Instrumentation and safety

The amount of hydrogen crossing the membrane is nearly independent of current. It crosses because of a pressure difference, so much the same amount comes through even when you pass less current. Oxygen, by contrast, is produced in proportion to current: drop the current to 40 percent and the oxygen drops to 40 percent too. On the same stack, lowering the current is the same as lowering the current density — that is the figure quoted in A/cm² in the data. What is managed is the ratio of hydrogen to oxygen. The hydrogen coming across stays as it was while the oxygen falls, so the ratio rises. The same hydrogen is mixing into less oxygen. Pressure raises it by a different route. Higher pressure widens the hydrogen pressure difference across the membrane, so more hydrogen crosses in the first place. Low load and high pressure together bring both effects at once. The measured figures: on a 23-cell HXS-2 (0.1 M KOH, 43–45 °C), 0.39 percent at rated load and ambient pressure, 0.66 percent at 40 percent load and ambient, 0.98 percent at rated load and 6 barg, and 1.75 percent at 40 percent load and 6 barg. Lowest to highest is a factor of 4.5, and at 6 barg with 40 percent load only 0.25 percentage points remain to the 2 percent control limit.

The Measuring principle column says what the instrument physically responds to. For the same oxygen concentration, each principle is strong over a different range and carries different consumables and calibration intervals. Read this column first and it becomes clear why some models belong at trace level and others at percent level. For concentration there are two families. Principles that consume the target gas reach lower, but the sensor has a service life. Principles that do not consume it last longer and hold calibration longer, but lose resolution as the concentration falls. Flow meters and pressure instruments sit outside those two families: they measure how much is flowing, or a pressure, rather than what the gas is made of. On pressure the table covers the ones that put out an electrical signal (piezoresistive, ceramic capacitive), the mechanical ones you read off a pointer without power (Bourdon tube, diaphragm), and electromechanical contacts that make or break the circuit themselves. How far down a given model reaches varies within a principle. That is what the Range column in the table records. Infrared (NDIR), the method most people reach for first in leak detection, cannot be used on hydrogen. The hydrogen molecule does not absorb infrared, so there is no signal to read. Do not move an infrared detector used for another flammable gas onto a hydrogen duty. Hydrogen leaks are caught with catalytic bead or electrochemical sensors. Leak detection splits two ways. A fixed detector is mounted on a wall or a post to watch the concentration in the area continuously, and it puts out 4-20 mA or a contact so it can drive an alarm and an interlock. A portable one is carried to find where a leak you already know about is coming from. Neither substitutes for the other, which is why the table is split on that axis: a portable unit left in place sees nothing while nobody is holding it, and a fixed one cannot be walked along a run of fittings joint by joint.

Electrochemical, fuel-cell type

Measures the current produced as the target gas reacts at an electrode; the current is proportional to concentration. The cell generates its own current like a fuel cell, so no external supply is needed. Strong at low levels, reaching ppm and, on some models, below. The electrode and electrolyte are consumed by the reaction, so the cell is a replacement item

Polymer electrochemical

A solid polymer membrane in place of a liquid electrolyte. Nothing to dry out or leak, so there is no top-up and no restriction on mounting orientation

Thermal conductivity

Uses the fact that gases carry heat at different rates: the heat leaving a warmed element reads the composition. Hydrogen conducts about seven times as well as air, so the contrast is large and this principle suits it particularly well. No consumables, but it only works on a two-component mixture; a third gas leaves the reading ambiguous

Thermo-paramagnetic

Oxygen is drawn into a magnetic field and other gases are barely affected. The strength of that pull gives the oxygen concentration. It does not consume oxygen, so it is long-lived and strong at percent-level concentrations

Optical luminescence

A fluorescent layer emits light when illuminated, and oxygen reaching it suppresses that emission. Concentration is read from how far the emission drops and how long it takes to recover. Oxygen is not consumed, so calibration holds for longer; the fluorescent cap is the periodic replacement item

Thermal mass flow

Measures the heat a flowing gas carries away from a heated element. The correction factor differs by gas, so the unit has to arrive configured for hydrogen

Laminar differential pressure mass flow

Uses the fact that the pressure drop across a laminar element is proportional to flow. Pressure and temperature are measured alongside it and converted to mass flow. Response is fast

Catalytic bead (pellistor)

A flammable gas burns on a catalyst-coated bead and the heat raises the resistance of a platinum coil, which is what is measured. It is read as a percentage of the lower explosive limit and used to watch a room for leaks. Combustion has to happen, so a minimum oxygen concentration is required; in oxygen-poor or inert atmospheres the reading runs low. Catalyst poisoning sets the life

Tracer-gas sniffing

The part under test is filled with a hydrogen tracer gas, usually 5% hydrogen in 95% nitrogen, and a probe carrying a sensor that responds only to hydrogen is passed over the surface to find where it escapes. The scale is ppm rather than %LEL, so it catches very small holes and the probe pinpoints where they are. The gas has to be put into the part, so it cannot watch the concentration in a room around running equipment

Piezoresistive

Pressure deflects a diaphragm and changes the resistance of a silicon chip behind it. Wide span and low cost make it the default for pressure. The wetted face is metal, so which metal it is decides media compatibility

Capacitive (ceramic)

The capacitance between a ceramic diaphragm and a backplate changes as the diaphragm deflects. The wetted face is ceramic rather than metal and there is no fill fluid. This is why it is the one to look at instead of a metal diaphragm for alkaline and acidic media

Bourdon tube (mechanical)

