Predictive Maintenance Machine Learning Model

A Simple MATLAB-based machine learning model for predicting equipment failures using sensor data and maintenance records.

Overview

This project implements a Random Forest classifier to predict potential equipment failures based on various operational parameters including:

• Air Temperature (K)
• Process Temperature (K)
• Rotational Speed (RPM)
• Torque (Nm)
• Tool Wear (minutes)

Data

The dataset consists of 10,000 data points with the following features:

• UID: Unique identifier (1 to 10000)

• ProductID: Product quality variant identifier

  • L: Low (50% of products)
  • M: Medium (30% of products)
  • H: High (20% of products)

• Air Temperature [K]: Normally distributed around 300K with σ = 2K

• Process Temperature [K]: Air temperature + 10K with σ = 1K

• Rotational Speed [rpm]: Derived from 2860W power with normal noise

• Torque [Nm]: Normally distributed around 40Nm with σ = 10Nm

• Tool Wear [min]: Varies by product quality

  • H: +5 minutes
  • M: +3 minutes
  • L: +2 minutes

• Target Variables: The model predicts two targets:

  • Binary classification of machine failure (Yes/No)
  • Type of failure (when applicable)

⚠️ Important: Both targets should be treated as prediction targets and not used as features to avoid data leakage.

Link: https://www.kaggle.com/datasets/shivamb/machine-predictive-maintenance-classification

Features

• Data preprocessing and feature selection
• 80-20 train-test split
• Random Forest model training with 100 trees
• Feature importance analysis
• Model evaluation using confusion matrix
• Prediction function for new data points
• Model persistence (save/load functionality)

Usage

Prerequisites

• MATLAB
  • Statistics and Machine Learning Toolbox
  • Deep Learning Toolbox

Code Structure:

1. Data Loading and Preprocessing
2. Feature Selection
3. Train-Test Split (80-20)
4. Model Training (Random Forest with 100 trees)
5. Prediction Function
6. Model Evaluation
7. Model Persistence

Training the Model

• Place your 'maintenance_data.csv' in the working directory
• Run the main script to:
  • Load and preprocess data
  • Train the model
  • Evaluate performance
  • Save the model
• Use the predictMaintenance function for new predictions

  % Example prediction
  newData = [299, 311, 1000, 2.6, 70];
  [predicted_failure, probability] = predictMaintenance(model, newData);
  

• The trained model is automatically saved as 'maintenance_model.mat'. To load the saved model:

  load('maintenance_model.mat');
  

Author

Zenilus

Date

Last Updated: November 30, 2024