INVITED SPEAKER

Ramla SADDEM is an Associate Professor at the EiSINE School of Industrial and Digital Sciences Engineering at the University of Reims Champagne-Ardenne (URCA) in France. She has been a member of the CReSTIC laboratory since September 2013. She’s academic journey began with her Engineering degree from the National School of Computer Sciences (ENSI) in Tunisia, which she achieved in 2006. She furthered her education by obtaining a Master's degree in Artificial Intelligence and Decision Support Systems in 2007. She gained practical industry experience as an analysis and development engineer at International Information Development, Euro Information, the IT subsidiary of the CM-CIC banking group in Tunisia. Then, she embarked on a Ph.D. program in industrial computer science and automation, jointly supervised by Centrale Lille in France, from 2009 to 2012. She earned her Ph.D. degree in industrial computer science and automation, as well as a Ph.D. degree in Computer Science from ENSI. To expand her expertise, Ramla SADDEM completed a PostDoc at Centrale Lille from January to August 2013. Ramla SADDEM's primary research interests revolve around the application of Artificial Intelligence (AI) and machine learning (ML) techniques in the field of fault diagnosis of Discrete Event Systems. Through her work, she aims to contribute to the advancement of AI and ML, particularly in their application to improve fault diagnosis processes using Digital Twin one of the tools of the industry of the future. She has published more than 40 papers in peer-reviewed international journals and proceedings of international conferences and has supervised 3 Ph.D.

Title: Benefits of Using Digital Twin for Online Fault Diagnosis of a Manufacturing System

Abstract: 

In this presentation, we highlight the significance of utilizing a digital twin for online fault diagnosis in a manufacturing system. The manufacturing system under consideration is equipped with sensors and actuators that provide binary signals, which can be effectively modeled as Discrete Event Systems. Traditional fault diagnosis solutions are frequently non-industrializable, prompting the need for a novel data-driven approach. To address this, we propose an intelligent diagnostic solution that leverages the power of simulation to learn from plant behaviors. Our methodology involves the utilization of recurrent neural networks (RNN) with both short-term and long-term memory capabilities, specifically employing Long Short-Term Memory (LSTM) networks.

By employing a digital twin, we can create a virtual replica of the manufacturing system. This digital representation allows us to simulate various operating scenarios, generating valuable data that can be used for training and testing the RNN-based fault diagnosis model. The LSTM architecture is well-suited for capturing temporal dependencies and patterns within the data, making it an ideal choice for this task. Our proposed approach offers several advantages over traditional methods. Firstly, it provides a more industrially viable solution by leveraging the power of data-driven techniques. The ability to learn from simulated plant behaviors enables the model to adapt to real-world conditions effectively. Secondly, the use of RNNs with LSTM facilitates the detection of complex fault patterns and improves diagnostic accuracy. Through this presentation, we aim to showcase the potential of utilizing a digital twin coupled with RNN-based fault diagnosis in the manufacturing domain. Our research demonstrates how this innovative approach can enhance fault detection and diagnosis capabilities, leading to improved system reliability, reduced downtime, and increased productivity.

Keywords: Digital twin, Online fault diagnosis, Discrete event systems, Automated production systems