Component Failure Analysis From a Stationary Proton Exchange Membrane Fuel Cell Demonstration

The durability and failure mechanisms of polymer electrolyte membranes are well known [1,2], and while a great deal of research effort focuses on fuel cell components such as membranes, electrocatalysts, and bipolar plates, the fuel cell power plant, as an entire system, presents its own set of challenges. Rastler noted how some hindrances of fuel cells in stationary applications are the distribution and interconnection issues [3]. The subsystems within fuel processing, power conditioning, and balance of plant have components that can degrade rapidly and affect the overall system efficiency [4], and these ancillary components account for a majority of the outages in the fuel cell power plant [5–7]. This study analyzes the components most frequently replaced during the demonstration of a fleet of PEM fuel cell power plants.

Since the early 1990 s, the U.S. Army Engineer Research and Development Center, Construction Engineering Research Laboratory (ERDC–CERL) has performed demonstrations of fuel cell systems. From fiscal years 2001–2004 (FY01–FY04), Congress appropriated funding to demonstrate domestically produced, residential-scale PEM fuel cells at federal facilities. The Department of Defense (DoD) maintains a large inventory of fixed facilities at its bases including office buildings, hospitals, industrial facilities barracks, and gymnasiums. Many of these facilities can benefit from distributed generation, in particular fuel cells, to augment their power, heat, reliability, and security requirements with an environmentally friendly system. This project sought to maximize diversity of location, environment, and applications [8–10], and also to provide a wealth of site-specific data [11]. The data were acquired from (and the conclusions apply to) a broad range of similar installed systems. The results from the overall demonstration supported the development of useful conclusions, despite a continuously changing, multifaceted field.

The fuel cell contract awardees were responsible for all siting and installation requirements, and for gathering and reporting performance data. One contract requirement, and primary goal of the demonstration, was to provide a minimum of 12 months of fuel cell power with at least 90% unit availability. Each of the systems was also covered under a comprehensive maintenance contract for the duration of the demonstration.

Information gathering and reporting matured throughout the project, allowing improvements in the data analysis. This study analyzes data based on demonstrations that contributed sufficient maintenance records to the overall body of knowledge. Because this analysis is based on real world demonstration data rather than information gathered from testing in a controlled laboratory setting, component replacement and reasoning varied greatly by each contractor. Part replacements were the result of normal maintenance needs, faulty components, extreme environmental conditions including lightning and natural disasters, or component exchanges performed due to small reductions in system performance. An important consideration, and possible source of error within this paper, is the root cause analysis. The fuel cell technician may not have always performed a root cause analysis on the fuel cell system each time a failure occurred; therefore the system analysis can be classified as a general overview of the most common causes of failures within the system. Nevertheless, this paper may assist a typical user with information on possible component failures and how they affect system availability. The results of this analysis are analyzed by fuel type and age to provide the most objective comparisons, and offer the greatest overlap of component types.

Table 1 summarizes the operational performance of the entire fleet of fuel cells operating through the course of the demonstration. A total of 91 fuel cell units from five different manufacturers were scheduled to be installed at 56 U.S. Government facilities in a variety of stationary applications, 81 of these fuel cells were installed and provided data. The first fuel cell unit from the FY01 fleet was installed in Jan. 2002, and the last fuel cell from the FY04 fleet was decommissioned in Oct. 2008. A majority of the fuel cells were sent back to the manufacturer for analysis following the 12-month demonstration, but several units were able to remain at the site for continued operation. Although the FY04 fleet showed a significant overall decrease in availability (68%), all of the backup units demonstrated during that year were able to meet the minimum requirement. The number of units that achieved each level of availability is shown in Fig. 1. The overall availability of the fleet reached 85%, which is slightly less than the minimum 90% availability requirement, but 58% of the units that provided data were able to reach that requirement.

Figure 2 shows the total component outages and average component lifetime for each of the fuel cell power plant subsystems. Possible sources of error would occur in the classification and categorization of each component. The Stack and Manifold system had the highest number of outages, while the Reformer Water Circuit had the fewest number of component replacements. A noteworthy trend is the general increase in average component lifetime as the number of actual component replacements decreased. With the exception of the Reformer Water Circuit and Heat and Water Management system (consisting primarily of a mixed bed water filter and steam filters and pumps, respectively), the remaining subsystem component lifetime averages tended to increase with decreasing replacements.

Table 2 and Fig. 3 detail the component outage statistics for the entire fleet of fuel cell units. A total of 719 component replacements were reported from the contractors. The results of the analysis indicate that the balance of plant components required the most component replacements for the fuel cell units, accounting for 60.6% of the total outages during the demonstration, which is similar to the results reported by Feitelberg [6]. The stack system replacements could have been lower, but as will be discussed in subsequent sections, the majority of the Stack and Manifold components replaced were removable fuel cell cartridges that were designed for periodic replacement.

Figure 4 shows the general trend of component replacements per fuel cell unit as the demonstration progressed from FY01 to FY04. The five most-replaced subsystems, as noted in sequential order in the chart from Table 2, ranged from 0.09 to 2.81 components replaced per fuel cell unit over the course of each 12-month demonstration. The increase in component replacements for the Stack and Manifold system in FY02 was due to the number of individual cartridges replaced at one of the sites, as described in subsequent sections of this paper. A positive and noteworthy trend is the overall decrease in components replaced per unit over the course of the demonstration. The range of component replacements in FY01 was 0.67–2.48, while in FY04 that range decreased to 0.09–1.09 component replacements per unit.

Figure 5 shows a more in-depth breakout of the components most frequently replaced for each system. One of the most glaring aspects of the figure is the values for the hydrogen cartridge replacements. Although the component replacement values appear to be high, and the average component life is low, this is not necessarily an indication of poor performance results. Eight separate sites in the demonstration reported on the frequency of the cartridge replacement. These units were unique because they did not contain a solid fuel cell stack, as the other demonstration units did, but rather an array of small fuel cell cartridges that could be individually replaced without disrupting operation of the other fuel cell cartridges. These systems were used for premium (backup) power and tracked the reliability from each start signal. The systems operated under programmed on/off sequences to mirror the start/stop requirements demanded from a premium power supply.

The data in Table 3 show a deeper level investigation of the component replacements for each subsystem with values on the minimum and maximum operational lifetime. This list does not encompass every component outage, but rather the frequently replaced items. The list is sorted in descending order from the subsystems with the highest frequency of outages. Beyond the Stack and Manifold system replacements, the next most common component replacement occurred with the various filters.

The Reverse Osmosis (RO), Carbon, Deionization (DI), Air-intake, and Mixed-bed water filters accounted for more than 24% (174) of the total component outages. With the exception of the Mixed-bed filter (part of the Reformer system), the rest of the filters were categorized as part of the balance of plant. Filters often required early replacement due to premature clogging, and it became apparent early in the demonstration phases that an alternative was required to allow proper purification throughout a system maintenance cycle.

Table 4 lists the number of replaced components in the Stack and Manifold subsystem for the various input fuels over the course of the demonstration. Natural gas-fueled units accounted for 56 of the site application demonstrations that reported component outages followed by propane (11), hydrogen cartridges (9), and hydrogen stacks (5). The data in Table 4 indicate a significant improvement in the number of required stack/cartridge replacement for each fuel cell unit as the demonstration progressed.

The data in Table 4 also indicate that the required number of hydrogen cartridge replacements improved greatly over the course of the demonstration. Although the replacements for the unit in FY01 were not officially counted, the contractor indicated that all cartridges were replaced multiple times during the 12-month demonstration. The unit demonstrated in FY02 reported 51 cartridge replacements. The units demonstrated in FY03 reported replacing an average of 5.6 cartridges per unit, while the FY04 units reported no replacements throughout the demonstration.

