Belt Conveyor Health Monitoring System

A Python and Streamlit-based dashboard for predictive maintenance and health monitoring of belt conveyor equipment, using sensor data and machine learning for real-time condition classification, visualization, and maintenance planning[1][2].

---

Features

• Sensor Data Ingestion

  • Upload CSV files with sensor readings (timestamp, tension, vibration, remaining_life)
  • Manual entry of sensor features for quick predictions
  • Demo and real-time data simulation modes

• Preprocessing & Feature Engineering

  • Time-based interpolation and resampling
  • Rolling window statistics (mean, std) for tension
  • FFT-based vibration feature extraction
  • Automatic health condition labeling: Critical, Warning, Healthy

• Machine Learning Model

  • Random Forest classifier for predicting equipment condition
  • Model training, evaluation, and persistence (equipment_fault_predictor.pkl)
  • Predicts health condition from engineered features

• Interactive Dashboard (Streamlit)

  • Data preview and prediction results table
  • Multiple visualization options:
    • Matplotlib, Plotly, and Seaborn trend plots
    • Interactive custom trend explorer
    • Vibration spectrogram and 3D vibration analysis
  • Anomaly detection using Isolation Forest
  • Statistical health metrics (MTBF, failure probability)
  • Feature importance visualization

• Maintenance & Operations Tools

  • Equipment profile manager (length, load)
  • Sensor calibration log with expiry warnings
  • Maintenance scheduler and overdue alerts
  • Spare parts inventory tracking
  • Maintenance history timeline
  • PDF report generation (placeholder)

• User Access Control

  • Role-based sidebar (Operator, Engineer, Admin)
  • Read-only and advanced access modes

---

Installation

Requirements

• Python 3.8+
• pip

Install Dependencies

pip install pandas numpy scipy scikit-learn matplotlib seaborn plotly streamlit joblib

---

How to Run

1. Train the Model (Optional)

If you want to retrain the model with your own data:

python your_script_name.py

• Expects a file named sensor_maintenance_dataset.csv in the same directory.
• Outputs equipment_fault_predictor.pkl after training.

2. Launch the Dashboard

streamlit run your_streamlit_dashboard.py

• The dashboard will open in your browser.

---

Usage

A. Data Upload Workflow

1. Select "Upload CSV File" in the sidebar.
2. Upload your sensor data file (must include columns: timestamp, tension, vibration, remaining_life).
3. View data preview, predictions, and visualizations.
4. Explore trends: Tension/vibration over time, rolling means, spectrograms, and custom charts.
5. Review anomaly detection, maintenance schedules, inventory, and history.

B. Manual Input Workflow

1. Select "Manual Sensor Input" in the sidebar.
2. Enter values for tension_mean, tension_std, and vibration_fft_max.
3. Click "Predict Health Condition" to get an instant classification and recommended action.

C. Demo & Simulation

• Use sidebar buttons to load demo data or simulate real-time sensor streams.
• Useful for testing and demonstration without real sensor input.

---

File Structure

File	Purpose
predictive_maintenance.py	Model training and CLI prediction
belt_dashboard.py	Streamlit dashboard app
equipment_fault_predictor.pkl	Trained ML model (auto-generated)
sensor_maintenance_dataset.csv	Example training data

---

Notes

• The dashboard expects preprocessed features: tension_mean, tension_std, vibration_fft_max.
• Health condition labels are derived from remaining_life:
  • Critical: ≤ 50
  • Warning: 51–100
  • Healthy: >100
• Some features (PDF report, advanced filtering) are placeholders for future extension.

---

Example

Manual Prediction:

import pandas as pd
import joblib

model = joblib.load('equipment_fault_predictor.pkl')
new_data = pd.DataFrame({
    'tension_mean': [1200.0],
    'tension_std': [20.0],
    'vibration_fft_max': [85.0]
})
prediction = model.predict(new_data)
print(f"Predicted Condition: {prediction[0]}")

---

Developed with ❤️ using Streamlit