Digital Twins And Neural Networks For The Diagnosis Of Industrial Systems

4 Jun 2023

Job Information

Organisation/Company

RAMLA SADDEM

Research Field

Engineering

Computer science » Informatics

Researcher Profile

Recognised Researcher (R2)

Leading Researcher (R4)

First Stage Researcher (R1)

Established Researcher (R3)

Country

France

Application Deadline

11 Jun 2023 - 22:00 (UTC)

Type of Contract

Temporary

Job Status

Full-time

Offer Starting Date

2 Oct 2023

Is the job funded through the EU Research Framework Programme?

Not funded by an EU programme

Is the Job related to staff position within a Research Infrastructure?

No

Offer Description

Directeur (ou directrice) de thèse : Bart LAMIROY – Université de Reims Champagne-Ardenne CReSTIC –

Co-encadrante : Ramla SADDEM – Université de Reims Champagne-Ardenne CReSTIC –

This PhD. revolves around Industrial Systems diagnosis, aiming to enhance their flexibility and resilience through advancements in analysis techniques and the integration of Machine Learning. This integration seeks to extend the diagnosis beyond the state of the art to more complex systems while optimizing computational resources. Simultaneously, it should provide human operators with the ability to control and initiate corrective actions through interpretable results generated by the diagnosis and isolation tools.

Initially, the research will focus on systems equipped with sensors and actuators that generate binary signals, commonly called Discrete Event Systems (DES). However, there is potential to extend the study to include Hybrid Systems that incorporate both discrete and continuous sensors and actuators. The research will primarily consider the following operational cycle: a) acquiring inputs by recording the states of non-controllable sensor variables, b) executing the control program, and c) updating the controllable actuator variables. These steps are cyclically repeated.

The project will tackle the diagnosis problem by leveraging automatic analysis techniques applied to operational data, without relying on explicit models. Instead, the emphasis will be on implementing Machine Learning approaches, particularly neural networks, to detect and isolate faults.

The PhD. will pursue the following two objectives:

Extend the preliminary diagnosis results presented in (Saddem et al., 2022a,b), and be able to automatically dimension the size of the neural networks and their hyper-parameters in function of the monitored system.

Ensure that the developed approaches go beyond the stage of black-box classifiers or predictors, and that they can provide a useful explicability level to a human operator; either by formalizing verification approaches (Fawzi et al. 2022) or others, or by analysing the trained networks (Rojat et al., 2022).

This work will rely on Cellflex4.0 at the University of Reims Champagne- Ardenne (https: // crestic.univ-reims.fr/fr/plateformes/cellflex-4-0) and its associated Digital Twins for the data acquisition and experimental validation. It will also benefit of HPC Romeo platform computing capacities.

The proposed course of the PhD. is outlined as follows:

Step 1: Familiarization with the research topic and inception of potential approaches to the problem by conducting a comprehensive literature review on Industrial Systems Diagnosis and Machine Learning techniques specifically focused on diagnosis.

Step 2: Collection and preparation of required data for the identified learning approaches. This step involves defining available data sources and selecting appropriate techniques for data collection. The collected data will then be prepared for utilization in the selected online diagnosis methods.

Step 3: Development and implementation of diagnosis algorithms using Machine Learning techniques. Algorithms will be trained to accurately return the plant state (normal or faulty and if faulty return the fault class). Experimental validation of the algorithms on both Digital Twins and Cellflex platform.

Step 4: Explaining the diagnosis provided by the neural networks through the identification the faulty components. This can be achieved through explicit search in the solution space using deep reinforcement learning or by employing traditional eXplainable Artificial Intelligence (XAI) approaches to analyze the trained networks.

Special attention will be given to the dissemination and exploitation the research findings, notably through scientific publications and the development of demonstration software.

By following this outlined plan, the thesis aims to contribute to the field of industrial system diagnosis, enabling more flexible and resilient operations while ensuring human operators can understand and utilize the diagnostic results effectively.

Funding category: Contrat doctoral

PHD Country: France

Requirements

Specific Requirements

Ce sujet de thèse s'adresse aux étudiant.e.s ayant un diplôme de Master II ou d'ingénieur, soit issu.e.s d'un parcours ayant donné accès à des connaissances approfondies en systèmes industriels (notamment systèmes à évènements discrets et hybrides) et avec une envie d'investir le champ de l'Intelligence Artificielle, soit des étudiant.e.s de formation informatique avec de très bons acquis théoriques et applicatifs en apprentissage automatique souhaitant les appliquer à des contextes de systèmes industriels.

Au-delà des bases scientifiques citées ci-dessus, les autres compétences principales recherchées sont la curiosité et l'envie d'apprendre, la capacité de travail en équipe, la rigueur scientifique et la capacité de formalisation du raisonnement. Une grande partie du travail étant consacrée à la validation expérimentale des modèles, des compétences solides en programmation seront également nécessaires.

Additional Information Work Location(s)

Number of offers available

1

Company/Institute

RAMLA SADDEM

Country

France

City

REIMS

Where to apply

Website

https: // www. abg.asso.fr/fr/candidatOffres/show/idoffre/114982

Contact

Website

https: // crestic.univ-reims.fr/

STATUS: EXPIRED