Industry collaboration with Borçelik — an AI-enhanced digital twin for HVAC-type fan-coil cooling systems. Combines ML-based temperature prediction, PID thermal control, and real-time Streamlit visualisation to demonstrate energy-efficient cooling optimisation.
| Feature | Description |
|---|---|
| CAD visualisation | Live 2D render of coil + 8-fan layout, colour-mapped to temperature |
| Digital twin | Animated simulation of coil temperature and fan response over time |
| Model validation | Upload real measurement CSV → computes MSE against the simulated model |
| Parameter control | Sidebar sliders for temperature range, fan speed, and simulation settings |
| Export | Animated output via imageio (GIF/video) |
| Library | Role |
|---|---|
streamlit |
Web dashboard and interactive controls |
numpy |
Numerical simulation and data generation |
scipy.optimize |
Parameter fitting and optimisation |
scikit-learn |
MSE calculation for model validation |
matplotlib |
CAD-style coil/fan visualisation and charts |
imageio |
Animation export |
# 1. Clone
git clone https://github.com/salamon30/cool-cooling.git
cd cool-cooling
# 2. Create virtual environment
python3 -m venv .venv
source .venv/bin/activate # Windows: .venv\Scripts\activate
# 3. Install dependencies
pip install streamlit numpy scipy scikit-learn matplotlib imageio pandas
# 4. Run
streamlit run bobin_arayuz.py
# → http://localhost:8501Real Measurement CSV
↓
Date & Sensor Parsing → MSE vs. Simulated Model
↓
Digital Twin Simulation
↓
8-Fan CAD Animation → Temperature colour map (coolwarm)
The dashboard has two modes:
- Model Validation (sidebar) — upload a CSV with timestamped temperature readings; the app fits the simulation model and reports MSE.
- Digital Twin Simulation — run the animated coil + fan system forward in time with adjustable parameters.
Developed in collaboration with Borçelik (steel manufacturing, part of Arcelor Mittal group) as a graduation project at Kadir Has University. The goal: apply digital twin methodology to an industrial fan-coil cooling system to enable predictive maintenance and energy efficiency analysis.
Key engineering contributions:
- ML temperature prediction model trained on real operating data from Borçelik's production environment
- PID-based thermal control loop implemented in Python/MATLAB/Simulink
- SciPy parameter fitting to align the simulated model with measured sensor readings
- Real-time dashboard enabling plant engineers to validate model accuracy via CSV upload
Recep Uzun — AI Master's Student @ Deggendorf Institute of Technology
