PhD Position – Hybridization of Simulation and Learning Models in Digital Twins for Mechanical Systems

Position Overview

We are seeking a motivated PhD candidate to contribute to the Engineering Digital Twins (EDT) program within Catalyst: the Reliable Hybrid Model Forge.

Digital twins are emerging as a key technology to support the continuous engineering process of mechanical systems. They combine model-based simulation with data-driven approaches to provide monitoring, prediction, optimization, and decision support capabilities.

Modeling and simulation traditionally rely on analytical or physics-based models to describe system behavior, while data science and machine learning exploit observational data to build predictive models. Rather than opposing these approaches, this project explores how they can be combined through principled hybridization strategies. This would enable a coordinated use of both techniques in complex scenarios (e.g., analytical models for explanation, and data model for recurrent pattern retrieval). Moreover, the hybridization also opens the door to adaptive modeling, where one model is inferred or refined by the others, and vice-versa (e.g., inferring or refining an analytical model from a learning model, and better tuning and explaining a learning model thanks to an analytical model).

Research Focus

The main goal is to establish the first unifying theory for both model simulation and learning models. More precisely, this PhD will investigate concepts, abstractions, and mechanisms enabling hybrid modeling through three main research directions.

Hybrid Modeling

This objective focuses on the coordinated use of heterogeneous models in digital twin environments.

The research will define concepts and abstractions to specify hybrid modeling scenarios combining techniques such as: modeling and simulation,machine learning, data mining and statistical modeling.

These concepts will provide the foundations for implementing hybrid models in digital twin services.

Adaptive Modeling

Hybrid models also enable adaptive modeling workflows, where models interact and improve each other.

Examples include:

• refining analytical models based on learned patterns
• constraining machine learning models using physical knowledge
• calibrating simulation models using observational data.

This objective aims to formalize adaptive interactions between models, enabling continuous refinement and improvement of digital twin models.

Model Interfaces and Protocols

To enable hybrid and adaptive modeling scenarios, models must interact through well-defined interfaces and protocols.

This research direction focuses on defining a hybrid platform capable of:

• orchestrating heterogeneous models
• coordinating simulation and learning models
• supporting model adaptation workflows.

Application Domain

The research results will be validated on industrial mechanical systems, particularly in the domain of process equipment and industrial machinery.

Target application areas include:

• fluid process systems
• mobile off-road working machines
• industrial production machines such as welding or machining systems.

An experimental platform available at CETIM will support the validation of the research results.

The JNEM Hydraulic Loop

The JNEM loop is an instrumented hydraulic process loop representative of industrial fluid systems.

The system includes: a pump, a heat exchanger, a tank, a regulation and multiple piping sections. It measures key physical quantities such as flow, pressure, and temperature, and includes devices allowing artificial generation of faults.

This experimental setup will enable the development and validation of digital twin applications including:

• predictive maintenance
• process optimization
• decision support
• virtual sensing and monitoring

Examples include detecting pipe clogging, estimating performance degradation, optimizing pump and valve settings, or predicting system behavior through virtual sensors.

Research Environment

This PhD is funded by CETIM (the French Technical Center for Mechanical Industries) in collaboration with Inria.

The candidate will join either:

• the Inria DiverSE team in Rennes
• or the Inria Kairos team in Sophia Antipolis.

Qualifications

Required

• Master’s degree in Computer Science, Data Science, Software Engineering, or Applied Mathematics
• strong programming skills
• background in modeling, simulation, or machine learning
• interest in cyber-physical systems and digital twins
• good communication skills in English

Preferred

• knowledge of scientific computing or numerical simulation
• experience with machine learning methods
• interest in modeling languages and engineering tools
• curiosity for mechanical systems and industrial applications

Application Process

Please submit your application including:

1. Cover Letter explaining your motivation and research interests
2. Curriculum Vitae
3. Academic Transcripts
4. Short Research Statement (1–2 pages)
5. Contact information for two references

Contact Information

Main supervisor: Supervisor: Julien Deantoni
Institution: Université Côte d’Azur (UniCA)
Email: julien.deantoni@univ-cotedazur.fr

Co-supervisor: Supervisor: Benoit Combemale
Institution: Université de Rennes / Inria
Email: benoit.combemale@inria.fr

For scientific questions regarding the PhD topic, please contact the main supervisor directly.

Funding and Benefits

• Duration: 3 years
• Salary: Competitive PhD stipend according to French standards
• Benefits: Social security, health coverage, research travel support
• Travel: Support for conference attendance and research collaborations

About the EDT Program

The Engineering Digital Twins (EDT) program is a major French research initiative bringing together leading institutions to advance the science and engineering of digital twin technologies. The program aims to establish international leadership in digital twin engineering through interdisciplinary collaboration.