Experimental Data

Due to the lack of data currently available for fuel cell degradation and reliability analysis, key failure modes that currently have no data readily available in the literature were studied using the developed experimental set-up.

One of the failure modes that is mentioned numerous times throughout the literature is the issue surrounding BIP torque. However, no data was found in the literature that shows how changes in BIP torque affect the performance of the cell, and as such, is one of the failure modes where experimentation was undertaken.

Experimentation of how BIP Torque affects fuel cell performance

The fuel cell provided by Pragma industries had an associated specification sheet that showed the compression pressure felt in the cell due to a corresponding endplate torque (see Figure 6.19).

Figure 6.19: Look-up table of endplate bolt torques against MEA pressure [72]

However, there was no information available that shows how a change in pressure affects the

performance of the cell in general. Due to this lack of data, a set of experiments was undertaken to fill the knowledge gap, and use the data gained to inform the degradation model.

Polarisation curves were performed in-between tightening the endplate bolts from 2.5 Nm to 6.5 Nm in 1 Nm increments, and are shown in Figure 6.20.

As can be seen in Figure 6.20, from the initial endplate torque of 2.5 Nm, fuel cell perfor- mance increased with torque. This corresponds to an initial pressure on the MEA of around 0.4 MPa, and a final MEA compression pressure of around 2.0 MPa.

The MEAs used were unused and unconditioned, and as such, a large increase in performance was seen from initial to final torque levels due to the initial compression of the components of the MEA. When a new MEA is created, there is a large resistance involved due to the lack of interfacial contact between the core components. When the fuel cell is tightened as in this experiment, the contact patch is increased as the GDL material is crushed between the membrane and the BIPs. Interfacial contact resistance is therefore reduced through the increased endplate bolt torque.

After a period of conditioning of the new MEA, the fuel cell was stopped and the endplate bolts torques reduced back to the initial 2.5 Nm. The polarisation curves were then performed using the same torque levels up until 5.5 Nm as 6.5 Nm was at the very upper limit of the BIP’s mechanical integrity and as such was avoided. These results are shown in Figure 6.21.

0 10 20 30 40 50 60 70 80 90 0.4 0.5 0.6 0.7 0.8 0.9 1 Current (A) Voltage (V) 2.5 Nm 3.5 Nm 4.5 Nm 5.5 Nm 6.5 Nm

0 20 40 60 80 100 0.4 0.5 0.6 0.7 0.8 0.9 1

Polarisation curves at different endplate bolt torques

Current (A) Potential (V) 2.5 Nm 3.5 Nm 4.5 Nm 5.5 Nm

Figure 6.21: Polarisation curves for different endplate torques after conditioning

As can be seen in Figure 6.21, the fuel cell out-performs the pre-conditioning runs by a significant amount. Where the initial runs at 2.5 Nm had a peak current of 27.01 A, the latter run reached a peak current of 89.55 A. This significant gain in performance is partially due to the conditioning period for the MEA, however the newly compressed MEA has irreversible mechanical changes to its structure, and therefore the interfacial contact resistance is still reduced as when compressed to a higher level. It can be seen from Figure 6.21 that the lower torque level still performs badly compared to the higher increments.

EIS

EIS techniques were used to determine the resistance experienced under different endplate torques at 0.2 A/cm2_{, 0.4 A/cm}2_{and 0.6 A/cm}2_{. Galvanostatic EIS sweeps from 500 to 0.1 Hz}

were undertaken with an amplitude of 20mV. The results were compared and analysed to show resistance in the cell.

As can be seen in Figure 6.22, the EIS Nyquist plots were very successful, having little to no noise in the signal. A lot of useful information can be determined from Nyquist plots in relation to resistance.

