Back to search

NANO2021-Nanoteknologi og nye materiale

Semantic-based Material Twin and Co-Simulation Platform for Solid Oxide Fuel Cells (MEDIATE)

Alternative title: Semantisk basert materialutvikling og plattform for koplede simuleringer av fastoksid brennselceller

Awarded: NOK 5.0 mill.

The MEDIATE project has made significant progress in advancing the development of a comprehensive knowledge-based computational approach and platform for the modelling and design of electrochemical fuel cells, specifically Solid Oxide Fuel Cells (SOFCs). The project – which is worked on jointly by the Luxembourg Institute of Science and Technology (LIST), Technische Universität Dresden (TUD), SINTEF AS (SINTEF) and the Norwegian University of Science and Technology (NTNU) - successfully established a robust connection between microstructure and SOFC performance, achieving progress in generating large-scale Representative Volume Elements (RVEs) through digital reconstruction and tomography images. A multiscale electrochemical model was developed to simulate thermo-fluid cell behaviour, complemented by the seamless integration of a multi-field first-order computational homogenization framework into a finite element program by LIST and TUD. Semantic interoperability efforts have been directed toward extending the domain ontology for SOFC. SINTEF created data models for application interfaces and is developing the semantic interoperability framework, in particular the Data Space, which is the gateway to utilizing the entire semantic interoperability technology stack. All development pertaining to semantic interoperability is available open source on GitHub. Utilizing ProMo, the project has been generating mathematical descriptions of fundamental model building blocks, laying a solid foundation for semantic interoperability within the co-simulation platform. NTNU generated a knowledge graph containing the mathematical description of all basic entities. It contains the ontologised mathematical language and the variable/expression multi-bipartite graph. The first version is made available on GitHub. NTNU is working on a first revision, implementing a more straightforward mathematical language based on multi-linear algebra concepts. Dedicated to the development and use of data analytics, optimization, and machine learning models, the project created data-driven models for microstructure-property-performance. This includes the development and testing of a surrogate model for Finite Element (FE) simulations using Artificial Neural Networks (ANNs) through two distinct training approaches. Initially, a Conjugate Gradient (CG) optimization algorithm applicable for training networks using small datasets has been implemented and employed successfully. Secondly, considering the challenges posed by the involvement of the Hessian matrix, substantial memory requirements, and other limitations associated with the first approach, a larger dataset has been provided and modern Stochastic Gradient Descent (SGD)-based training algorithms such as Adam have been adopted. Ongoing work involves implementing a Bayesian optimization approach to identify uncertain parameters in the SOFC model.

The demand for energy generation has increased dramatically over the last few years, as well as the need to reduce its production impact on the environment. Fuel cell systems are one of the most promising technologies that can help achieve these objectives. Among the existing fuel cells systems, Solid Oxide Fuel Cells are promising technology that offers a clean alternative to fossil fuels due to their high kinetic activity, their fuel flexibility and their fuel reforming within the cell unit. The main characteristics of fuel cells are their lower noise, pollution emission and their higher energy conversion efficiency compared to most conventional thermomechanical-based power generation processes. We propose in this project to leverage the concept Digital Twin (DT) based approach and framework to support material microstructural design with a view to optimal design solid oxide fuel cells. This approach includes the development of a semantic-based interoperable material twin environment that factors in a wide range of multi-physics and multi-scale data sources and their underpinning semantic models, augmented with optimisation, data analytics and uncertainty management, to deliver an end-to-end material simulation, prediction and design capability. MEDIATE will develop standards and protocols for semantic interoperability for SOFC modelling that are extendable to other materials modelling applications through adherence to the EMMO. Scientists and technologies in the field of SOFCs and other fields employing multiscale modelling and multi-physical modelling of materials and processes will be able to use MEDIATE modules with much less detailed knowledge of the solvers.

Publications from Cristin

No publications found

No publications found

No publications found

No publications found

Funding scheme:

NANO2021-Nanoteknologi og nye materiale