Confined Space Hazards, Risks and Controls

A confined space does not invent new hazards. It concentrates and hides ordinary ones. Rust is a chemical curiosity in open air and a killer inside a sealed tank. A solvent spill evaporates harmlessly on a workshop floor and builds a flammable layer in a pit. What makes this work lethal out of proportion to how often it happens is that the enclosure removes the margins we rely on everywhere else: dilution by moving air, easy escape, line of sight, and the ability of a colleague to simply walk over and help.

The message up front: atmospheric hazards are among the most serious and most common causes of confined-space fatalities, and the most dangerous property of an oxygen-deficient atmosphere is that it does not feel dangerous. Engulfment, entrapment and the other physical hazards are covered here too, and they kill people as well. The atmosphere is also not a fixed state you test once; it is a process the work itself changes. Every control worth discussing sits somewhere on the hierarchy of controls, and the most effective one by a wide margin is not entering at all.

1. Why an enclosure changes the physics of ordinary hazards

The defining features of a confined space are enclosure, restricted access and lack of designed occupancy. None is a hazard by itself. What they do is remove the mitigations that keep ordinary hazards from becoming fatal ones.

Enclosure removes dilution. A gas released from a residue disperses into an effectively infinite volume outdoors and into a few cubic metres inside a vessel, where it stays. Anything produced inside accumulates; anything consumed inside is not replenished. That symmetry is the root of both toxic build-up and oxygen depletion.

Enclosure creates stratification. Still air does not mix. Gases denser than air settle into the lowest reachable point as a layer; lighter gases collect in the crown. A confined space is therefore rarely one atmosphere but several stacked on each other, and a person moving downward can pass out of a breathable layer into an unbreathable one without crossing any visible boundary.

Restricted access removes escape and delays help. This is the quiet multiplier behind almost every confined-space death: an exposure survivable if the person is out in seconds becomes fatal when extraction takes minutes, and a manway a healthy person climbs through comfortably is very hard to pull an unconscious adult through. Nothing in the space was designed for people either, so everything a person needs is carried in, and every item carried in brings hazards of its own.

Identifying confined spaces, the entry programme, the roles and the ways programmes fail are covered in the confined space requirements and procedures pillar. The permit itself sits in the confined space permit guide. This article stays with the hazards.

2. Oxygen deficiency: the hazard that gives no warning

Oxygen deficiency is the most common cause of death in confined spaces and the least intuitive hazard in industrial safety. Oxygen is almost never removed deliberately. It is consumed or displaced by processes that look entirely innocent from outside.

• Oxidation of exposed metal. Rusting consumes oxygen. Irrelevant in the open; inside a damp steel tank with a large internal surface area, the steel is quietly and continuously taking oxygen out of the air. A tank closed for weeks can be dangerously depleted with nothing having happened inside it, and with no smell and no visible sign.
• Microbial action and decomposition. Wherever there is sludge, effluent, silt or standing water, micro-organisms are respiring. Decomposition is a double hazard: it lowers oxygen and adds toxicity at once.
• Combustion, and absorption by damp organic material. Any flame or engine consumes oxygen and adds combustion products. Separately, damp grain, wood chip and certain wastes absorb oxygen out of the air in contact with them, which is why bins and holds storing organic material are a recognised case.
• Displacement by an inert gas. Purge and blanketing gases are used because they are unreactive, and that same property makes them lethal to breathe. A space left blanketed, or one into which an inert supply has leaked from a connected system, can be almost free of oxygen while looking and smelling like ordinary air.
• Displacement by a denser gas pooling in a low point. A heavier-than-air gas does not need to fill the space to kill. It only needs to fill the part a person's head will occupy. Pits, sumps, manholes, drainage chambers and vessel bottoms all collect denser gases, and the person kneeling to work is exactly where that layer is.

That is a documented and repeated pattern rather than a theoretical risk: a substantial share of confined-space deaths are would-be rescuers who went in to help someone who had collapsed, because the space looked normal, which it did, because the hazard is invisible. Acceptable and unacceptable oxygen concentrations are specified in the applicable regulations, standards and guidance for your jurisdiction, and in the assessment for the specific space. I reproduce no figure, because a number half-remembered from an article and applied to the wrong space, instrument or jurisdiction is worse than no number at all.

