🔧 Sensor Fault Detection System

An end-to-end Machine Learning–based Sensor Fault Detection System designed to automatically identify faulty and non-faulty sensor readings across multiple sensor types using a unified architecture.
The system supports Wafer, Gas, Temperature, Light (LDR), and Soil Moisture sensors and provides real-time predictions through a Streamlit web interface.

🔗 Live Demo: https://sensor-fault-detection-system.streamlit.app/

---

📌 Project Highlights

• ✅ Multi-sensor fault detection using Machine Learning
• ✅ All-in-One intelligent router for automatic sensor type detection
• ✅ Robust preprocessing with imputation, scaling, and class balancing
• ✅ Separate optimized models for each sensor type
• ✅ Interactive Streamlit-based web interface
• ✅ Production-ready modular project structure

---

🧠 System Architecture

1. Input Layer

   • CSV file upload or manual sensor input

2. All-in-One Router

   • Automatically detects sensor type based on schema

3. Preprocessing Pipeline

   • Missing value handling (KNN / Simple Imputer)
   • Feature scaling (RobustScaler / StandardScaler)
   • Class imbalance correction (SMOTETomek)
   • Feature engineering

4. ML Model Layer

   • Wafer → Support Vector Classifier (SVC)
   • Gas → Random Forest
   • Temperature → Random Forest
   • Light (LDR) → XGBoost
   • Soil Moisture → XGBoost

5. Prediction & Visualization

   • Fault classification
   • Downloadable results
   • Visual insights

---

📂 Supported Sensors

Sensor Type	Fault Labels
Wafer	0 = Normal, 1 = Faulty
Gas	0 = Normal, 1 = Faulty
Temperature	0 = Faulty, 1 = Normal
Light (LDR)	-1 = Faulty, 1 = Normal
Soil Moisture	-1 = Faulty, 1 = Normal

---

🛠️ Technologies Used

• Programming Language: Python
• Machine Learning: Scikit-learn, XGBoost
• Data Processing: NumPy, Pandas
• Imbalance Handling: SMOTETomek
• Model Persistence: Pickle (.pkl)
• Web Framework: Streamlit
• Version Control: Git & GitHub

---

📁 Project Structure

Sensor-Fault-Detection/
│
├── data/                   # Sensor datasets
├── models/                 # Trained models (.pkl)
├── pipelines/              # Preprocessing pipelines
├── scripts/                # Training & utility scripts
├── custom_transformers.py  # Feature engineering logic
├── model_utils.py          # Model load/save utilities
├── all_in_one_router.py    # Sensor auto-detection logic
├── streamlit_app.py        # Web application
├── main.py                 # Training & prediction entry
├── requirements.txt        # Dependencies
└── README.md               # Project documentation

---

🚀 How to Run Locally

1️⃣ Clone the Repository

git clone https://github.com/debjithacks/sensor-fault-detection.git
cd sensor-fault-detection

2️⃣ Install Dependencies

pip install -r requirements.txt

3️⃣ Run the Application

streamlit run streamlit_app.py

---

📊 Model Evaluation

The models are evaluated using:

• Accuracy
• Precision
• Recall
• F1-Score
• ROC-AUC Curve
• Confusion Matrix

All models achieved strong performance with ROC-AUC scores above 0.85, confirming reliability for real-world usage.

---

👥 Team Members

This project was developed as a team project by:

• Aditya Maity
• Souvik Pramanik
• Avijit Ray
• Subhadeep Hatai
• Debjit Ghosh

---

📌 Use Cases

• Industrial sensor health monitoring
• Smart agriculture systems
• IoT-based automation platforms
• Preventive maintenance systems

---

🔮 Future Enhancements

• Real-time sensor streaming (IoT integration)
• Deep learning-based fault detection
• Sensor fusion and anomaly explanation
• Cloud-native deployment

---

📜 License

This project is released under the MIT License – free to use, modify, and distribute with attribution.

---

⭐ If you find this project useful, please consider giving it a star!