CSV or Excel file.

Recommended columns

• Inputs:
  • S1_mm, fin_height_fh_mm, fin_spacing_fs_mm
• CFD outputs (suggested names):
  • Q_CFD (or Qpp_CFD)
  • dP_CFD (or DeltaP_CFD)

cd /path/to/HeatExchanger
ls -lh data | grep -i cfd

4.2 Run validation script

cd /path/to/HeatExchanger

python3 scripts/validation/validate_with_cfd.py   --train data/total_2D_Data.xlsx   --cfd   data/CFD_validation.csv   --out   data/validation_report.csv

Expected outputs:

• data/validation_report.csv (per-sample error summary)
• (Optional) parity plots / error histograms (if implemented)

4.3 What to report

Common validation metrics:

• R² (CFD vs surrogate)
• MAE / RMSE for Q'' and ΔP
• Parity plot (y = x) and error distribution
• Worst-case error samples (top-k)

---

II. Dataset

A) Prebuilt dataset (download)

Place into:

• HeatExchanger/data/

Dataset	Purpose	Target path (repo)	File name	Download link	Notes
porousdata	Prebuilt porous parameters dataset (1/K, C2, etc.)	data/	porousdata.xlsx	Download	Put the file exactly at HeatExchanger/data/porousdata.xlsx
total_2D_Data	CFD/2D training dataset for surrogate (Q'', ΔP)	data/	total_2D_Data.xlsx	Download	Used by GP surrogate training + GA optimization

Examples:

• data/porousdata.xlsx
• data/total_2D_Data.xlsx

B) Locally generated dataset

Generated by:

• scripts/surrogate/sampling.py

Output example:

• data/LHS_Porous_Params_Result.csv

---