3. Oxygen enrichment: dangerous, not beneficial

The instinctive reading of "more oxygen" is "safer air". It is the opposite. The usual sources are equipment: an oxy-fuel cutting set left inside between shifts, a leaking hose or fitting, a cylinder valve not fully closed, an oxygen line passing through the space. Because the gas is odourless and the leak may be small, enrichment builds over hours with nobody aware. There is also the occasional and lethal misconception that an oxygen supply can freshen stale air.

The harm is fire. Raising the proportion of oxygen dramatically increases ease of ignition, speed of flame spread and intensity of burning. Materials that ordinarily smoulder or self-extinguish burn readily and fiercely, including clothing, hair, textile gloves, plastics, lagging and greasy residues, and marginal ignition sources become sufficient. In plain terms the person inside is wearing fuel, standing in fuel and surrounded by fuel, with the ignition threshold lowered. Oxygen also saturates porous materials, so a garment removed from an enriched atmosphere stays unusually flammable for a while afterwards.

The control principle is easy to state and routinely neglected: gas supplies are not left in a confined space, they are isolated at source rather than at the torch, and hoses come out when work stops. That is an isolation control, belonging with the lockout/tagout discipline.

4. Toxic atmospheres: sources, and how testing misses them

Toxic atmospheres arrive from four directions, and a competent assessment considers all four rather than only the obvious first.

• What the space held, and what is left. Residues, sludges, scale and product films in tanks, sumps and interceptors keep releasing vapour long after the space is drained and looks empty. "Empty" and "clean" are different words for good reason.
• Decomposition. Wherever organic material, effluent or standing water has sat without oxygen, biological breakdown generates gas. This characterises sewers, drainage systems, wet wells, slurry systems and any space that has held waste, and the gases produced are among the most dangerous in this work, often both toxic and flammable.
• Adjacent and connected systems. Pipework, drains, vents, ducts, common headers and shared sumps connect the space to processes elsewhere, so something happening in a different part of the site can deliver gas into it while people are inside, with nobody there able to see it coming.
• What the work introduces. The category most often omitted, because the assessment is written about the space rather than the job. Solvents, paints, resins, coatings, adhesives, cleaning chemicals, welding fume, grinding dust and the exhaust of any combustion-powered equipment all create a contaminated atmosphere that was not there when the space was tested.

Density defeats a test at the opening. Several of the most dangerous gases here are denser than air. They collect in the lowest part of the space, precisely where people kneel, crouch and put their heads to work. A test taken at the access opening, at the top of a vessel, or at chest height on a ladder can read as entirely acceptable while a hazardous layer sits undisturbed below. Testing therefore has to be representative of the whole space and taken at the depths and locations where people will actually be, including remote ends, side branches, low points and behind internal structures. The principle is representativeness, not a single reading.

Smell is not an instrument, and with some gases it actively misleads. Some have no odour. Others smell strongly at low concentration but deaden the sense of smell at higher concentration, so the person experiences a powerful smell that then appears to fade. The natural reading is that the problem has gone; the correct reading is that it has got worse and the warning mechanism has been switched off. A strong smell followed by no smell is an escalation, not reassurance.

Disturbance releases gas into a space that tested clear. Sludge, scale, silt and product films trap gas under and within them, and walking on a deposit, scraping it, jetting it or lifting it can release that gas in seconds. A space can be tested, found clear, entered legitimately, and turn hazardous the moment the first tool touches the residue.

5. Flammable and explosive atmospheres

A flammable atmosphere needs fuel, oxygen and ignition. A confined space is very good at supplying the first, ordinarily supplies the second, and is routinely handed the third by the people working in it. The fuel comes from the same places as the toxic hazard, which is why the two so often coexist: vapour from a residue or product film, gas from decomposition, gas entering through connected pipework or a leaking valve, solvents brought in for cleaning, and combustible dust raised by disturbing a deposit. With no air movement, vapour does not disperse but accumulates, preferentially at whichever level its density dictates.

The ignition sources are almost always introduced. Welding and cutting are obvious; less obvious are grinding sparks, hot surfaces, unprotected portable lighting, power tools and their cabling, static discharge from cleaning or jetting, and metal moving against metal. Every item taken in has to be considered a potential ignition source rather than neutral kit.