Curve interpretation

The point at which the curve intersects the X axis at the left of the graph is indicative of the overall resistance experienced in the cell, and therefore the Ohmic losses. Therefore for 5.5 NM at 0.6 A/cm2_{, the resistance of the cell can be considered to be 0.2047 Ω − cm}2_{. The area}

of the Nyquist plot from the first intersect to the last is indicative of the anode and cathode activation losses and the mass transfer effects. In a perfect world there would be a small curve

0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 Z’ (Ω−cm2) Z’’ ( Ω −cm 2) 2.5 Nm 3.5 Nm 4.5 Nm 5.5 Nm (a) 0.2 A / cm2 0.2 0.25 0.3 0.35 0.4 0.45 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 Z’ (Ω−cm2) Z’’ ( Ω −cm 2) 2.5 Nm 3.5 Nm 4.5 Nm 5.5 Nm (b) 0.4 A / cm2 0.2 0.25 0.3 0.35 0.4 0.45 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 Z’ (Ω−cm2) Z’’ ( Ω −cm 2) 2.5 Nm 3.5 Nm 4.5 Nm 5.5 Nm (c) 0.6 A / cm2

Figure 6.22: Electrochemical Impedance Spectroscopy of different endplate bolt torques at dif- ferent current densities

for the anode activation losses, a larger curve for the cathode activation losses, and finally a smaller curve for the mass transport losses as in Figure 6.11. However, in practise EIS is more likely to produce either one large curve as in Figure 6.22a which shows the anode, cathode activation and mass transport losses combined, or one large curve and one smaller as in Figure 6.22c which shows the anode and cathode activation combined, and the mass transport region is clearly visible.

Extrapolation of degradation data

Data affecting the performance of the cell was extracted from the above experimentation anal- ysis of BIP torque in fuel cells. The overall resistance was taken from the x axis intercept for each current density and averaged to give the average resistance per torque level over the entire current range of the cell. The average resistance for 2.5 Nm was 0.249Ω − cm2, with average resistances of 0.234, 0.221 and 0.2145Ω − cm2_{for 3.5, 4.5 and 5.5 Nm torque levels respectively.}

These values were then plotted against the corresponding cell interior pressures with the data from Table 6.19.

Figure 6.23: RH and resistance equation derivation

Once the resistances for each pressure have been determined, the data can be plotted against data ascertained from the literature regarding operating conditions that would affect pressure; the relationship between water content in the membrane and pressure. A study by Serincan & Pasaogullari [73] detailed the relationship between RH across the cell and the pressure in the cell in their 2011 study. The pressures in the cell for 0%, 30% and 70% RH were taken from this work and plotted against the data obtained from the experimentation (see Figure 6.23) to give the relationship equation; −0.0009 ∗ RelativeHumidity + 0.264.

This can be used by the degradation model to show how an increase in cell pressure can increase cell resistance, thus affecting the Ohmic region of the polarisation curve, as in Figure 5.20.

6.5

Conclusions

The 1D model used for this work is an accurate mathematical representation of the chemical interactions of a FC, proved by the validation experimentation. The model was validated against an experimental rig, and the correlation between model and test data was very strong. Advanced techniques such as EIS were used to ascertain improved data over basic polarisation curves to look into resistances in the cell. The EIS data ascertained in this work, combined with numerous data sources from the literature regarding cell pressure and RH, mean that a highly accurate relationship between cell RH and performance can be implemented in the the Petri-net analysis. During operation, failure occurred that could be considered to match failure of other sources, this occurred when a gas inlet channel was damaged, partially blocking the gasses from entering the cell. This could be assumed to emulate the occurrence of an ice blockage in the channel, and as such, is used in the quantification of the Petri-net analysis.

Chapter 7

Degradation Simulation Results

Each of the failure rates, both proposed and taken from literature, were inserted into the Petri- Net model and ran both individually and as part of different operating modes to ascertain fuel cell lifetime. The entirety of the individual Petri-Net modules are omitted for brevity, however key examples are listed below. Additionally, the total results of the interactions of all Petri-Net modules are presented.

This chapter starts with two examples of what degradation logic occurs on an individual basis for two failure conditions. Startup Shutdown Cycling (SSC) results are presented that show where the degradation data comes from, and what the result is when only SSC is being observed in a FC. The degradation modelled through Petri-Net analysis is then validated by being run under the same conditions as a test undertaken on the FC experimental rig, with results shown. Finally, the Petri-Net model results are shown for the entire interaction of all the Petri-Net modules during operation under an automotive life cycle. Conclusions are then drawn in light of the results.