Read enough confined space incident reports and a pattern appears that is uncomfortable for anyone specifying equipment. The detector was present. It was calibrated. It alarmed. And the worker inside still died.
The mechanism is usually the same. A portable gas detector sits at the entry point or just inside it. It measures correctly. On a threshold breach it sounds an alarm and flashes a light — at the detector, which is on the ground beside a manhole, while the worker is eight metres along a culvert, facing away, wearing hearing protection, doing a task that requires both hands and most of their attention.
The three failures that hide behind “the equipment worked”
First, the alarm does not reach the entrant. This is the commonest and the most avoidable. An alarm that signals at a fixed point is an alarm for the attendant, not for the person in the space. If the entrant wears nothing that can warn them directly, the detection has been converted into a notification for someone who is already safe.
Second, the attendant is watching the wrong thing. The attendant's job is simultaneously to monitor the atmosphere, maintain the entry log, watch the entrant, keep the access clear for rescue and communicate with the supervisor. In practice they split attention between a screen, a clipboard and a hole in the ground, and attention is the resource that runs out first. Asking one person to hold six threads is a staffing plan that fails quietly and then catastrophically.
Third, the rescue is not retrievable. Even where the alarm is raised correctly and immediately, confined space rescue in a vessel with a vertical manhole, or a tunnel with a single access point, takes time that the casualty does not have. Hydrogen sulfide at high concentration can render a person unconscious in a single breath. If the control depends on rescue, the control is too late.
What changes when the design assumption changes
The design assumption worth changing is that monitoring exists to produce a record of the atmosphere. It does not. Monitoring exists to interrupt the sequence that leads to a fatality, and the sequence runs through the entrant, who is the only person in a position to withdraw.
That assumption leads to a different equipment brief. The person inside needs to be alarmed directly, on their body, without depending on a network, a speaker, or anyone noticing. The atmosphere should be sampled where they are working rather than at the entrance, because stratified vapour means those are different readings. Their vital signs should be measured, because a worker who is affected gradually may not recognise it in themselves. And the communication path out of the space should not depend on radio propagating through steel, because it does not.
The practical test to apply to your own site
Here is a question worth asking of any confined space monitoring arrangement, including ours. Take an entrant standing at the furthest point they ever work inside the space, at the moment when the atmosphere crosses a threshold. Trace the path of that alarm. Who receives it first? How long until it reaches the entrant? And what happens between those two moments?
If the answer involves the attendant noticing a screen and then shouting down a tunnel, the system is producing records rather than control. That is not a criticism of the detector — it is a statement about where the last metre of the design sits, and it is the metre that decides the outcome.
Where this leads in practice. It is why our monitoring stations alarm on the entrant's wristband as well as at the entry point, why sampling is pump-suction with an extendable hose rather than fixed at the manhole, and why internal units communicate over a cabled bus or LoRa rather than waiting for a mobile signal that will not be there.