This is where confined-space work and hot work overlap, and the overlap is dangerous because two regimes each adequate alone can leave a gap between them. Hot work inside a confined space is a compound hazard: the hot work supplies ignition and consumes oxygen while the space supplies accumulated fuel and prevents escape. It deserves one combined assessment rather than two permits that never look at each other. The specifics are in the hot work hazards and permits guide. Flammability limits and the readings that permit entry or hot work are set in the applicable standards and regulations, including the Dangerous Substances and Explosive Atmospheres Regulations 2002 (as amended) in Great Britain. I reproduce no figure.

6. The atmosphere is not a state, it is a process

This is the conceptual correction that does the most good and is most often missing from training. People are taught to test and then, if the result is acceptable, to enter. That framing treats the atmosphere as a fixed property of the space, like the diameter of the manway. It is not. It changes continuously, and working in the space is one of the things that changes it.

• The work changes it. Welding adds fume and consumes oxygen, solvent cleaning and coating add vapour, grinding adds dust, engine-driven equipment adds exhaust. In almost every case the atmosphere at the end of the task is worse than the one tested before it.
• Something was disturbed, or conditions changed. The trapped-gas release under a deposit can happen minutes into the job. Rising temperature increases evaporation from a residue, pressure changes can draw gas out of drainage systems, and a shift in wind alters what a ventilation inlet is drawing in.
• A connected system leaked or was operated. A valve cracked open elsewhere, a pump started, a drain flushed, a line broken for unrelated work. None of that is visible from inside.
• A control failed or was defeated. Ventilation switched off because it was noisy or because the lead was needed for a tool. An isolation lifted by another crew. A purge supply drifting open.

Once you accept that the atmosphere is a process, continuous monitoring stops being an optional extra and becomes the logical consequence of the hazard. Pre-entry testing describes a moment that has already passed by the time anyone is inside; monitoring during occupancy describes the condition people are actually in. That is the principle. What is monitored, how, with which instruments, how often and with what alarm arrangements are matters for the applicable legal framework, the manufacturer's instructions and a competent assessment, and I specify none of them.

7. Atmospheric hazards at a glance

Control class below refers to the level of the hierarchy of controls that genuinely addresses the hazard, not to a specific measure.

8. Physical hazards: the ones that trap, crush, drown and exhaust

Atmospheric hazards kill the most people, but physical hazards cause a great deal of serious harm, and several have the additional property of obstructing rescue.

Engulfment. Free-flowing solids and liquids can surround and bury a person, killing by asphyxiation or crushing. Stored granular material deserves particular attention because of its bridging behaviour: as material is drawn from the bottom of a silo, the remainder can arch over the void and form a crust that looks like a solid, walkable floor. It is not a floor, it is a roof over a hole, and a person standing on it collapses the bridge and is buried in seconds with no prospect of self-rescue. Liquids and slurries can also flow in from connected pipework faster than anyone can climb out.

Entrapment. Converging, tapering and funnel-shaped sections, hoppers and chutes narrow as they descend, so a person who slips down cannot climb back and may become wedged, while internal baffles, plates and agitator assemblies restrict movement and block the route out. The consequence is doubled: entrapment holds the entrant next to whatever else is harming them, and physically prevents rescuers reaching them.

Residual and inadvertent energy. This is what isolation exists to address. Agitators, augers, conveyors, fans and dampers can start, by automated sequence, remote action or someone at a panel who does not know anyone is inside. Stored energy is as dangerous as supplied energy: a spring, a raised component, a pressurised line, residual charge, stored heat or a head of liquid behind a valve can all act after the supply is switched off. Every system feeding the space is a route by which energy arrives, and the isolation has to be proven rather than assumed. See the lockout/tagout guide.

Heat. An enclosure concentrates heat three ways at once: restricted ventilation prevents it leaving, physical work in an awkward posture generates a lot of it, and protective clothing worn for other hazards prevents the body shedding it. Heat strain develops faster than people expect and faster than their own sense of how hot they are. Its real significance is not collapse, though that is real, but that it degrades judgement, coordination and reaction time, so the person becomes less likely to notice a change, read it correctly or act promptly. Heat is a hazard multiplier, not just a hazard.

Restricted access and noise. Small openings, long routes and awkward geometry force bad posture and limit what can be taken in, but more decisively they govern the rescue: the tighter the access, the longer extraction takes, and the more exposures that would have been survivable become fatal. Noise compounds this, because a hard-surfaced enclosure is highly reverberant, so sound reflects and builds instead of dissipating, damaging hearing and destroying the entrant-to-attendant communication by which distress is meant to be detected.

