Paper Title
ONLINE MACHINE LEARNING IN SELF-LEARNING DIGITAL TWINS: A CASE STUDY ON 42SICR STEEL OPTIMIZATION

Abstract
Digital twins have become crucial in modern manufacturing, enabling the virtual replication and optimization of physical systems. Traditional digital twins are limited by static models that cannot adapt to real-time changes, leading to inefficiencies in dynamic environments. This paper introduces a framework for self-learning digital twins that employs online learning to integrate real-time data, continuously enhancing predictions of material properties such as tensile strength and elongation. XGBoost was utilized to incrementally update the model, allowing for faster adaptation and improved prediction accuracy, saving time in the manufacturing process. The use case applied to 42SiCr steel illustrates the benefit of maintaining an up-to-date model that optimizes material performance under varying conditions. The results establish a foundation for responsive and autonomous manufacturing, with future research focusing on refining model adaptability and exploring its scalability to diverse scenarios. Keywords - Self-Learning Digital Twin, Material Property Optimization, Online-Machine Learning Model, Industry 5.0