Gas detection in refrigeration, part 1.
A refrigerant leak almost never starts as an emergency. It starts at a gasket, a valve seal, a joint that worked loose from vibration. The system keeps running, the cold room stays at temperature and nobody notices a thing. That is exactly the problem: by the time the leak becomes obvious, it has been costing money for quite a while.
MSA Safety has been making this point on its blog, with a focus on halocarbon refrigerants, supermarkets and data centers. In this series we take it where it matters most to us: the industrial ammonia plant, cross-referencing it with the IIAR standards we have been covering every Tuesday in Refrigeration Insights.
What an unseen leak costs
The immediate cost is the refrigerant that has to be replaced. But that is the smallest one. With a reduced charge, the system runs longer to hold the same setpoint temperatures, and that translates into:
1) Higher energy consumption, with nothing on the control panel to explain it.
2) More wear on compressors, motors and other critical components, which are overworked long before any alarm goes off.
3) Loss of capacity: if the leak continues, the system eventually can no longer keep up with the heat load, trips on low pressure and shuts down. In a product cold room, that shutdown is the most expensive cost of all.
MSA sums it up with a list any maintenance manager will recognize: the leak is only discovered once there are significant refrigerant losses, efficiency has dropped, equipment is compromised, product or people have been affected, or an emergency service call is needed. In multi-site organizations, the maintenance team also ends up chasing emergencies instead of following the plan.
“With ammonia, you can smell the leak”
It is the most common objection, and it is partly right: ammonia announces itself by its odor at very low concentrations, long before it becomes dangerous. But your nose is not a detection system:
1) It only works where someone is present. An unmanned machinery room at night, a rooftop with evaporators, a cold room closed over the weekend: there, the leak progresses with no witnesses.
2) People get used to it. Those who work in the plant every day stop noticing a background odor, and a small, steady leak becomes “the usual smell”.
3) It leaves no record and triggers nothing. It does not start the ventilation, does not alert anyone outside the room and does not close any valve.
That is why the standard does not rely on the nose.
What we have already covered in the standard
During our IIAR standards season we introduced ANSI/IIAR 2-2021, the cornerstone standard for industrial ammonia, and at Tecno Fidta we boiled it down to one sentence: the standard does not ask for a detector, it asks for a sequence. For the machinery room, in short:
1) From 25 ppm: visual and audible alarms, and notification to a monitored location.
2) From 150 ppm (half the IDLH value): in addition, emergency ventilation of at least 30 air changes per hour, with manual reset. General ventilation and room air conditioning that are not part of the emergency system shut down.
3) Above the detector's upper limit or 40,000 ppm (25% of the lower flammable limit), whichever comes first: in addition, automatic de-energization of compressors, refrigerant pumps and normally closed automatic valves that are not part of the emergency control.
Outside the machinery room, in evaporator areas, the standard calls for closing the liquid and hot gas valves of the affected zone from 25 ppm.
And this week, with IIAR 5-2025 on startup and commissioning, we covered the step before: safety systems are verified before the first ammonia charge. If the detection sequence was not tested during startup, nobody knows whether it works.
What happens below the alarm
So far, the standard: safety thresholds. What MSA's perspective adds is the other half, what happens before those thresholds are reached.
A small leak at an evaporator can stay below 25 ppm for weeks without triggering anything. Yet it is already losing charge and already signaling that a gasket or seal is starting to fail. A detector that reports its reading continuously, not only when it alarms, makes that trend visible and lets you send maintenance before the leak grows. That is where detection stops being just a safety obligation and becomes a maintenance tool.
From detecting to monitoring
The difference between a detector that alarms and a system that informs is visibility. As MSA puts it: knowing at all times the status of the detectors, the alarm history and the operating conditions makes it possible to detect earlier, lose less refrigerant, plan maintenance and reduce unplanned downtime.
In practice, with the MSA Bacharach line this is built on three levels:
1) Point detection: MGS-400 series detectors integrate with the MGS-408 controller or directly with the plant BMS via Modbus (RS-485). At the room entrance, the MGS-401 monitor displays the concentration from outside, so nobody walks into a room that is already in alarm.
2) Multipoint sampling: the Multi-Zone draws air from up to 16 points (next to chillers, compressors and valves) and analyzes it with a central infrared sensor, with a minimum detection limit of 1 ppm for halocarbon gases.
3) Remote monitoring: MSA Parasense lets you see detector status, alarms and system health for one or several plants from a single place.
CO2 and halocarbons
CO2. Here, smell is no help at all: CO2 is odorless and colorless. As we saw in May, ANSI/IIAR CO2-2021 sets the alarm according to occupancy (up to 5,000 ppm where the public is present, up to 15,000 ppm in industrial occupancies), and sensors are installed 30 cm (12 in) or less above the floor, because CO2 sinks. The 30,000 ppm that appear in the standard are not an alarm setpoint: they are the design limit for the charge in rooms with CO2 equipment. We will devote a full Refrigeration Insights issue to it on December 1.
Halocarbons (HFC, HFO and A2L). Here the economic argument from the first part applies in full, plus the regulatory pressure on refrigerant losses.
A detector without records does not exist
ANSI/IIAR 6-2025, which we cover next Tuesday in Refrigeration Insights, requires every inspection, testing and maintenance task to have a defined frequency and a record. For detectors, that means documented functional testing and calibration. With the MGS-400 app (the same one we demonstrated live at Tecno Fidta) you can test relays, LEDs and buzzer, perform zero and span calibration, and generate the calibration certificate in the field. That is the record the standard requires and the one an audit will look for.
Practical advice
Three questions to ask about your plant today:
1) When was each detector last calibrated, and where is the certificate?
2) What happens automatically in your machinery room at 25, 150 and 40,000 ppm? Has anyone tested it since startup?
3) If a small leak starts at an evaporator tomorrow, how many days until you find out?
If any of the three has no clear answer, that is where the first job is.
In the next issue: which sensor technology is best for each gas and each point in the plant.
Sources: MSA Safety, “Why Waiting for a Refrigerant Leak Can Cost More Than You Think” and “Three Ways Refrigerant Leaks Can Affect Data Center Cooling Performance” (Spotlight on Safety blog, 2026); ANSI/IIAR 2-2021, sections 6.13.2, 6.14 and chapter 17; ANSI/IIAR 5-2025; ANSI/IIAR 6-2025; ANSI/IIAR CO2-2021, International Institute of Ammonia Refrigeration. Thermomac is a corporate member of IIAR and a channel partner for the MSA Bacharach gas detection line.
