Skip to content

README.md

Latest commit

 

History

History
91 lines (76 loc) · 5.06 KB

README.md

File metadata and controls

91 lines (76 loc) · 5.06 KB

✈️ Machine Learning Project for Predicting Broach Maintenance in Airplane Engine Manufacturing ✈️

Overview

This project was developed as a focus on machine learning techniques for clustering and regression analysis. It explores real-world datasets to solve challenges and extract meaningful insights. Specifically, it addresses the critical task of predicting when to replace broaches used in manufacturing airplane engines.

The work was a collaborative effort by:

Goals

The primary objectives of this project are:

  1. Clustering Analysis: Identifying natural groupings within the data to understand broach wear patterns.
  2. Regression Modeling: Developing models to predict when broaches should be replaced to minimize downtime and costs.
  3. Impact on Manufacturing: Improving product quality and reducing manufacturing costs for ITP Aero.

Key Features

  • 🔀 Clustering Techniques: Analysis using various clustering algorithms, including K-Means and hierarchical clustering.
  • ⚛️ Regression Models: Implementation of advanced regression techniques, including Lasso and GBM (Gradient Boosting Machine), tailored for industrial datasets.
  • 📊 Dataset Handling: Preprocessing, cleaning, and validating the data for robust model performance.

Files

  • clustering_analysis.R: Script implementing clustering techniques.
  • ITPaero.R: Regression modeling script using ITP Aero dataset.
  • TrabajoFinal.R: Final integration script combining insights from all analyses.
  • ITPaero.csv: Dataset provided for regression modeling.

Dataset ✈️🚀

The dataset used in this project comes from ITP Aero, focusing on broach calibration and wear patterns. Key features of the dataset include:

  • Training Dataset: Includes parameters such as broach usage time, wear indicators, and operating conditions.
  • Validation Dataset: Contains the target variables to evaluate model accuracy.
  • Target Variables: Four key target variables were analyzed: XCMM, ZCMM, BCMM, and CCMM, representing deviations across different axes (X, Z, B, C).
  • Metrics Evaluated:
    • Absolute Maximum Error: Less than 1 for the corrector on the X-axis and 0.15 for others.
    • RMSE: Less than 0.25 for the corrector on the X-axis and 0.025 for others.
  • Significant Variables: Factors like tooling, machine type, broach type, and manufacturing order were found to be highly significant.

Data Insights

  • Normal Distribution: All target variables show a reasonably normal distribution, with XCMM having the highest variance, making it the most challenging to predict.
  • Correlation Analysis: Strong correlations were observed, e.g., XCMM with BCMM, and nbrochahss with nusos (correlation = 0.75).
  • Optimal Clustering: Two clusters were identified as the most optimal separation for the data, with minimal overlapping.

Compliance with Requirements

The models developed in this project meet all requirements specified:

  1. Error Metrics:
    • Absolute Maximum Error and RMSE values comply with ITP Aero thresholds for all axes.
    • Models were extensively validated to ensure accuracy and precision.
  2. Computational Efficiency:
    • Optimization techniques were applied to ensure models run efficiently, reducing computational time while maintaining accuracy.
  3. Originality and Applicability:
    • Novel preprocessing techniques were implemented to handle noisy and missing data.
    • The methodology is adaptable to other predictive maintenance scenarios.
  4. Presentation and Clarity:
    • The process, results, and insights are clearly documented and visualized for ease of interpretation.

Results

  1. ✈️ Clustering:

    • Identified two distinct clusters representing different broach wear stages.
    • Clear separation ensures actionable insights for maintenance scheduling.

  2. 🚀 Regression:

    • GBM (Gradient Boosting Machine) emerged as the best-performing model.
    • Achieved RMSE below the threshold of 0.25 on specific axes.
    • Models successfully predict deviations and recommend corrective actions for broach replacement.

Instructions

🔧 Setup

  1. Clone this repository: 
    git clone https://github.com/dantesc03/BroachAlign-Machine-Learning.git
  2. Install required dependencies in R. 
    install.packages(c("dplyr", "ggplot2", "caret", "tidyr", "readr", "cluster", "factoextra", "gbm"))
  3. Load the scripts and dataset into your R environment.

💡 Execution

  • Run clustering_analysis.R for clustering insights.
  • Use ITPaero.R for regression analysis.
  • Execute TrabajoFinal.R for a comprehensive summary.

Acknowledgments

  • ITP Aero for providing the dataset and the challenge.
  • Our professor and classmates for guidance and support.

📢 Contact

For any inquiries or feedback, reach out to us: