Hydrogen Compression Reliability: Where the Downtime Actually Comes From

A hydrogen compressor fails on its seals more often than on anything else, and the numbers are published. In a national-laboratory test of a diaphragm machine on a fuelling duty, 753 operating hours and 81 start-stop cycles produced five seal failures — four in the compressor head, one on the second-stage discharge check valve — at a mean time between failures of 49 days. The same programme found the compressor first among station components for both downtime and maintenance event count, accounting for about one third of station maintenance hours.

What follows is the mechanism behind those numbers, the detection gap that lets a seal leak become a seal failure, and the specification lines a buyer can use to act on both.

Where the problem sits

In the station reliability work published by the US Department of Energy's hydrogen programme, the compressor is first among station components for downtime and first for maintenance event count, and accounts for about one third of station maintenance hours. That is a statement about the station as a whole: the compressor is the one high-pressure, cycling, moving assembly in the fuelling path, and it carries the reliability load for everything downstream of it.

The same programme published the test record behind the finding. A commercially available diaphragm compressor — the family usually reached for at the top of a hydrogen pressure range, because its gas path is closed by a metallic diaphragm rather than by a sliding seal — was run at duty cycles of up to 20 hours a day, compressing from about 100 psi suction to discharge pressures between 3,000 and 6,000 psi, for roughly 800 hours in total. The recorded figures:

Two of those lines matter more than the rest. The efficiency figure is the one that gets quoted in cost discussions. For reliability, the pair that matters is the cycle count against the failure count — 81 starts and 49 days between failures is a machine whose duty pattern is part of its failure record, not only its selection case.

The mechanism: seals

The second finding of the same work was that seal weakness is the main failure mechanism. Five seal failures occurred during the test — four in the compressor head and one on the check valve on the second-stage discharge — and the report notes that the mechanism is consistent with field data on compressor failures held by a separate station data-collection programme.

Parallel work on the same machine family identified what drives it. Contamination and debris in the hydrogen, and improper priming procedures when restarting after a stop, were identified as two important factors reducing diaphragm life. A finite-element model of the diaphragm showed why: at operating pressure, a hard particle trapped against the head profile produces local plastic deformation and residual stress, and that residual stress reduces fatigue life. The model's predicted failure location matched where the failures are observed in service. A second contamination route was found on the machine itself: a non-native coating on the interior wall of the process tubing was sampled and identified as a siloxane — a compound present in some commercial vacuum greases and harmful to fuel cells — and the recommendation was a non-siloxane lubricant on the compressor seals.

The practical reading is that in this duty the seal is not only a wear item. It is the part that concentrates every other defect the duty produces: particles from the gas, lubricant from assembly, and the pressure and temperature excursions of the start-stop cycle.

What a seal failure costs

The same programme recorded the repair burden for each failure type, and the distribution is not what most buyers expect:

The wrench time is small and the parts cost is small. The lead time is not. Six weeks for a first- or second-stage seal, against a mean time between failures of 49 days, means a site that orders the seal when the failure is detected will be down for the length of the delivery. The spare belongs on the shelf before the failure, not in a purchase requisition after it. Two of the four repairs also need two people and a hoist, which makes lifting provision and access a line on the specification rather than a matter for the maintenance planner.

The detection gap

The fourth finding is the one a buyer can act on without changing the machine: significant downtime can be avoided by monitoring the leak-detection circuit. Compressor manufacturers typically set a limit on the pressure allowed in that circuit, and pressures below that limit — commonly 15 psi — raise no alarm for the operator, yet can be a precursor to seal failure.

That is the whole gap in one sentence. A leak-detection circuit built to a threshold is a failure annunciator rather than a condition monitor: below the threshold the machine runs through the precursor stage silently, and the operator first learns of the leak at the point where the seal has already failed. The specification lines that follow are therefore not "is there a leak alarm" but three others — where the threshold is set, what the machine does with a reading below it, and whether the sub-threshold pressure is trended and logged rather than only compared against a trip value.

One further recommendation from the same work belongs on the same list. The overpump valve setting should be monitored, because the setting can change over time, and a drifted setting puts pressure or temperature in the compressor head above where it should be, which damages seals. It can be checked at the valve's test port with a check valve and a gauge.

What the standards already fix for hydrogen

The findings above are about what fails. The standards are about the variables that were left free — and for hydrogen, three of them are not left free at all.

