Predictive Maintenance using IoT

Overview

This project demonstrates an IoT-based Predictive Maintenance System. It uses real-time sensor data, machine learning models, and Firebase integration to monitor equipment health, predict failures, and determine reasons for malfunctions. The system includes visualization using Streamlit and Matplotlib, real-time data fetching from Firebase, and automated machine health and fault prediction.

---

Features

• Real-time Data Monitoring: Fetch sensor data from Firebase and visualize it dynamically.
• Machine Health Prediction: Predict the condition of a machine using a pre-trained Random Forest Classifier.
• Fault Diagnosis: Identify the reasons behind machine faults with another Random Forest model.
• Historical Data Logging: Append predictions and sensor data to a CSV file for future analysis.
• Interactive Visualizations: Real-time sensor plots displayed via Streamlit and Matplotlib.
• Automated Model Training: Train Random Forest models for health and fault prediction.

---

Technologies Used

• Languages: Python
• Libraries:
  • Data Handling: pandas, numpy
  • Machine Learning: scikit-learn, joblib
  • Visualization: matplotlib, streamlit
• Firebase:
  • Realtime Database for sensor data
  • Admin SDK for secure data access
• Others: subprocess for calling scripts dynamically

---

Project Structure

📂 Predictive-Maintenance-IoT  
├── testing.py                # Streamlit-based dashboard for real-time monitoring  
├── script1.py             # Script for prediction and data logging  
├── script2.py             # Script for model training  
├── updateddataset2.csv    # Main dataset for training and predictions  
├── machine_health_model.pkl  # Pre-trained machine health model  
├── reason_model.pkl       # Pre-trained fault diagnosis model  
├── firebase-adminsdk.json # Firebase service account key  

---

Workflow

1. Real-Time Monitoring

• Sensor data is fetched continuously from Firebase.
• Live plots for temperature, RPM, vibration, and sound levels are updated in the Streamlit dashboard.

2. Machine Health Prediction

• script1.py processes incoming data and predicts machine health (Healthy, Warning, Critical).
• The root cause of any issue is also predicted using the trained model.

3. Data Logging

• Predictions and real-time sensor data are appended to the updateddataset2.csv file for analysis.

4. Model Training

• script2.py trains Random Forest Classifiers for health and fault prediction using the main dataset.

---

Setup Instructions

Prerequisites

1. Install Python (>=3.8).

2. Install the required libraries using:

   pip install pandas numpy scikit-learn matplotlib firebase-admin streamlit  

3. Set up a Firebase Realtime Database and download the service account key JSON file.

Running the Project

1. Start the real-time monitoring:

   streamlit run testing.py  

2. Train the models (if needed):

   python script2.py  

3. Perform predictions:

   python script1.py  

---

Dataset Details

The dataset (updateddataset2.csv) includes the following features:

• Temperature: Sensor data in degrees Celsius.
• Humidity: Percentage of relative humidity.
• RPM: Rotational speed of the machine.
• Sound (dB): Noise levels recorded.
• Amplitude: Vibration amplitude of the machine.
• Frequency: Vibration frequency in Hz.
• Machine Condition: Health labels (Healthy, Warning, Critical).
• Reason: Root cause labels for machine faults.

---

Future Enhancements

• Integration with IoT edge devices like ESP32 for direct data streaming.
• Advanced ML models (e.g., Deep Learning) for better accuracy.
• Mobile app integration for real-time health notifications.

---

Screenshots

Streamlit Dashboard

Real-Time Sensor Plots

---

Feel free to fork and customize the project for your use case! 🌟