3D-AmplifAI: An Ensemble Machine Learning Approach to Digital Twin Fault Monitoring for Additive Manufacturing in Smart Factories

Abstract

In the digital age, the digital twin eliminates physical barriers and risks, facilitating seamless activities in both real and virtual worlds. In the context of additive manufacturing, testing 3D printers can be resource-intensive and prone to printing issues. This research introduces a digital twin-based system that employs the innovative ensemble 3D-AmplifAI algorithm for fault monitoring in 3D printers. The system continuously monitors real-time temperature values and detects faults to prevent potential damage to the printer. Through an ensemble method, the 3D-AmplifAI algorithm combines multiple machine learning models to enhance fault detection in 3D printers. The digital twin environment, developed using Unity, serves as the bridge connecting the physical printer to the virtual world. Comparative evaluations against state-of-the-art algorithms, including Ridge Regression, XGBoost, InceptionTime, Time Series Transformer (TST), Rocket Ridge, Logistic Regression, Rocket XGBoost, ResNet, and Rocket Ridge Regression, demonstrate the superior performance of the 3D-AmplifAI algorithm in terms of accuracy, precision, recall, and F1-score.