Specialty Gas Compressor Selection Guide: From Gas to Machine Type
The gas comes first, the pressure second. Reverse that order and the shortlist is wrong before you start.
Select the gas first, then the machine. That is the order our tool works in, and it is the opposite of how most equipment enquiries are written, which usually start with the pressure the buyer already has in mind.
This guide walks the five steps in order. It gives the gas-to-machine mapping for the 12 gases we cover first, states plainly where our coverage stops, and flags the one input we never convert.
Read the steps in sequence, and you can see which machine family your duty point survives to — and whether it survives at all.
Step 1 — Which gas properties eliminate a family before anything else?
Four properties do most of the eliminating. They come from the gas data record, not from the buyer's process description.
| Property | What it eliminates |
|---|---|
| Pyrophoric — ignites on contact with air | Everything. Silane, phosphine and arsine go straight to a human, and they are the reason this category has a category at all |
| Purity-critical / oxygen service | Oil-injected machines, and for oxygen, any oil contact anywhere. Oxygen duty also prohibits copper alloys |
| Corrosive — wet | Clean-machine options. Wet chlorine, wet H₂S and wet CO₂ point at liquid ring or at a human review |
| Embrittlement — hydrogen | High-strength steels. Hydrogen duty is specified with materials that do not embrittle |
Two more properties matter later rather than first: molecular weight, which decides whether a very large flow at low pressure is a positive-displacement duty at all, and the autoignition temperature, which feeds the discharge-temperature calculation.
Step 2 — What does the tool need to know?
| Input | Required | What it decides |
|---|---|---|
| Gas | Yes | Everything in Step 1, plus molecular weight and the isentropic exponent |
| Flow, with a unit | Yes | Whether the duty is in the positive-displacement window at all |
| Suction pressure, bar(a) or bar(g) | Yes | The pressure ratio. There is no pump equivalent of this input, which is why compressor selection starts with two pressures rather than one |
| Discharge pressure, bar(a) or bar(g) | Yes | The machine family |
| Suction temperature, °C | Yes | The discharge-temperature estimate, and the unit conversion |
| Oil-free required | No | Rules oil-injected machines out |
| Gas contains liquid or solids | Yes, three-state | Rules out or redirects the machine family |
| Duty — continuous or intermittent | No | Tilts between reciprocating and diaphragm |
| ATEX / hazardous area | No | Carries through to the specification sheet |
Notice what is not asked. There is no request for a stage count, no request for intercooling temperatures, no request for a compressor efficiency. Those are design outputs, and a screening tool that produces them is producing guesses.
Step 3 — Which unit always goes to a human?
Our main unit is Nm³/h at a 0°C reference (32°F). Two other units are converted, with the conversion declared in the report:
| Your unit | What we do |
|---|---|
| Nm³/h (0°C reference) | Accepted as entered |
| Sm³/h (15°C / 60°F reference) | × 0.935 to Nm³/h |
| scfm | × 1.699 to Nm³/h |
| Actual m³/h at your suction conditions | Not converted. The duty goes to a human. |
Note that the last row looks unhelpful. It is the most deliberate line in the entire input set.
Actual volumetric flow depends on suction pressure, suction temperature and gas composition. Converting it approximately would produce a number that looks precise and is wrong, and everything downstream — machine family, stage count sensitivity, discharge temperature — inherits the error.
A unit that carries its own uncertainty is a signal of engineering seriousness, not a reason to guess. The handover is the correct output.
Step 4 — Which machine does your gas point to?
This is the mapping our tool applies for the first 12 gases. It is a starting position, not the final answer: pressure, flow and phase then narrow it further.
| Gas | Recommended starting family | The rule that matters most |
|---|---|---|
| Hydrogen | High-pressure pure hydrogen → diaphragm or reciprocating. Low-pressure, large-flow hydrogen below 45 bar → dry screw, which is routine in the industry, not an exception | Molecular weight 2.02 and embrittlement. No high-strength steel |
| Nitrogen | Any family that fits the pressure and flow | The baseline gas, no special constraints |
| Oxygen | Any family that fits, with oil-free construction mandatory | Oil plus oxygen is an explosion path. No copper alloys at high pressure, and the discharge-temperature treatment is stricter |
| Natural gas | Oil-injected screw for boosting, reciprocating for other duties | ATEX and the fuel-gas specification |
| Biogas | Any family that fits, with material selection starting at 316 stainless | Wet H₂S corrosion. The desulphurisation state of the gas is part of the question, not a detail |
| Carbon dioxide | Screw for large-flow low-ratio duty, diaphragm or reciprocating at high pressure | High density and dry-ice risk. A phase-change prompt applies at low pressure ratios |
| Ammonia | Any family that fits, with copper and copper alloys prohibited throughout | Toxicity, and a hard material prohibition |
| Chlorine | Dry chlorine in carbon steel is routine. Wet chlorine → liquid ring or a human review | Dry and wet chlorine are two different problems |
| Helium | Diaphragm first, because of leakage | Molecular weight 4 and the highest leakage tendency in the set. Valuable, so recovery duty is common |
| Hydrogen sulfide | Zero-leakage machine families required, with a human review behind them | Toxic and corrosive at the same time |
| Carbon monoxide | As hydrogen sulfide | Toxic and flammable |
| Vinyl chloride | Liquid ring is common in this service | Polymerisation risk with a carcinogenic gas. Discharge temperature is limited to 100°C (212°F), not 135°C (275°F) |
Step 5 — Where does the tool stop?