Similar results are evident with the natural gas and propane-fueled units. The natural gas values decreased from 1.16 to 0.67 replacements per unit from FY01–FY04. The propane units progressed in the order of 6, 0.33, 1, 0 for each fiscal year. The overall values for replacements per unit progressed as 1.3, 2.8, 1.6, 0.45; however, if the hydrogen cartridge replacements are excluded, the progression would have been 1.3, 0.9, 0.5, 0.45. This indicates significant improvement in fuel cell performance over the course of the demonstration.

Details on the minimum and maximum operational lifetimes of the stacks/cartridges as well as the average lifetime and standard deviation are listed in Table 5. The natural gas and propane units had maximum operational lifetime values of 10,250 and 8300, respectively; the average lifetimes were 3056 and 2571, respectively. The systems in these applications were most often replaced when they initially begin to decrease in performance. They were replaced under a planned shutdown to avoid a catastrophic system failure. The technicians remotely monitored the systems in the field to prevent system failures by predicting and detecting component failures. This observation system helped determine stack failures by the degradation of individual cell voltages over time; therefore, the fuel cell systems in the demonstration never reached total failure in the field.

The standard deviation of the average operational lifetime for the natural gas and propane-fueled units was 2582 and 2194 h, respectively. This indicated that the average values varied greatly from each unit. Figure 6 shows the breakout of the average operational lifetime of the natural gas and propane fuel cell stacks, while Fig. 7 shows the operational lifetime of the hydrogen cartridges.

For the natural gas units, almost 50% operated between 0 and 2000 h before needing to be replaced, while overall, 83.3% were able to operate between 0 and 7000 h. Seven units operated 7000–8000 h, while the stack in the unit installed at Fort Jackson, SC surpassed the 12-month demonstration requirement and operated for 10,250 h before requiring replacement. The units with propane-fueled stacks followed a similar trend as the natural gas units.

One hydrogen cartridge reached 2139 h of operation, but with that outlier removed, the average operational lifetime of the remaining units was 224 h. In the early part of the demonstration, the cartridge replacements were caused by a low-level hydrogen-leak signal. The time shown for these components is equal to the hours the fuel cells actually operated. This is notable because these fuel cells operated typically for 1–2 h per day.

It is important to note there were several factors that affected availability other than the fuel cell components themselves, as has been the case with many past projects as well [12]. Each demonstration provided an opportunity for the fuel cell manufacturers to make modifications to improve their systems. Many challenges arose from technical failures of surrounding systems, including faulty space heaters that caused parts of the fuel cells to freeze and water damage caused by a leaky roof directly over a unit housed indoors. There were also some bureaucratic challenges with the installation and operation of the systems, which included fuel supply interruptions and security issues that prevented the installation of high-speed data lines.

Another significant cause of failures was the varying climates and operating conditions at the installation sites. On coastal sites such as Hawaii and Iceland, there were challenges with high humidity and corrosion. In Arizona, the summer heat led to cooling issues that required the radiator fans to be replaced. The high elevation of the site in Colorado led to difficulties with high wind and air intake blowers. There were also issues with air quality that caused outages. The Fort Knox site, for example, had chlorine in the air because of its proximity to a pool facility, requiring additional chemical filters to be added. A large amount of airborne particulates at another site led to premature clogging of the inlet filters. All of these challenges led to changes by the fuel cell manufacturers to adapt to the various environments.

PEM fuel cell design and manufacturing has greatly evolved. This evaluation provided a review of possible product enhancements to increase system availability and reliability. Many of the component system failures were replaced with improved parts to increase the overall system availability. These more robust versions were able to provide increased availability and reliability over many of the systems demonstrated.

The authors would like to thank Dr. Michael Binder (retired), Frank Holcomb, and the rest of the team from ERDC-CERL who have provided support of this demonstration throughout the years. The views expressed in this paper are those of the authors and do not reflect the official policy or position of the Department of the Army, the Department of Defense, or the U.S. Government.