The excessive presence of liquid water in a gas diffusion layer (GDL) hinders the access of reactant gases to the active sites of the catalyst layer leading to decreased performance of a polymer electrolyte membrane fuel cell (PEMFC). Therefore, GDLs are usually treated with a hydrophobic agent to render their fibres more hydrophobic in order to facilitate gas transport and water removal. Numerous studies have been conducted to investigate water transport in PEMFCs in recent years; however, the behaviour of liquid water in a GDL at a pore-level is poorly understood. Macroscopic models fail to incorporate the influence of the structural morphology of GDLs on liquid water transport behaviour. Experimental methods are not conducive towards a good understanding at a microscopic level because of the diminutive size of the GDLs porous structure. Alternatively, the Lattice Boltzmann (LB) method has gathered interest as it is found to be particularly useful in fluid flow simulations in porous media due to its capability to incorporate the complex boundaries of actual GDL structures. To date, most studies on fluid transport in GDLs integrated artificial structures generated by stochastic simulation techniques to the LB models. The stochastic-based model, however, does not represent closely the microscopic features of the actual GDL as manufactured. In addition, comparison of liquid water transport behaviour in different GDL structures using the LB method is rare since only a single GDL material has been utilised in most of those studies. This thesis aims to develop our understanding of liquid water transport behaviour in GDLs with morphologically different structures under varying wettability conditions based on the LB method and the X-ray computed tomography (XCT) technique. GDLs with paper and felt structures were reconstructed into 3D digital volumetric models via the XCT process. The digital models were then incorporated into a LB solver to model water saturation distribution through the GDL domains. The GDL wettability was also altered so that the effect on liquid water behaviour in the GDL could be examined. This project is divided into three main sections. In the sensitivity analysis, the effect of image resolution on gas permeability through the X-ray reconstructed GDL was carried out using a single-phase LB model. It was found that the resolution variation could significantly affect the resulting gas permeability in both principal and off-principal directions, as well as computational time. An optimum resolution, however, exists at 2.72 μm/pixel, which consumed 400 times less computational time with less than 8% difference in the resulting permeability compared to the base resolution. This study also served as a guideline for selecting a resolution for generating the XCT images of the GDLs which were utilised in the following studies. In the structure analysis, the structures of the paper and felt GDLs were generated using the XCT and the key properties of each GDL, including thickness, porosity, permeability and tortuosity, were characterised. The thickness and the through-plane porosity distributions of each GDL were examined based on the tomography images. The resulting local through-plane porosity distributions were then used to calculate through-plane permeability and tortuosity distributions using an analytical model available in the literature. This study revealed the heterogeneity of the GDLs and how the heterogeneous nature of the GDL structures affects others properties of the GDLs. In this study, the absolute through-plane permeability and tortuosity of the X-ray-reconstructed GDL samples were also characterised using the single-phase LB model. The results from the two models were then compared and validated against data in the literature. In the water transport analysis, the two-phase LB model was employed to examine the effects of GDL structures on the behaviour of liquid water in the GDLs, including invasion patterns, saturation distribution and breakthrough behaviour under varying GDL wettability conditions. It was found that wettability was responsible for invasion patterns and water saturation levels whilst the GDL structure was mostly responsible for breakthrough occurrence and saturation distribution. It was observed that water travelled with stable displacement saturating all pores in hydrophilic GDLs, while it travelled with capillary fingering causing decreased saturation in hydrophobic GDLs, about 50% in the highly hydrophobic cases. The GDL structure was found to play a key role in breakthrough behaviour in the hydrophilic GDL as it was seen that the through-plane fibres in the felt structure and the through-plane binders in the paper structure encouraged water removal from the GDL in the thickness direction. Conversely, the GDL structure was found to have negligible influence on breakthrough in the hydrophobic GDL. Each GDL structure, however, contributed to a distinct difference in water distribution in the GDL with hydrophobic wettability. The work presented in this thesis contributes to the understanding of liquid water transport behaviour in the GDLs under the combined effects of the GDL structures and wettability conditions, which is essential for the development of effective PEMFC water management and the design of future GDL materials.