After the gas, three numbers narrow the family:
- Discharge pressure. Above 45 bar (650 psi), dry screw machines are out; oil-injected screw follows at 50 barg (725 psi). Above 400 bar (5,800 psi), only diaphragm remains, up to 900 bar (13,050 psi).
- Differential pressure. For dry screw machines, 15 bar (218 psi) across the machine, with a human review above it.
- Flow. Below 300 m³/h suction, screw machines drop out. Above 100,000 m³/h, the duty has left this category.
These cut-offs are the range over which the tool answers automatically, and on the diaphragm side they follow documented production: Chinese makers have high-pressure diaphragm machines in series production to 70 MPa (700 bar) and 90 MPa (13,050 psi) machines delivered, with published ceilings at 100 MPa (14,500 psi) — which is why the window runs to 900 bar. Oil-lubricated reciprocating machines are published to 1,000 bar (14,500 psi) by European manufacturers, while domestic evidence for that design now reaches 40–50 MPa (400–500 bar); on that side a gap remains, and it is stated rather than smoothed over.
The visible result is that a large share of duty points end in a handover rather than in an automatic report. That is the intended behaviour. A tool that answers every question is a tool that is guessing somewhere, and in process gas the cost of a wrong answer is not symmetric with the cost of an unanswered one.
What people assume, and what the mapping says
| Assumption | What the mapping actually says |
|---|---|
| "Air compressor experience transfers." | It mostly does not. A process gas duty has a molecular weight, an isentropic exponent, a purity requirement and a phase risk that air does not. |
| "Oil-free is a preference." | For oxygen and for any purity-critical gas it is a constraint. Oil-injected machines are excluded, not derated. |
| "Higher pressure just needs a bigger machine." | Higher pressure changes the machine family. Above 50 barg the screw families leave the table entirely, and above 400 bar only one family remains. |
| "Wet and dry chlorine are the same duty with different corrosion allowances." | They are different machines. Dry chlorine in carbon steel is ordinary; wet chlorine is a liquid-ring or human-review question. |
| "Discharge temperature is a cooling problem, not a selection problem." | It is the selection problem. It is what decides stage count, and stage count above four on one crankshaft is a handover. |
| "The tool left out centrifugal machines." | It did, and the reason is that this guide maps positive-displacement families. A dynamic machine serves large continuous flow at a low pressure ratio, and those duties are bought by a different group of contractors on a different supply chain. Mixing the two into one table would answer neither question well. |
Questions we get about this guide
Can I use an air compressor for a process gas? The pressure and flow arithmetic may line up. The gas properties will not. Oxygen service prohibits any oil contact; ammonia prohibits copper; hydrogen requires materials that do not embrittle. These are not derating factors, they are exclusions.
Why do you ask for suction pressure as well as discharge pressure? Because the compression work depends on the ratio, not the final pressure alone. Ten bar of lift from 1 bar is a very different machine from ten bar of lift from 40 bar.
What if my gas is not in the list? The duty goes to a human. The library covers 12 gases in the first release and a further eight in the second; anything outside it is routed rather than approximated.
Do you calculate the compressor power? No. We give a boundary classification and an RFQ specification sheet. Power, cooling arrangement and the mechanical design belong to the manufacturer's engineering team.
What does the tool output, exactly? A recommended machine family within its envelope, the gas-specific constraints that apply, and the specification lines a supplier needs. Where the envelope is exceeded, it outputs the boundary that was exceeded and hands the enquiry on.
What this article does not cover
Note that the guide does not calculate compressor power, size a motor, design an intercooling arrangement or produce a stage count. It maps a duty point to a machine family and states the boundaries. Power, cooling and mechanical design belong to the manufacturer's engineering team.
Sources
| # | Basis |
|---|---|
| 1 | GasCompressorSelect boundary and gas data set v1.3, §1 (scope), §2 (machine families and exclusions), §3.1 (input fields), §3.2 (unit normalisation rules), §3.3 (pressure ratio, stage count, discharge temperature), §4 (gas data schema), §5 (Tier 1 and Tier 2 gas lists with selection hooks), §6 (handover rules), §7.C–D (envelope rulings 2026-09-19) |
| 2 | Gas property values quoted — hydrogen molecular weight 2.02, isentropic exponent 1.41, embrittlement flag, autoignition 560°C; oxygen oil-free requirement; ammonia copper prohibition; vinyl chloride 100°C discharge limit — gas data set §4 and §5 |
| 3 | Envelope figures and domestic manufacturer evidence — appendix B of the data set |
| 4 | Dynamic machine exclusion rationale — §1.0, §1.2, decided 2026-09-14 |