Meta-models for linking powertrain design - Thèse ITN H/F (Recherche & Développement/Autre)

Job description

Filière/Métier : Recherche & Développement/Autre

Contrat : Doctorat CIFRE

Description du poste :

This PhD position is part of the “ECO DRIVE” project, funded by the European Commission through the H2020 “Marie Skłodowska-Curie Innovative Training Networks” program (grant number 858018) and focusing on noise and vibration in eco-efficient powertrains of future vehicles. Different universities and companies in Europe are part of the project, which includes various training and exchange periods for all the PhD candidates involved.
The application context is the global design process for the next generation of electrified powertrains (including hybridized engines, electric motors, gearboxes, drivelines, etc.) The ongoing massive electrification of powertrains brings quite a large number of challenges in car industry, particularly in engineering and design concerning acoustic issues.

Scientific framework and global objectives :

This research work aims at developing advanced vibroacoustic “meta-”models to help engineering and design process of automotive electric powertrains. These models will typically use recent advances in probabilities, Bayesian inference and other IA/ML-related tools. They will use information both from experimental (more accurate but expensive) and numerical (cheaper but uncertain) models, which today are considered rather separately to justify engineering decisions. More precisely, the meta-model function is to infer a probabilistic vibro-acoustic performance estimator from a set of known design parameters, which are mostly geometrical and related to the dynamic structural properties. The vibro-acoustic performance estimators will be defined using global quantities coming from the different models, such as injected structural power, source equivalent forces, overall radiated power, etc. One key feature of the model is its reduced order/CPU efficiency and, then, its ability to perform in global multi-objective optimization processes. Another feature of these meta-models should be their ability to get easily updated and improved when the number of experimental data and knowledge increase with time, thereby increasing the meta-model's performance. The end-user target is a knowledge model able to guide the early-stage engineering decision processes on a rational, comprehensive and objective basis.


The candidate must have completed an M2 level with excellent academic results in (both) applied mathematics and mechanical engineering.

They also have to meet the eligibility conditions of ETN projects:

• no residence in France longer than 12 months in the past 3 years immediately before the date of recruitment
• not been involved in research for more than 4 years (full time equivalent) starting to count the date this person graduates his/her first MSc degree.

The PhD will take place in France (Paris area), with training periods in K.U. Leuven (Belgium) and INSA Lyon (University of Lyon).

Ville : Vélizy-Villacoublay

Niveau de diplôme préparé : Bac+8