Pressure straightens a curved tube and that movement drives a pointer. It reads without power, which is what you want for commissioning and walk-round checks, but it sends no signal, so it cannot serve logging or an interlock

Diaphragm (mechanical)

A thin diaphragm deflects under pressure and drives the pointer. It holds lower pressures better than a Bourdon tube, and a PTFE lining or a special-material pressure chamber makes it fit corrosive media

Electromechanical contacts

Above a set pressure a spring and lever physically make or break a contact. Nothing electronic sits in the path, so it still acts when the controller is down — which is why it is used to wire a shutdown interlock independently

If the instrument goes into a hazardous area it has to be rated for that area. A single line such as Ex db IIB+H₂ T6 Gb in a catalogue is four separate things written end to end. The db at the front is the protection method. Flameproof (the Ex d family) contains an ignition inside a strong enclosure so it cannot spread; intrinsic safety (Ex ia, Ex ib) keeps the energy in the circuit below what it takes to ignite anything in the first place. Analysers with a display and mains power are usually flameproof; small transmitters with no display are usually intrinsically safe. The IIB or IIC in the middle is the gas group. The further along it goes the easier the gas is to ignite, and hydrogen sits in the most demanding group, IIC. A unit certified only to IIB cannot be used in a hydrogen area. Some markings add hydrogen on its own as IIB+H₂, so do not rule a unit out on the letters IIB alone. T6 is the temperature class. It says the surface stays below the ignition temperature of the gas, and T1 through T6 guarantees progressively lower surface temperatures. The Gb at the end is the equipment protection level: Ga corresponds to Zone 0, Gb to Zone 1, Gc to Zone 2. The certification schemes differ by country: ATEX in Europe, IECEx internationally, KCs in Korea. The same unit may still need a separate KCs certificate to be installed in Korea, so tell us the country of installation first. Below is only what we have confirmed against manufacturer documentation. A model left out is not a model without a hazardous-area version; it is one we have not confirmed. Note also that the hazardous-area version is usually a separate model number rather than an option on the same one. Get the part number wrong on the order and a general-purpose unit arrives.

Michell XTC601

Flameproof Ex db IIB+H₂ T6 Gb; ATEX, IECEx, KCs

Michell XTP601

Three build codes: general purpose GP1 and GP2, hazardous area EX1. EX1 is flameproof Ex db IIB+H₂ T6 Gb, dust Ex tb IIIC T85°C Db (IP66); ATEX, UKCA, IECEx, cQPSus, TR CU Ex

Messkonzept FTC 320

Non-flameproof; FTC 320 EX is a separate model

Chang AI CI-PC551-2

Flameproof Ex d IIB T6 Gb, IP65. The manufacturer manual prints the same field as Ex d C T6 Gb

GASDNA DA-600S

Flameproof Ex db IIC T6 Gb (IP6X), dust Ex tb IIIC T85°C Db; KCs, ATEX, IECEx, JPEx, CE

Southland OMD-507

Non-Ex as standard. An intrinsically safe option exists, but the manufacturer rates it for Class I Div 2, which is not the Zone 1 flameproof grade this column means

Michell Easidew

Easidew I.S. is the separate intrinsically safe model; IECEx, ATEX, UKCA

PRIGNITZ SPT

PMP-S122-Exi is the separate intrinsically safe part number; II 1G Ex ia IIC T4 Ga; CSA, ATEX, IECEx. Its datasheet states hydrogen approval

STORK DEWCom II

CE marking only in the manufacturer documentation; no hazardous-area rating confirmed

Bronkhorst EL-FLOW Select

EX-FLOW is the separate hazardous-area range

MSA ULTIMA X5000

ATEX, IECEx, CSA, FM, INMETRO and DNV-GL approvals; SIL 2 suitable. The marking depends on the part number and sensor

Honeywell Sensepoint XCD

ATEX II 2 GD Ex d IIC T6/T5 Gb, Ex tb IIIC T85°C Db; the same under IECEx. IIC gives hydrogen room to spare

Everything else

Models we have not yet confirmed against manufacturer documentation

The same measurement is read in different places depending on the model. That is what the Display column records. A screen on the instrument is the simplest case. Where the sensor and the screen are one body it is integrated with the sensor; where it is built to drop into a cabinet door it is a panel-mount type. Either way somebody reads the number on site. A model without a screen only puts the value out as a current. That is the 4-20 mA signal, and the number is visible only if something at the other end receives it: a PLC, a recorder, or the manufacturer's own separate readout. Miss this and the instrument is fine while nobody can see what it reads. Not all 4-20 mA wiring is the same. A two-wire loop carries power and signal on the same pair; a three-wire unit takes its power separately. Connectors are standardised too, so a model using a DIN 43650 connector needs the matching socket ordered with it. Instruments with a wide span often add auto-ranging: the unit changes its own measuring range as the concentration moves. On a bare 4-20 mA link that means the same 20 mA is 1000 ppm at one moment and 100% at another. Decide whether you also want the signal that identifies the current range.

This is where instrumentation most often goes wrong against upper-level control. Whether the command reached the unit, whether the unit accepted it, whether it is running, whether it finished, whether it failed: each of these has to be a separate state. Without that separation you get a screen that reads normal while nothing is actually happening.

The same quantity comes in different units from different instruments. This table alone has oxygen-in-hydrogen reported in percent and in ppm. If the tag units are not matched and the 4-20 mA scaling is carried over as is, the reading is out by a factor of ten thousand, and it often sits in a plausible range long enough that nobody notices. If you also bring pressure in, settle barg against bara for the same reason.