⚡ Echo-Power: AI-Assisted Voltage Abnormality Detection for Medical-Grade Transformers

🌍 Project Summary

Echo-Power is an open-source AI-powered detection system for identifying abnormal voltage conditions in miniature constant-voltage transformers used in medical devices, especially for low-resource environments.

Developed with real-time 1D-CNN model inference optimized for Raspberry Pi CM4, STM32, and ESP32.

📁 Directory Structure

echo-power/
├── echo_power/                 # Python package
│   ├── detector.py             # TFLite model loader and inference
│   ├── preprocessor.py         # Signal normalization
│   ├── relay_control.py        # GPIO relay activation
│   ├── config.yaml             # Threshold configuration
│   └── example_run.py          # End-to-end inference demo
├── hardware/
│   ├── Echo-Power_Hardware_Design_Guide.docx     # Full schematic
│   ├── echo_power_hardware_schematic.png         # PCB layout
│   └── Echo-Power_BOM_List.xlsx                  # Bill of materials
├── LICENSE
└── README.md

🚀 How to Use

1. Clone this repository:

git clone https://github.com/ykkims99/echo-power.git
cd echo-power

2. Prepare Raspberry Pi GPIO and copy your TFLite model to model/echo_power.tflite.

3. Run demo:

python3 echo_power/example_run.py

📦 Model Details

• Input: Normalized voltage & current sequences (1000 samples)
• Output: Anomaly classification probabilities (e.g., normal, surge, undervoltage)
• Size: < 128KB TFLite
• Latency: ~20ms on CM4

📜 License

This project is licensed under the MIT License.

✉️ Citation

Kim, Y.K. (2025). Echo-Power: AI-Assisted Voltage Abnormality Detection for Medical Transformers. In submission to Scientific African.

---

Maintained with ❤️ by @ykkims99