e-powerTrain prEdictive mAintenance using physics inforMed learnING

Informacje na temat projektu

TEAMING

Identyfikator umowy o grant: 101131278

DOI

10.3030/101131278

Data podpisania przez KE 11 Października 2023

Data rozpoczęcia 1 Stycznia 2024

Data zakończenia 31 Grudnia 2027

Finansowanie w ramach

• Marie Skłodowska-Curie Actions (MSCA)

Koszt całkowity

Brak danych

Wkład UE

€ 1 283 400,00

Koordynowany przez

COMMISSARIAT A L ENERGIE ATOMIQUE ET AUX ENERGIES ALTERNATIVES
France

Periodic Reporting for period 1 - TEAMING (e-powerTrain prEdictive mAintenance using physics inforMed learnING)

Okres sprawozdawczy: 2024-01-01 do 2025-12-31

Podsumowanie kontekstu i ogólnych celów projektu

Mobility electrification plays a critical role in the economy decarbonisation, and we are on the edge of an industrial revolution linked to the massive deployment of the electric vehicle (EV). Their technologies readiness level has significantly increased, and the EV can now replace the thermal vehicle in terms of service provided, supporting the EU decarbonisation effort. Besides the reduction of critical material, and decrease of cost, optimising the lifetime of the EV components is essential to ease their adoption, especially the powertrain sub-components that have the major impact on EV cost and CO2 emissions. A new-generation of diagnostic and prognostic systems for the powertrain will be a game changer to ensure EV adoption, because they will estimate its degradation, anticipate failures, and ease reparability thus extending its lifespan. With significant improvement of sensors, complex modelling and data processing methods such as Artificial Intelligence (AI), predictive maintenance (PdM) has gained a lot of interest in different fields. Development of PdM methods for the sub-components of the EV powertrain (battery, fuel cell, e-motor, power electronics) is at the heart of TEAMING. Thanks to international staff exchanges and through experts’ knowledge sharing and transfer, TEAMING will significantly improve the different facets of the PdM solution: sensors, modelling, Digital Twins, adapted AI, and Physics-Informed Machine Learning methods are at the centre of the studies and present a major potential in term of innovation. TEAMING will advance PdM system to better diagnose the internal physical phenomena of the different EV powertrain components, predict the behaviour of the sub-components in various scenarios of operation, and take corrective measures to optimize their performance, lifetime, safety, and reliability.
As such, TEAMING objectives can be described:
• Create and consolidate research collaboration links to be leveraged toward future collaborative projects
• Share knowledge among partners through secondments, internal workshop, trainings
• Foster acceleration of research activities through knowledge sharing and dissemination
• Improve performance and durability of eV powertrain elements through failure anticipation and predictive monitoring
• Optimize eV powertrain elements lifetime (durability) through DT approaches augmented by machine learning
• Improve prediction of remaining useful life for eV powertrain elements and the powertrain as a whole
• Ensure technico-economic and societal impact of improved predictive maintenance on eV usage through LCA

Prace wykonane od początku projektu do końca okresu sprawozdawczego oraz najważniejsze dotychczasowe rezultaty

TEAMING project was launched in January 2024 with a KO meeting organized in Grenoble.
During the first months, TEAMING was engaged in setting up the procedure to follow and the different materials needed to manage the mobility plan.
Through the RP1 implemented secondments, and the following outcomes have been achieved at M24
- Substantial progress in developing high-fidelity models and digital twins for electric vehicle powertrain components: development of accurate physics(based models and advanced computational methods for power sources, cooling systems and key powertrain components. Essential work to establish the modelling foundation for the predictive maintenance methodologies developed in later WPs.
- Implementation of a new sensing concept based on LC resonant sensors; design and implementation of a multimodal data structures and semantic backend dealing with the data model required or PdM of eV powertrains; design of layered digital architecture; incorporating edge computing, secure communication, cloud orchestration and enterprise integration.
- Development of physics-informed machine learning methods for different components of the eV such as batteries, fuel-cells and e-motors (battery aging diagnosis and prognosis, conditiong monitoring, degradation evaluation and prediction suitable for e-motors, anomaly detection and classification for different component of the powertrain).
- Preliminary specifications and design of the simulation platform that will integrate models + algorithls + data infrastructure developed across the different WPs; consolidation of the functional specifications of the integration framework with sepcific attention given to ensure compatibility with the semantic interopearbility developed in WP2 and the hybrid physics-informed/data driven approaches developed in WP3.
- Good progress towards establishing a LCA framewirk for eV powertrain components with a particular focus on enabling future evaluation of PdM strategies: creation of a robust LCA infrastructure.


Secondment implementation has resulted in the acceptance of more than a dozen of publications and others still in validation process.
The international secondments have strengthened the consortium's expertise across eV powertrain components and the collaboration to develop the predictive maintenance solution under its different facets, to be further rolled out.

Innowacyjność oraz oczekiwany potencjalny wpływ (w tym dotychczasowe znaczenie społeczno-gospodarcze i szersze implikacje społeczne projektu)

Progress has been strongest for battery modelling with:
- Major breakthroughs achieved in computational efficiency and physics-informed machine learning (Development of a Model-Integrated Neural Networks (MINN) that convert differential-algebraic equations to ordinary differential equations, enabling real-time battery digital twins with 182× speedup compared to traditional solvers, published in IEEE Transactions on Pattern Analysis and Machine Intelligence)
- Development of a the Heterogeneous-Optimal Temperature (HOT) framework for battery thermal characterization, introducing an effective temperature metric that enables lumped models to accurately predict cell resistance and overpotential, with journal papers published in 2024 and 202
- Contribution to the modelling of the battery thermal behaviours with equivalent circuit models and develoment of parametrization methods with AI for such models with field data from battery industry aimed at accurate aging estimation.
- Comprehensive review of high-frequency models and development of condition monitoring approach using GaN-based excitation to detect insulation aging through resonance frequency shifts, with two journal articles planned for 2026.
- The consortium has demonstrated significant advances beyond the state of the art in computational efficiency, novel thermal characterization methods, and physics-informed neural network architectures.

Development of an innovative sensing concept based on LC resonant factors deisgned to be integrated into the battery cells (sensitivity to temperature, pressure, electrolyte salt concentration and and electrolyte volume intercalation in polymer electrode materials).


Progress towards defining life-cycle inventories (LCIs) has been strongest for battery technologies with:
- the development of preliminary inventories for sodium-ion battery components, covering key materials, manufacturing steps and energy inputs. This work is to be presented at the Faraday Institution's ECR conference (UK) and the Advanced Battery Power conference (Germany) in the first part of 2026.