Enhanced Changeover Detection in Industry 4.0 Environments with Machine Learning.

Area of Science:

• Manufacturing Engineering
• Data Science
• Industrial Automation

Background:

• Changeover times significantly impact Overall Equipment Effectiveness (OEE) in production.
• Accurate detection of changeovers is crucial for optimizing manufacturing processes.
• Manual tracking of changeovers is time-consuming and prone to errors.

Purpose of the Study:

• To develop and evaluate a machine learning (ML) approach for automatic changeover detection in a shopfloor environment.
• To compare the performance of various ML algorithms for this task.
• To identify optimal model configurations for accurate changeover phase identification.

Main Methods:

• Utilized sensor data from a milling machine, including door status, coolant flow, power consumption, and operator GPS.
• Implemented and compared Decision Trees, Support Vector Machines, Random Forest, and Neural Networks.
• Evaluated model performance using metrics such as F1 score and AUC.

Main Results:

• The Random Forest ML model demonstrated superior performance, achieving a 97% F1 score and 99.72% AUC score.
• Optimal model performance was observed when classifying only two phases: changeover and production.
• Reducing the number of sub-phases improved overall model accuracy.

Conclusions:

• An ML-based approach using external sensors effectively automates changeover detection.
• The Random Forest algorithm is a highly effective method for this application.
• Simplifying the classification task to binary (changeover vs. production) enhances detection accuracy, aiding OEE improvement.