A Digital Twin Framework for LSTM-Based Fault Diagnosis in Discrete Event Manufacturing Systems

Abstract

Digital Twin (DT) technology is increasingly used in manufacturing to enable realtime monitoring, prediction and decision support. In this work, we propose a DT dedicated to fault diagnosis in manufacturing systems modeled as Discrete Event Systems. Building on our previous contribution, which introduced a data-driven diagnostic method based on Long Short-Term Memory (LSTM) neural networks, we present an improved version of this approach and deliver a turnkey solution suitable for both shop-floor operators and plant managers. The effectiveness of the proposed DT is demonstrated using the CellFlex plant, a training and research platform at the URCA. CellFlex plant consists of eight stations operating around a central conveyor, forming a flexible miniaturized bottling line connected through industrialstandard networks. The obtained results confirm the relevance and practical applicability of the proposed approach for online fault diagnosis in industrial environments.

Mehr zum Titel