Water, falls and darkness. Liquid can enter from connected drainage, an upstream operation, groundwater, rainfall or a failed isolation, and very little depth drowns a person who is trapped or unable to raise their head. Falls into a space, and falls within it from a ladder or an unstable deposit, leave the casualty injured somewhere hard to reach and possibly in a low-level atmospheric layer they were previously above. Darkness and obscuring dust, steam or mist cause slips and contact injuries and stop the attendant seeing the entrant, and the control for that is the next hazard: lighting and powered tools are electrical equipment entering an enclosure often wet and conductive on all sides including the floor, and potentially flammable. Equipment selection for such a space is a competent-engineering decision, not a matter of using whatever is on the van.

Biological hazards. Spaces that have held effluent, sewage, standing water or decaying organic material present exposure through contact, ingestion, inhalation of aerosols and contamination of skin, cuts and clothing. I name no organism and no organism-specific control, because that guidance should come from competent occupational-health assessment and the applicable framework, which in Great Britain means the Control of Substances Hazardous to Health Regulations 2002 (as amended) and, for the space itself, the Confined Spaces Regulations 1997.

9. The controls, arranged by the hierarchy of controls

Confined-space controls are usually presented as a list, which makes them look like options of roughly equal standing. They are not. They are ranked, and the ranking is the most useful structuring idea available. The hierarchy of controls is a principle rather than a standard in its own right. It is required by ISO 45001:2018, as amended by Amd 1:2024, at clause 8.1.2, and by ANSI/ASSP Z10.0-2019 at section 8.4 in the United States. It is described by NIOSH on free public pages, which is where the familiar inverted triangle originates. Be precise about attribution: US OSHA does not define the hierarchy in any regulation, and ISO places reorganisation of work inside the engineering tier rather than with administrative controls, so the ISO and NIOSH lists are not identical. The general treatment is in the hierarchy of controls guide; what follows is the confined-space mapping.

Elimination: the most effective confined-space control is not entering

This deserves to be taken seriously rather than acknowledged and skipped, which is what usually happens. If nobody enters, every hazard in this article is eliminated at once: no atmospheric hazard, no engulfment, no entrapment, no rescue problem. Nothing further down comes close, and yet assessments routinely begin by assuming entry and working out how to do it safely.

The realistic options are broader than people assume. Remote inspection by camera, borescope, pole or crawler unit answers most inspection questions from outside, and a great many entries exist only to look at something. Permanent instrumentation brings the reading out to the reader, removing recurring entries rather than one at a time. Sample points let material be taken without anyone going in, and cleaning in place addresses the commonest reason for entry. Relocating a valve, filter or sensor so it is reached externally converts a recurring entry into a routine task. Best of all, designing new plant so entry is never needed is the highest-value intervention available and the one with the shortest window: at design stage external access, cleaning provision and instrumentation cost comparatively little, and after commissioning they cost so much that they never happen.

Substitution, then engineering controls

Where entry cannot be eliminated, ask whether the hazard inside can be replaced with a lesser one: a less hazardous cleaning agent instead of a solvent, a mechanical method instead of a chemical one, a cold method instead of hot work. Substitution is under-used here because the hazard is thought of as belonging to the space rather than to the job, and it is the job that is easier to change.

Engineering controls act on the space and the plant rather than on behaviour, which is why they outrank everything that follows. As principles: isolation, separating the space from every system that could deliver energy, process fluid, gas or liquid into it, and proving that separation; cleaning and emptying, which removes the source directly; purging, noting that an inert purge is itself a hazard-generating control and a purged space is not a breathable one until it has been made so; and ventilation, the workhorse control for atmospheric hazards. Rates, arrangements, inlet siting and verification are specific to the space and the contaminant and belong to a competent assessment rather than to an article.

Administrative controls

These do not remove the hazard; they manage how people interact with it, which is why they sit below engineering controls despite absorbing most of the visible effort. Each is a substantial subject owned elsewhere, so I name rather than teach them: the entry permit, within the wider discipline of the permit to work guide; competence and training, which is a distinct competence for entrants, attendants, supervisors, testers and rescuers rather than one course; the attendant, who exists precisely to compensate for the entrant's inability to detect some hazards; communication that works in the real noise and geometry; monitoring during occupancy, because the atmosphere is a process; and rescue arrangements planned, resourced, practised and matched to the actual space so nobody has to improvise an entry. Roles and rescue in depth belong to the requirements and procedures pillar, and all of it rests on an assessment that identified the hazards correctly: see the hazard identification guide, the risk assessment guide and, for the wider frame, what HSE actually covers.