Due to escalating energy costs and limited fossil fuel resources, much attention has been given to polymer electrolyte membrane (PEM) fuel cells. Gas diffusion layers (GDLs) play a vital role in a fuel cell such as (1) water removal, (2) cooling, (3) structural backing, (4) electrical conduction and (5) transporting gases towards the active catalyst sites where the reactions take place. The power density of a PEM fuel cell in part is dependent upon how uniform the gases are distributed to the active sites. To this end, research is being conducted to understand the mechanisms that influence gas distribution across the fuel cell. Emerging PEM fuel cell designs have shown that higher power density can be achieved; however this requires significant changes to existing components, particularly the GDL. For instance, some emerging concepts require higher through-plane gas permeability than in-plane gas permeability (i.e., anisotropic resistance) which is contrary to conventional GDLs (e.g., carbon paper and carbon cloth), to obtain a uniform gas distribution across the active sites. This is the foundation on which this thesis is centered. A numerical study is conducted in order to investigate the effect of the gas permeability profile on the expected current density in the catalyst layer. An experimental study is done to characterize the effects of the weave structure on gas permeability in woven GDLs. Numerical simulations are developed using Fluent version 6.3.26 and COMSOL Multiphysics version 3.5 to create an anisotropic resistance profile in the unconventional GDL, while maintaining similar performance to conventional GDL designs. The effects of (1) changing the permeability profile in the in-plane and through-plane direction, (2) changing the thickness of the unconventional GDL and (3) changing the gas stoichiometry on the current density and pressure drop through the unconventional GDL are investigated. It is found that the permeability profile and thickness of the unconventional GDL have a minimal effect on the average current density and current density distribution. As a tradeoff, an unconventional GDL with a lower permeability will exhibit a higher pressure drop. Once the fuel cell has a sufficient amount of oxygen to sustain reactions, the gas stoichiometry has a minimal effect on increases in performance. Woven GDL samples with varying tightness and weave patterns are made on a hand loom, and their in-plane and through-plane permeability are measured using in-house test equipment. The porosity of the samples is measured using mercury intrusion porosimetry. It is found that the in-plane permeability is higher than the through-plane permeability for all weave patterns tested, except for the twill weave with 8 tows/cm in the warp direction and 4 tows/cm in the weft direction, which exhibited a through-plane permeability which was 20% higher than the in-plane permeability. It is also concluded that the permeability of twill woven fabrics is higher than the permeability of plain woven fabrics, and that the percentage of macropores, ranging in size from 50-400 µm, is a driving force in determining the through-plane permeability of a woven GDL. From these studies, it was found that the graduated permeability profile in the unconventional GDL had a minimal effect on gas flow. However, a graduated permeability may have an impact on liquid water transport. In addition, it was found that graduating the catalyst loading, thereby employing a non-uniform catalyst loading has a greater effect on creating a uniform current density than graduating the permeability profile.

The gas diffusion layer (GDL) is one of the key components in a polymer electrolyte membrane (PEM) fuel cell. It performs several functions including the transport of reactant gases and product water to and from the catalyst layer, conduction of both electrons and heat produced in the catalyst layer, as well as mechanical support for the membrane. The overarching goal of this work is to thoroughly examine the GDL structure and properties for use in PEM fuel cells, and more specifically, to determine how to characterize the GDL experimentally ex-situ, to understand its performance in-situ, and to relate theory to performance through controlled experimentation. Thus, the impact of readily measured effective water vapor diffusivity on the performance of the GDL is investigated and shown to correlate to the wet limiting current density, as a surrogate of the oxygen diffusivity to which it is more directly related. The influence of microporous layer (MPL) design and construction on the fuel cell performance is studied and recommendations are made for optimal MPL designs for different operating conditions. A method for modifying the PTFE (Teflon) distribution within the GDL is proposed and the impact of distribution of PTFE in the GDL on fuel cell performance is studied. A method for characterizing the surface roughness of the GDL is developed and the impact of surface roughness on various ex-situ GDL properties is investigated. Finally, a detailed analysis of the physical structure and permeability of the GDL is provided and a theoretical model is proposed to predict both dry and wet gas flow within a GDL based on mercury intrusion porosimetry and porometry data. It is hoped that this work will contribute to an improved understanding of the functioning and structure of the GDL and hence advance PEM fuel cell technology.

