README.md

metadata

title: Predictive Maintenance System
emoji: 🔧
colorFrom: blue
colorTo: green
sdk: streamlit
sdk_version: 1.29.0
app_file: app.py
pinned: false

Predictive Maintenance System - AI4I 2020 Dataset

Project Overview

This project implements a comprehensive predictive maintenance system using machine learning to predict when industrial machinery needs maintenance. The system analyzes the AI4I 2020 Predictive Maintenance Dataset and provides interactive visualizations and real-time predictions through a Streamlit web application.

Features

• Comprehensive EDA: 15+ different exploratory data analyses
• Machine Learning Model: Random Forest Classifier for failure prediction
• Real-time Predictions: Interactive interface for runtime predictions
• Maintenance Scheduling: Estimates time to failure and maintenance urgency
• Interactive Visualizations: Dynamic charts and graphs using Plotly
• Batch Processing: Upload CSV files for batch predictions

Dataset

The AI4I 2020 Predictive Maintenance Dataset contains:

• 10,000 machine records
• 14 features including temperature, rotational speed, torque, and tool wear
• Binary target: Machine failure (0 = no failure, 1 = failure)
• 5 failure types: TWF, HDF, PWF, OSF, RNF

Project Structure

project/
├── ai4i2020.csv              # Dataset file
├── analysis.py               # EDA analysis script
├── preprocessing.py          # Data preprocessing module
├── model.py                  # Machine learning model
├── train_model.py           # Script to train and save model
├── app.py                    # Streamlit web application
└── README.md                 # Project documentation

Installation

1. Clone or download the project
2. Install dependencies (requirements listed below):

pip install pandas==2.1.4 numpy==1.26.2 matplotlib==3.8.2 seaborn==0.13.0 scikit-learn==1.4.0 streamlit==1.29.0 plotly==5.18.0

3. Ensure the dataset file (ai4i2020.csv) is in the project directory

Usage

Option 1: Run Streamlit App Directly (Recommended)

The app will train the model automatically on first run:

streamlit run app.py

Option 2: Train Model First, Then Run App

1. Train the model:

python train_model.py

2. Run the Streamlit app:

streamlit run app.py

Option 3: Run EDA Analysis Only

python analysis.py

Streamlit Application

The web application includes four main sections:

1. Introduction

• Dataset overview and statistics
• Project goals and features
• Dataset preview

2. Exploratory Data Analysis

Interactive visualizations including:

• Summary statistics
• Target distribution
• Feature distributions
• Correlation analysis
• Failure analysis by machine type
• Tool wear analysis
• Temperature analysis
• Outlier detection
• Pairwise relationships
• Failure type breakdown

3. Model & Predictions

• Model performance metrics
• Feature importance visualization
• Runtime prediction: Enter machine parameters to predict maintenance needs
• Batch prediction: Upload CSV file for multiple predictions
• Maintenance urgency assessment
• Time-to-failure estimation

4. Conclusion

• Key findings and takeaways
• Applications and future improvements

EDA Analyses Performed

1. Summary statistics (mean, median, mode, etc.)
2. Missing value analysis
3. Data types and unique value counts
4. Target variable distribution
5. Correlation analysis
6. Outlier detection (IQR method)
7. Feature distribution analysis
8. Failure analysis by machine type
9. Tool wear analysis
10. Temperature analysis
11. Power and rotational speed analysis
12. Pairwise feature relationships
13. Detailed failure type analysis
14. Time to failure estimation
15. Grouped aggregations

Machine Learning Model

• Algorithm: Random Forest Classifier

• Features: 11 engineered features including:

  • Air temperature, Process temperature
  • Rotational speed, Torque
  • Tool wear
  • Temperature difference
  • Power
  • Machine type encoding

• Evaluation Metrics:

  • Accuracy
  • Precision
  • Recall
  • F1-Score
  • ROC-AUC

Predictive Features

The model predicts:

1. Machine Failure: Binary prediction (Yes/No)
2. Failure Probability: Probability score (0-1)
3. Time to Failure: Estimated minutes until maintenance needed
4. Maintenance Status: Current maintenance requirement status
5. Maintenance Urgency: CRITICAL, HIGH, MEDIUM, or LOW

Runtime Prediction

Users can input machine parameters:

• Machine Type (L, M, H)
• Air Temperature (K)
• Process Temperature (K)
• Rotational Speed (rpm)
• Torque (Nm)
• Tool Wear (minutes)

The system provides:

• Failure prediction
• Maintenance urgency level
• Estimated time to failure
• Detailed recommendations

Technical Stack

• Python 3.8+
• Data Processing: Pandas, NumPy
• Visualization: Matplotlib, Seaborn, Plotly
• Machine Learning: Scikit-learn
• Web Framework: Streamlit

Key Insights

1. Tool wear is the most critical indicator of machine health
2. Temperature difference between process and air temperature correlates with failures
3. Machine type affects failure rates differently
4. Early detection can prevent costly downtime
5. Proactive maintenance scheduling can optimize operations

Applications

• Industrial manufacturing
• Equipment monitoring systems
• Preventive maintenance scheduling
• Cost reduction through failure prevention
• Production optimization

Future Enhancements

1. Real-time data streaming integration
2. IoT sensor integration
3. Advanced ensemble methods
4. Time-series analysis
5. Automated alert system
6. Historical maintenance record integration

Author

Developed as part of the Introduction to Data Science course project.

License

This project is for educational purposes.

Acknowledgments

• AI4I 2020 Predictive Maintenance Dataset
• Scikit-learn documentation
• Streamlit documentation

---

Note: Make sure the ai4i2020.csv file is in the same directory as the scripts before running the application.