Bayesian Active Learning for Determining Physical Parameters in SOFC Anode Degradation
Please login to view abstract download link
Solid-oxide Fuel Cells (SOFCs) emerge as leading contenders for replacing fossil fuels with renewable energy sources. Their high efficiency in generating clean electricity makes them suitable for powering large-scale systems such as industrial facilities and data centers. For long-term viability, it is crucial to understand the durability and lifespan of their components. The microstructural evolution of SOFC components during operation significantly impacts their electrochemical properties and overall performance. Multiphase-field simulation studies play a vital role in understanding the underlying microstructural changes and the resulting property alterations in SOFCs over time. However, these simulations face considerable challenges in identifying a suitable model and model parameterization. This research introduces an innovative framework that leverages Active Learning (AL) with Bayesian optimization to determine optimal parameter sets for phase-field models simulating the aging processes in Nickel-Gadolinium Doped Ceria (Ni-GDC) anodes. By combining experimental microstructural data with computational methods, our approach efficiently explores the high-dimensional material parameter space and finds unknown physical parameters. The methodology systematically navigates the complex material parameter space, evaluates simulation outcomes against experimental benchmarks, and iteratively refines the parameter selection process. This approach significantly reduces computational requirements while enhancing model accuracy. Our findings demonstrate that the AL-Bayesian framework not only improves the fidelity of phase-field simulations but also provides a versatile methodology applicable to broader SOFC development.
complas_sec@cimne.upc.edu/ Telf. + 34 - 93 405 46 94
Copyright © 2024 CIMNE, All Rights Reserved. Terms of service