Aufgrund des weltweit steigenden Energiebedarfs, dessen Deckung derzeit größtenteils auf fossilen Brennstoffen basiert, ist es nötig geworden, die Entwicklung alternativer Möglichkeiten zur Erzeugung von Elektroenergie als Primärenergie voranzutreiben. Eine dieser alternativen Möglichkeiten ist die Brennstoffezellentechnologie, welche sowohl in stationären als auch mobilen Anwendungen zum Einsatz kommen kann. Ihrer weitreichenden Verbreitung stehen bislang die aufgrund des großen Fertigungsaufwandes hohen Herstellungskosten der benötigten Komponenten im Wege. Hierzu zählen die Gasdiffusionslagen des weit verbreiteten Typs der wasserstoffbetriebenen Polymerelektrolytbrennstoffzelle. Es treten zwischen den einzelnen Fertigungsschritten im Herstellungsprozess dieser Gasdiffusionslagen Wechselwirkungen auf, die zu unerwünschten Materialveränderungen führen. Die Ursachen dieser Wechselwirkungen sind nicht vollends verstanden. Eine Vertiefung des Verständnisses der Herstellungsprozesse soll die Grundlage für eine Optimierung der Prozessführung bilden. Es sollen eine Kostenreduktion sowie eine Leistungssteigerung der Gasdiffusionslagen ermöglicht werden.:1 Einleitung 1 2 Stand der Technik 5 2.1 Brennstoffzellen 5 2.2 Gasdiffusionslagen 11 3 Problemstellung und Zielsetzung 17 4 Analyse und Klassifizierung von GDL-Fehlern 20 4.1 Fehlerklassifizierung 22 4.2 Fehleridentifizierung 26 4.3 Auswahl zu analysierender Fehlerbilder 27 4.4 Charakteristika der ausgewählten Fehlerbilder 42 4.4.1 Bahndeformationen 42 4.4.2 Umlaufende Verdickungen von Wickeln in Umfangsrichtung 44 4.4.3 Längs- und Queraufrauhungen sowie Rauhspuren 45 5 Theoretische Grundlagen 49 5.1 Physikalische und mechanische Grundlagen 49 5.1.1 Zug-, Biege- und Druckspannungen in Warenbahnen 49 5.1.2 Elastizitäts- und Kompressionsmoduli 52 5.1.3 Elastizität und Plastizität 53 5.1.4 Umformmechanismen im GDL-Basisvliesstoff und Versagensarten von Fasern 54 5.2 Statistik 55 5.2.1 Korrelationsanalyse 55 5.2.2 Regressionsanalyse 56 5.2.3 Zweistichproben-t-Tests und Konfidenzintervalle 56 5.2.4 Stichprobenumfang 57 5.3 Qualitätsregelkreise 58 6 Eingesetzte Untersuchungsmethoden 60 6.1 Mechanische Eigenschaften 64 6.1.1 Höchstzugkraft und Höchstzugkraftdehnung 64 6.1.2 Elastizitätsmodul und Kompressibilität 66 6.1.3 Elastische und plastische Deformation bei Zugbelastungen 67 6.1.4 Flächenmasse 70 6.1.5 Biegesteifigkeit 72 6.1.6 Dickenmessung 74 6.2 Thermische Eigenschaften 75 6.2.1 Wärmeleitfähigkeit 75 6.3 Bildgebende Verfahren 78 6.3.1 Schliffbildmikroskopie 78 6.3.2 Rasterelektronenmikroskopie 78 6.3.3 µ-Computertomographie 79 7 Herstellungsverfahren der untersuchten Gasdiffusionslagen im Überblick 81 8 Basisvliesstoffherstellung 84 8.1 Prozess der Vliesbildung und Verfestigung 84 8.2 Charakterisierung des GDL-Basisvliesstoffes 90 8.3 Fehlerbilder des GDL-Basisvliesstoffes 103 9 Dickenkalibrierung 113 9.1 Prozess der Dickenkalibrierung des GDL-Basisvliesstoffes 113 9.2 Charakterisierung des dickenkalibrierten GDL-Basisvliesstoffes 120 9.3 Fehlerbilder des dickenkalibrierten GDL-Basisvliesstoffes 130 9.3.1 Prozessbeobachtung 130 9.3.2 Hypothesenbildung und Verifikation 135 9.3.3 Maßnahmen zur Fehlervermeidung 146 10 Carbonisierung 156 10.1 Prozess der Carbonisierung 156 10.2 Charakterisierung carbonisierten GDL-Substrates 157 10.3 Fehlerbilder im Carbonisierprozess 163 11 Data Mining für die GDL-Herstellung 167 11.1 Datenerhebung 167 11.2 Auszuwertende Parameter 172 11.3 Ergebnisse der Parameteranalysen 173 12 Qualitätsregelkreise zum GDL-Produktionsprozess 178 12.1 Wulstbildung und Längsaufrauhung 178 12.2 Queraufrauhung 181 13 Zusammenfassung und Ausblick 184 14 Literaturverzeichnis 186 15 