PPE and respiratory protective equipment, last

The same logic applies to the rest. Protective clothing worn for a chemical hazard worsens heat strain. Gloves worn for a contact hazard reduce grip on a ladder. A harness is part of a rescue system only if there is a system to attach it to. PPE is chosen in full knowledge of what it costs elsewhere, which is the general point made in the PPE guide.

10. Where confined spaces really differ: hazards interact

Assessed one at a time, most of these hazards look manageable. Confined-space work is disproportionately fatal because they are not independent: they amplify each other, and several controls create new hazards while removing old ones.

Heat degrades judgement, which increases error, which increases exposure. A heat-affected entrant is slower to notice a change and more likely to misread it, so every other hazard becomes more dangerous because the person managing it is less capable than an hour earlier and less aware of that than anyone outside. Restricted access converts survivable exposures into fatal ones. The severity of an atmospheric exposure is a function of how long it lasts; access geometry determines how long extraction takes; therefore access geometry determines the outcome of an atmospheric event even though it is not itself an atmospheric hazard. The two are assessed in different sections of most assessments and multiplied together in reality.

Controls create hazards. This is the pattern to watch hardest, because a control is psychologically treated as a solution and stops being examined:

• Purge gas displaces oxygen. An inert purge is an excellent control for a flammable atmosphere and a direct cause of the most lethal atmospheric hazard there is. A space made safe from fire has been made lethal to breathe, and the transition between those states must be managed deliberately.
• Ventilation can import contamination. Forced ventilation supplies whatever is at the inlet, so an inlet sited near a vehicle exhaust, a generator, a vent stack or another work activity turns it into a delivery mechanism. Extraction can equally draw contamination from a connected part of the system into the occupied space.
• Lighting, tools and PPE. The control for poor visibility is electrical equipment inside a possibly flammable, possibly wet, conductive enclosure. Clothing selected for a chemical or thermal hazard reduces heat loss and mobility, increasing heat strain and slowing escape and extraction. Hearing protection worn for noise interferes with the mechanism by which the attendant detects that something is wrong.

The discipline is simple to state and takes rigour to do: for every control introduced, ask what it adds as well as what it removes, and re-examine the space as it will be with the control running rather than as it was when assessed. That is another reason the atmosphere has to be monitored rather than tested once. The controls themselves are changing it.

11. The hazards that are routinely under-assessed

Four categories recur, and they share a shape: the assessment looked at the space in its resting condition, and the harm came from somewhere else.

• Hazards created by the work itself. The assessment describes the tank. The fatality involves the solvent, the welding fume, the engine exhaust or the dust the job brought in. If the assessment does not cover what will be carried through the manway, it has assessed the wrong thing.
• Spaces not recognised as confined. Plant rooms, basements, trenches, pits, ducts, voids, cable tunnels and open-topped tanks get treated as ordinary workplaces because they do not match the mental picture of a manway on a tank. The test is not whether a space looks confined but whether it is enclosed and access-restricted enough for a foreseeable specified risk to arise, and an unrecognised space gets none of the controls.
• Hazards from adjacent and connected systems. The hazard is not in the space and never was. It is in the pipework, drainage, shared header or operation next door, and the space is simply where it arrives. Assessments that stop at the physical boundary miss this entirely.
• Hazards that appear only on disturbance. The space tested clear because the hazard was sealed under a deposit. A clear pre-entry test on a space containing residue is a statement about the air, not about the residue.

One further observation. A substantial share of confined-space harm involves people who were not doing confined-space work as such: someone reaching in, leaning in to retrieve a dropped item, going in for a minute. Duration is not a control, and a partial entry is an entry. Where the hazard exists, putting a head into the atmosphere is the exposure, whether or not the feet followed.

12. Law and standards, with jurisdictions attached

Nothing here creates or interprets a duty. This is only a map of where the duties live, and every entry carries its jurisdiction, because these instruments do not travel.

