Skip to main content

On-site oxygen generation for hospitals: sizing, purity and the honest case against cylinders

How PSA oxygen plants work, how to size one against a real clinical load rather than a pandemic peak, what to monitor continuously, and when cylinders or liquid oxygen remain the better answer for a Bangladeshi hospital.

Any hospital that ran a full oxygen ward through the pandemic learned the difference between having oxygen and having a delivery scheduled. Since then a great many Bangladeshi hospitals have installed, or been offered, a pressure swing adsorption oxygen plant. Some of those plants are well sized and running. Others sit idle because they were sized against a peak that will not return, or because the plant room has no ventilation, or because nobody budgeted for the sieve. The technology is not the difficult part. The sizing, the monitoring and the redundancy are.

What a PSA plant actually does

Air is roughly 78 per cent nitrogen, 21 per cent oxygen and about 1 per cent argon. A PSA plant compresses and dries that air and passes it through a vessel packed with a zeolite molecular sieve, which adsorbs nitrogen preferentially at pressure and lets oxygen through. When the sieve approaches saturation the vessel is depressurised, the nitrogen is released to atmosphere and the vessel is ready again. Two or more vessels alternate so that product flows continuously, which is where the twin tower arrangement in Oxymat's range comes from.

The argon is the part worth understanding, because it explains a question every hospital engineer eventually asks. The sieve separates nitrogen from oxygen, but argon behaves enough like oxygen that it passes through with the product. That is why concentrator oxygen sits in the low nineties per cent rather than above ninety-nine, and it is not a defect or a sign of a tired plant. The pharmacopoeias recognise this with a separate monograph for oxygen produced by concentrator, distinct from the monograph for the cryogenic product. What follows from that is practical: confirm that your ventilators, flowmeters, blenders and terminal units are documented as suitable for concentrator oxygen, because a small number of older devices are not.

The part that fails first is upstream of the sieve

In this climate, the compressor and the dryer decide the life of the plant. Zeolite is destroyed by liquid water. A dryer that cannot hold its dew point through a humid August, an aftercooler with a blocked drain, or a condensate trap nobody empties will slowly poison the sieve, and the first symptom is a purity that will not come back up. Specify the dryer against the worst ambient the plant room will see rather than a temperate design condition, put the drain traps somewhere a technician will actually look at them, and treat dew point as a monitored parameter rather than a commissioning number.

Sizing: count terminals, then argue about diversity

  1. Schedule every terminal unit by clinical area: intensive care, high dependency, neonatal, theatres, recovery, emergency, general wards, and any outpatient point.
  2. Ask the clinical team for the design flow per terminal for their case mix, and get it in writing from them rather than from a supplier's brochure. A neonatal point and an adult ICU point are different problems.
  3. Apply a diversity factor per area. Intensive care approaches full simultaneous use, general wards are far below it. Diversity is where sizing is won or lost, and it is the assumption to interrogate hardest in any proposal you receive.
  4. Add pipeline losses and the reserve requirement, then size the plant so that the design flow can still be met with one module or one compressor out of service.
  5. Sanity check against your actual consumption. If the hospital already buys cylinders or liquid, the delivery records are the single most honest data you have. Use twelve months of them.
  6. Plan the second module rather than buying it now. Leave the plant room space, the pipework tee and the electrical capacity for expansion, and add capacity when the load justifies it.

The mistake to avoid is sizing against the worst week of the pandemic. A plant sized that way runs at a small fraction of its design flow for years, and PSA plants do not enjoy heavy turndown: cycling changes, the compressor short-cycles and unloads, and efficiency per cubic metre gets worse rather than better. Oxymat's range spans compact and half-pallet units through to the larger O60 and O200 Nordic and Nordic Supreme machines precisely so that a hospital can match a plant to a real load, and a staged approach usually beats one oversized machine. Size the buffer vessel deliberately too, because it is what absorbs demand spikes and stops the plant cycling itself to death during a busy theatre list.

Purity and the parameters you watch every day

A hospital oxygen plant is a medicinal product manufacturing operation, whether or not anyone in the building describes it that way. That framing settles most arguments about instrumentation.

ParameterWhy it mattersWhat the system should do about it
Oxygen concentration at the outletIt is the product specification, and it falls before anything else fails visiblyContinuous analyser, low alarm, and automatic isolation from the pipeline before out-of-specification gas can reach a patient
Dew point after the dryerWet air destroys the sieve and can carry water into the pipelineContinuous measurement with an alarm, not a commissioning check
Carbon monoxide and carbon dioxideBoth indicate a contaminated air intake, most often engine exhaustMonitoring appropriate to the product, and an intake relocated away from generators, boilers and the ambulance bay
Delivery pressure and flowFalling pressure at peak demand is the early warning of undersizingTrended and alarmed at the plant and at the hospital engineering desk
Compressor and dryer run hoursService intervals and the electricity bill both follow from theseLogged, with consumables planned against them
Reserve manifold contentsThe reserve is worthless if nobody notices it has been drawn downContents gauge alarmed to a manned point

Two supporting points. First, the analyser is itself an instrument with a calibration interval, and an unverified analyser is not a control. Put it on the same calibration register as everything else in the hospital and keep the certificates. Second, the pipeline downstream is governed by medical gas pipeline practice, which expects proper identification, testing, purging and verification before any patient breathes from it, along with a permit system for any work on a live system. If the plant is a retrofit onto an older pipeline, budget for testing the pipeline as well as installing the plant.

Redundancy: what happens when the plant stops

Medical gas practice expects three sources of supply: a primary, a secondary and a reserve, each able to take over without clinical interruption. A PSA plant is the primary. What plays the other two roles is a decision, not a detail.

The economics, done honestly

Comparisons that show a plant paying for itself in a few months usually compare the plant against the hospital's worst month of cylinder purchasing. Build the comparison properly instead, on a delivered cost per cubic metre of oxygen actually consumed, over at least five years.

On the cylinder side, count more than the gas: cylinder rental and demurrage, transport, the losses left in a cylinder when it is changed, the porters' time spent moving them, the store space, and the clinical cost of the occasions when a delivery arrived late. On the plant side, count capital, installation and pipework, the electricity to run the compressor (this is the dominant running cost and it moves with your tariff), consumables, the service contract, the plant room space and its ventilation, and the cost of the reserve you must keep anyway.

Some honest conclusions from that arithmetic. A small clinic with low, intermittent demand is usually better off with cylinders, and the crossover point depends on your average daily consumption rather than your peak. Liquid oxygen with a vacuum insulated evaporator makes sense above a steady, substantial demand and where a dependable delivery route reaches the site, which outside the main cities is a logistics question rather than a price one. And a hospital with frequent load-shedding and no generator capacity to spare should price the generator fuel into the plant case, because oxygen production is a continuous electrical load and it does not pause when the grid does.

Installation details that decide whether it lasts

Back to all Insights