Kombination elektromagnetischer Feldsimulation und genetischer Algorithmen zur Designoptimierung von Brennstoffzellen und Modenverwirbelungskammern
Publication date
2026-02-04
Document type
Dissertation
Author
Zazai, Mohammad Farhad
Advisor
Referee
Granting institution
Helmut-Schmidt-Universität/Universität der Bundeswehr Hamburg
Exam date
2025-11-26
Organisational unit
Publisher
Universitätsbibliothek der HSU/UniBw H
Part of the university bibliography
✅
File(s)
Language
German
Abstract
This dissertation investigates simulation-based design optimization using genetic algorithms (GAs) in combination with finite-element method (FEM) simulations and demonstrates its applicability in two deliberately distinct domains: proton exchange membrane (PEM) fuel cells and electromagnetic reverberation (mode-stirred) chambers.
For fuel cells, numerous parameters, from membrane thickness to operating temperature, are tuned to reproduce a target power-density curve rather than maximize absolute power. A white-box, first-principles model is coupled with a GA to vary material, geometric, and operating parameters until the best match is obtained, revealing a family of parameter combinations consistent with measured performance and informing cell sizing and operating conditions.
The second application addresses field uniformity in electromagnetic compatibility (EMC) testing. Mode-stirred chambers generate field ensembles by varying stirrer geometry; homogeneity is assessed from distributions of spatially averaged field values. A gray-box model (physics-based core with empirical corrections and stochastic losses) is coupled with a GA; the Shapiro-Wilk test quantifies departures from normality, and p-values are maximized to identify geometry parameters that yield more uniform fields and support reproducible EMC testing.
Across both examples, the GA–FEM coupling efficiently explores high-dimensional, nonlinear design spaces, revealing multiple practically equivalent solutions. This provides manufacturing and cost flexibility for fuel cells and improved field uniformity for reverberation chambers, offering a transferable blueprint for GA-driven optimization in complex, coupled systems.
For fuel cells, numerous parameters, from membrane thickness to operating temperature, are tuned to reproduce a target power-density curve rather than maximize absolute power. A white-box, first-principles model is coupled with a GA to vary material, geometric, and operating parameters until the best match is obtained, revealing a family of parameter combinations consistent with measured performance and informing cell sizing and operating conditions.
The second application addresses field uniformity in electromagnetic compatibility (EMC) testing. Mode-stirred chambers generate field ensembles by varying stirrer geometry; homogeneity is assessed from distributions of spatially averaged field values. A gray-box model (physics-based core with empirical corrections and stochastic losses) is coupled with a GA; the Shapiro-Wilk test quantifies departures from normality, and p-values are maximized to identify geometry parameters that yield more uniform fields and support reproducible EMC testing.
Across both examples, the GA–FEM coupling efficiently explores high-dimensional, nonlinear design spaces, revealing multiple practically equivalent solutions. This provides manufacturing and cost flexibility for fuel cells and improved field uniformity for reverberation chambers, offering a transferable blueprint for GA-driven optimization in complex, coupled systems.
Version
Published version
Access right on openHSU
Open access