• United States, general industry. 29 CFR 1910.146, "Permit-required confined spaces". Its scope excludes construction, agriculture and shipyards, a real limit rather than a technicality, and many US states run their own OSHA-approved plans which may differ from the federal text.
• United States, construction. Covered separately by 29 CFR 1926 Subpart AA, "Confined Spaces in Construction", sections 1926.1201 to 1926.1213 (1926.1200 is reserved), final rule made in 2015. Applying the general-industry standard to construction work, or the reverse, is a common and consequential error.
• Great Britain. The Confined Spaces Regulations 1997 (SI 1997/1713), supported by L101, "Safe work in confined spaces", 3rd edition, December 2014. L101 is an Approved Code of Practice, which gives it special evidential status rather than being ordinary guidance. Northern Ireland has separate instruments with different years, so "the 1997 Regulations" is not the right citation there. Related duties sit in the Control of Substances Hazardous to Health Regulations 2002 (as amended) and the Dangerous Substances and Explosive Atmospheres Regulations 2002 (as amended).
• International standards. ISO 45001:2018, as amended by Amd 1:2024, is the certifiable occupational health and safety management system standard and requires the hierarchy of controls at clause 8.1.2. In the United States, ANSI/ASSP Z10.0-2019 addresses the same principle at section 8.4. Neither is law anywhere by itself; both bind through certification and contract. ISO 31000:2018 is guidance and is not certifiable, so there is no accredited organisational certification against it, and IEC 31010:2019 sets out risk assessment techniques. That designation is IEC 31010, not ISO 31010.
• Guidance without regulatory force. NIOSH in the United States publishes descriptive hazard material on free public pages and has no regulatory power.

I quote no clause or regulation text and reproduce no threshold from any of these documents. Read the applicable instrument for your jurisdiction, in its current version, and have a competent person apply it to your space. US OSHA standards and British HSE regulations have no legal force in the Gulf, where they serve as voluntary benchmarks while the binding duties come from national and emirate-level law. And none of the above is satisfied by having a document: a permit system that produces paperwork but does not verify isolation, does not test representatively and cannot mount a rescue is compliant in appearance only.

13. Where systems help, and where they do not

Permit and work-management systems do genuine good on the administrative tier, by holding isolations and competence records and preserving the evidence trail, which is covered in permit to work integration with a CMMS. What no system does is change where any of this sits in the hierarchy. A well-administered permit is still an administrative control, and digitising it does not make an atmosphere safe or shorten the time it takes to get an unconscious person out of a manway. The most valuable thing a system can do here is make the elimination options visible: a register showing how many entries an asset generates each year is the business case for the remote camera or the design change that removes them permanently.

The idea to walk away with

Confined-space hazards are ordinary hazards with their margins removed. Enclosure stops dilution and creates stacked layers of different atmospheres. Restricted access removes escape and stretches rescue, which turns survivable exposures into fatal ones. Two points carry more weight than the rest. An oxygen-deficient atmosphere gives the person in it no reliable warning and impairs their judgement before it stops their movement, so the control must be in place before entry and cannot depend on the entrant noticing anything. And the atmosphere is a process rather than a state: the work changes it, disturbance changes it, conditions change it, connected systems change it, and the controls themselves change it, which is why monitoring during occupancy is a principle and not a refinement. Everything else follows from arranging the controls honestly by the hierarchy, with the uncomfortable conclusion at the top: the most effective confined-space control available is not entering.

Final thoughts

The pattern I would encourage in any operation is to treat the number of confined-space entries as a metric worth reducing rather than a fact of life. Count them. Ask of each recurring entry why it exists, and whether the reason is inspection, measurement, sampling or cleaning, because all four have credible external alternatives. Take the answers to whoever specifies new plant, because access, cleaning provision and instrumentation are cheap at design stage and effectively unavailable afterwards. That is slow, unglamorous work, and it removes more risk than any amount of improvement to the permit form. For the entries that remain, the honest position is the one this article opened with: these hazards are not managed by reading about them, but by trained competent people working to the applicable legal framework for their jurisdiction, assessing the actual space with the actual job in it, with isolation proven rather than assumed, monitoring running rather than filed, and a rescue capability that matches the geometry.

Related reading: Confined space requirements and procedures, The confined space permit, Hierarchy of controls, Lockout/tagout, Hot work hazards and permits, Hazard identification, Risk assessment, PPE, Permit to work, What HSE covers.

Primary sources: US OSHA , UK HSE , NIOSH , ISO .