Predictive Maintenance System using Machine Learning

Project Overview: -

This project demonstrates the development of a predictive maintenance system utilizing machine learning techniques. The model predicts potential machine failures based on sensor data, aiming to reduce unplanned downtime, enhance equipment effectiveness, and optimize maintenance schedules. The system uses the AI4I 2020 dataset for training and evaluation.

Dataset: -

• Source: AI4I 2020 Predictive Maintenance Dataset
• Samples: 10,000 records
• Features:
  • Machine Type: L (Low), M (Medium), H (High)
  • Air Temperature: Measured in Celsius
  • Process Temperature: Measured in Celsius
  • Rotational Speed: Machine rotational speed (RPM)
  • Torque: Machine torque
  • Tool Wear: Wear of the machine tool
• Target: Machine Failure (0 = No, 1 = Yes)

Objective: -

The goal of this project is to:

• Predict machine failures in advance
• Reduce unplanned downtime
• Improve equipment effectiveness
• Optimize maintenance schedules

Technologies Used: -

• Python
  • Libraries: Pandas, NumPy, Seaborn, Matplotlib
  • Machine Learning: Scikit-learn, XGBoost, LightGBM
  • Data Preprocessing: Imbalanced-learn (SMOTE for target balancing)
  • Model Persistence: Joblib (to save and load models)
• Google Colab / Jupyter Notebook

Approach & Methodology: -

Data Preprocessing

• Handled missing values (if any)
• Standardized features using StandardScaler
• Balanced the target variable using SMOTE (Synthetic Minority Over-sampling Technique)

Model Training

Multiple machine learning models were trained and evaluated:

• Random Forest
• Support Vector Machine (SVM)
• XGBoost
• LightGBM

Model Evaluation

Model performance was evaluated using key classification metrics:

• Accuracy
• F1-Score

Model Selection

The Random Forest classifier outperformed others, achieving:

• Accuracy: 99.90%
• F1-Score: 0.9833

The best model was saved using Joblib for future use.

Results: -

Model	Accuracy	F1 Score
Random Forest	99.90%	0.9833
SVM	98.58%	0.9856
XGBoost	99.59%	0.9958
LightGBM	99.48%	0.9948

How to Run the Project: -

1. Clone the repository:

   git clone https://github.com/sujata1712/Predictive Maintenance System using Machine Learning.git
   

2. Navigate to the project directory:

   cd Predictive Maintenance System using Machine Learning
   

3. Install the required dependencies:

   pip install -r requirements.txt
   

4. Open and run the Predictive-Maintenance-System.ipynb notebook in Google Colab or Jupyter Notebook.

Conclusion: -

This project provides a comprehensive solution for predicting machine failures in industrial settings. By leveraging machine learning algorithms, the system optimizes maintenance schedules, reduces downtime, and enhances overall productivity. The approach outlined here can be applied to various industries to improve the efficiency of maintenance operations.