Where PhDs and companies meet

Already registered?

New user?

Multi-physics multi-scales seismological digital twin in complex geological media for industrial risks

ABG-105513 Thesis topic
2022-05-11 Public funding alone (i.e. government, region, European, international organization research grant)
Institut de Recherche en Constructibilité (IRC) - ESTP-Paris
Orléans - Centre Val de Loire - France
Multi-physics multi-scales seismological digital twin in complex geological media for industrial risks
  • Engineering sciences
  • Computer science
  • Mathematics
earthquake, digital twin, multi-physics multi-scales numerical simulation, interoperability

Topic description


The term « digital twin », in the manner of the artificial intelligence, of the machine learning, of the digital platforms, or even of the big data, the cloud computing or the smart cities, is part of the emergent terms whose signification evolves over time and disciplines. The term “digital twin”, although it is not explicitly mentioned, is close to the concept of “mirror worlds” employed in the nineties by Gelernter (1993) whose prolog starts with: “This book describes an event that will happen someday soon: You will look into a computer screen and see reality”. The exact term “digital twin” would have been established in Product Conception by Grieves (2014), who says himself in his article to have borrow the term from John Vickers from the NASA. In Computer-aided design (CAD), Grieves & Vickers (2017) define a digital twin like being the numerical replica of a physical object, in other words, a computer aided drafting. According to Grieves & Vickers (2017), it was initially “static”, and then evolved to become “dynamic” and being the support of simulations on which forces can be applied to model its dynamic response and associated stress-strain state. One year later, Batty (2018) took the digital twin concept over and defines it, as Gelernter did, like “a mirror image of a physical process that is articulated alongside the process in question, usually matching exactly the operation of the physical process which takes place in real time”. This definition starts coming up on online encyclopedia to define it as: “digital twins blend artificial intelligence, machine learning and data analysis to create numerical simulation models updating and changing as their physical counterpart change. A digital twin permanently learns and upgrade itself from multiple sources to represent its status […] in almost real-time” .
In earthquake engineering, a digital twin can be seen as a numeric replica allowing numerical simulation models to tend toward a realistic representation of the Earth. With this in mind, we propose a PhD subject whose objective is to understand the relative importance of the different processes governing earthquakes, via the construction of multi-physics multi-scales and interoperable digital twins by object of interest (fault, large seismological structures, lithostructure beneath a site of study, etc.) with the aim of improving earthquake ground motion prediction, as well as its spatial variability, whose accurate assessment is vital to ensure the integrity of critical industrial facilities when an earthquake occurs, as well as guarantying a sustainable development of societies under growing pressures with regards to land planning.
This PhD will investigate four fundamental processes to seismic risk assessment, namely: i) the kinematic rupture of the fault governed by the eikonal equation, ii) the wave propagation in multi-scales geological structures governed by the equation of motion, iii) the complex behavior (potentially non-linear) of the subsurface soils, and iv) the seismic response of civil engineering structures. The numerical simulations will be performed on regional or national high-performance computing architectures with the community open-source code EFISPEC3D  (De Martin et al., 2021). Workflows and Machine Learning tools will be developed in order to automatically post-process the large quantity of data (~ To) generated by earthquake scenarios.
More specifically: which phenomena – from the fault’s rupture to the site of study – predominantly govern the ground motion and which ones are negligible? What would be the portion of variability linked to each of these phenomena? Which spectral densities, correlation lengths and variances would be adequate to reproduce this variability? Which attenuation mechanism (i.e., scattering vs rheology) are predominant in this variability? Which characteristic physical quantities are easily transferable to engineering? Do we always need advanced physical models to predict those quantities? Which physical parameters influence the response spectra of the civil engineering structures? Which computational power is required to generate and compute realistic multi-scale heterogeneous media?
Although this PhD will not exhaustively tackle the above-mentioned questions, it will focus on the quantification, through the construction of multi-physics multi-scales seismic digital twins, of the spatial variability of seismic wave field with respect to the stochastic uncertainties of the fault rupture and to the epistemic uncertainties of the geological media within which the complexification of the wave field occurs. The study will take place on a site where geological and geophysical data are available, namely, an area of the Capesterre-Belle-Eau district in Guadeloupe including an industrial civil engineering structure: the Dumanoir dam for which a seismic digital twin has been initiated (De Martin et al., 2022) and which is the numerical support of earthquake ground motion predictions stemming from surrounding fault systems. The methodology done on the Dumanoir site could be applied to other sites under high seismic industrial risks, like the Le Teil one in Ardèche close to nuclear power plants or the Grenoble valley site.
It is worth to mention that the societal impacts contributing to land’s resilience will not be tackled in the framework of this PhD due to the substantial workload needed to treat the primary issue. Nonetheless, results would be released via general public communication vectors, like the Georisques platform , Grenoble’s major risks institute or the «Institut Séism» scientific interest consortium of which BRGM is member.


Starting date


Funding category

Public funding alone (i.e. government, region, European, international organization research grant)

Funding further details

The PhD will be funded by the chair, with a gross annual salary around 26 k€ including short-term contract allowance

Presentation of host institution and host laboratory

Institut de Recherche en Constructibilité (IRC) - ESTP-Paris

Welcoming team and place

The PhD will take place both at the French Geological Survey (BRGM, in the Seismic and Volcanic Risks Team of Orléans, France) and at the ESTP (Cachan or Orléans). The host periods between the institutes will have to be fairely shared (ex: year one at BRGM, year two at ESTP, year 3 to be defined).

The PhD will be managed by Prof. Fakhreddine ABABSA (ENSAM) and supervised by two university lecturers and researchers from ESTP and ENSAM, as well as by a BRGM expert (F. DE MARTIN).

Candidate's profile

  • Engineer or research master’s degree in applied mathematics or informatics, with a strong appetency to algorithmic coding related to Machine Learning and to the understanding of physical phenomena through the resolution of partial differential equation
  • Taste for applied research, knowledge of numerical method, Linux environment, Programmation language (Fortran, Python), etc.
  • Rigor and capacity to invest a lot in a research subject
  • Team work skill
  • Good writing skill and fluent in English (elementary French would be appreciated)
Partager via

Vous avez déjà un compte ?

Nouvel utilisateur ?