The discharge temperature. API 618 sets the maximum predicted discharge temperature at 150 °C (300 °F) generally, and then narrows it for this service: predicted discharge temperatures shall not exceed 135 °C (275 °F) for hydrogen-rich services, defined as molar mass less than or equal to 12. The standard's own note explains the reason — non-lubricated hydrogen services generally run at higher discharge temperatures than lubricated ones, because of internal leakage and because hydrogen can release heat when it expands, which is unlike the behaviour of the heavier gases in this category. The Chinese petrochemical engineering specification for reciprocating machines carries the same 135 °C for rich hydrogen at all specified conditions and loads, alongside 135 °C at the normal operating point and 150 °C elsewhere for oxygen-bearing gas in an oil-lubricated cylinder, and 130 °C for non-lubricated service at a gas pressure of 7 MPa(G) or above.

Alarm and trip. Both documents make the temperature observable rather than merely designed. API 618 recommends the discharge-temperature alarm and trip set points at +20 K (40 °F) and +30 K (50 °F) above the maximum predicted discharge temperature, caps the trip set point — 180 °C in the fifth edition, reduced to 175 °C in the sixth — and adds that lower limits should be considered for air, because of its oxygen content, above 20 bar(g) discharge. The Chinese specification sets its high alarm at +10 °C above the maximum expected discharge temperature and its trip, or high-high alarm, at +20 °C, and requires a separate measuring point at each cylinder's discharge outlet, as close as practicable to the discharge valve.

Speed. The same standard runs the argument the other way for oil-free machines: in general, their rotating speed and piston speed should be less than those of equivalent lubricated machines. Oil-free construction in this service is bought with speed, and the lower speed is the condition that makes the non-metallic parts last.

Three consequences are worth stating plainly. 135 °C is not a screening convention — it is the hydrogen-rich limit in the standard, which is why a duty in hydrogen-rich gas sits 15 K below the general limit for the same machine. The limit is also a reliability instrument rather than a thermal ceiling, because what it protects is the non-metallic part set whose life falls as temperature rises, and the alarm structure above it is what turns a design figure into something an operator can see. And the sixth edition of API 618 moves the general limit to 135 °C and the hydrogen-rich or non-lubricated limit to 120 °C, with the trip cap at 175 °C, so the direction of travel is tighter rather than looser.

What to put in the specification

Eight lines, in the order in which they change the answer:

1. The duty pattern, as a cycle count. Starts per day and a maximum per year, not "continuous" or "intermittent". The test record above is 81 starts against 49 days between failures.
2. Oil-free or lubricated. The two carry different discharge-temperature limits and different speed expectations, so the answer decides which limit applies.
3. The discharge temperature limit, by clause. Name the clause and the figure that applies to the duty, and state the gas's molecular weight, because that is what selects between them.
4. The alarm and trip structure, and the thermowell. Set points relative to the maximum predicted discharge temperature, and a measuring point at each cylinder close to the discharge valve.
5. The leak-detection threshold, and what happens below it. The trip value, whether sub-threshold pressure is trended and logged, and what the operator sees before the trip.
6. The lubricant. Non-siloxane, and compatible both with the seals and with whatever consumes the hydrogen downstream.
7. Gas cleanliness ahead of the machine. Filtration and knockout, with a particle size, because a trapped particle is a fatigue site.
8. Spare parts and access. The seal lead times above, and the space and lifting provision for a two-person repair.

A duty that combines more than four stages, an estimated discharge temperature above the applicable limit, or a gas carrying solids goes to engineering review rather than an automatic answer — which in this category is the normal route, not a rejection.

What this article does not cover

Note that this article does not select a compressor, size or specify a seal or a diaphragm, calculate stage count, power or efficiency, or give prices or maintenance intervals. It summarises published reliability findings for hydrogen compression and the standard clauses those findings interact with, and it cites only clauses whose text was read directly.

Related on this site

• Discharge Temperature: the 135°C Line, and Why Some Machines Get No Number
• Reciprocating or Diaphragm for High-Pressure Hydrogen?
• Diaphragm vs Piston vs Screw Compressors: A Comparison Table
• Specialty Gas Compressor Selection Guide: From Gas to Machine Type
• Flare Gas and Sour Gas Compression: What Changes
• Run a duty point: gascompressorselect.com

Sources