Abbildungsverzeichnis 192 16 Abkürzungsverzeichnis 201 17 Formelverzeichnis 203 18 Anlagenverzeichnis 204 / Due to worldwide increasing energy consumption, which is mainly covered by fossile fuels nowadays, it has become a necessity to further develop alternative possibilities to create electricity as primary energy. One alternative technology to accomplish this is fuel cell technology which can be used in stationary as well as in mobile applications. One aspect hindering its widespread use is the high manufacturing cost of the needed components due to the complicated production processes. Among these are gad diffusion layers of the commonly used hydrogen-driven polymer electrolyte fuel cells. There are interactions occurring between the several production steps leading to unwanted changes in material properties. The causes of these interactions are not completely understood. A deeper understanding of these shall be the basis for optimizations in process design and therefore cost reductions and improvements in performance of gas diffusion layers can be achieved.:1 Einleitung 1 2 Stand der Technik 5 2.1 Brennstoffzellen 5 2.2 Gasdiffusionslagen 11 3 Problemstellung und Zielsetzung 17 4 Analyse und Klassifizierung von GDL-Fehlern 20 4.1 Fehlerklassifizierung 22 4.2 Fehleridentifizierung 26 4.3 Auswahl zu analysierender Fehlerbilder 27 4.4 Charakteristika der ausgewählten Fehlerbilder 42 4.4.1 Bahndeformationen 42 4.4.2 Umlaufende Verdickungen von Wickeln in Umfangsrichtung 44 4.4.3 Längs- und Queraufrauhungen sowie Rauhspuren 45 5 Theoretische Grundlagen 49 5.1 Physikalische und mechanische Grundlagen 49 5.1.1 Zug-, Biege- und Druckspannungen in Warenbahnen 49 5.1.2 Elastizitäts- und Kompressionsmoduli 52 5.1.3 Elastizität und Plastizität 53 5.1.4 Umformmechanismen im GDL-Basisvliesstoff und Versagensarten von Fasern 54 5.2 Statistik 55 5.2.1 Korrelationsanalyse 55 5.2.2 Regressionsanalyse 56 5.2.3 Zweistichproben-t-Tests und Konfidenzintervalle 56 5.2.4 Stichprobenumfang 57 5.3 Qualitätsregelkreise 58 6 Eingesetzte Untersuchungsmethoden 60 6.1 Mechanische Eigenschaften 64 6.1.1 Höchstzugkraft und Höchstzugkraftdehnung 64 6.1.2 Elastizitätsmodul und Kompressibilität 66 6.1.3 Elastische und plastische Deformation bei Zugbelastungen 67 6.1.4 Flächenmasse 70 6.1.5 Biegesteifigkeit 72 6.1.6 Dickenmessung 74 6.2 Thermische Eigenschaften 75 6.2.1 Wärmeleitfähigkeit 75 6.3 Bildgebende Verfahren 78 6.3.1 Schliffbildmikroskopie 78 6.3.2 Rasterelektronenmikroskopie 78 6.3.3 µ-Computertomographie 79 7 Herstellungsverfahren der untersuchten Gasdiffusionslagen im Überblick 81 8 Basisvliesstoffherstellung 84 8.1 Prozess der Vliesbildung und Verfestigung 84 8.2 Charakterisierung des GDL-Basisvliesstoffes 90 8.3 Fehlerbilder des GDL-Basisvliesstoffes 103 9 Dickenkalibrierung 113 9.1 Prozess der Dickenkalibrierung des GDL-Basisvliesstoffes 113 9.2 Charakterisierung des dickenkalibrierten GDL-Basisvliesstoffes 120 9.3 Fehlerbilder des dickenkalibrierten GDL-Basisvliesstoffes 130 9.3.1 Prozessbeobachtung 130 9.3.2 Hypothesenbildung und Verifikation 135 9.3.3 Maßnahmen zur Fehlervermeidung 146 10 Carbonisierung 156 10.1 Prozess der Carbonisierung 156 10.2 Charakterisierung carbonisierten GDL-Substrates 157 10.3 Fehlerbilder im Carbonisierprozess 163 11 Data Mining für die GDL-Herstellung 167 11.1 Datenerhebung 167 11.2 Auszuwertende Parameter 172 11.3 Ergebnisse der Parameteranalysen 173 12 Qualitätsregelkreise zum GDL-Produktionsprozess 178 12.1 Wulstbildung und Längsaufrauhung 178 12.2 Queraufrauhung 181 13 Zusammenfassung und Ausblick 184 14 Literaturverzeichnis 186 15 Abbildungsverzeichnis 192 16 Abkürzungsverzeichnis 201 17 Formelverzeichnis 203 18 Anlagenverzeichnis 204