What Fall Restraint Is, and Why It Is the Better Answer

Why this matters

Fall arrest is the last thing in the order, not the first. A harness and a lanyard do not prevent a fall; they change what happens after one, and they bring a clearance calculation, a rescue clock and a set of forces that all have to be right. Restraint removes all of that by making the fall impossible to start. It is simpler in equipment and simpler to verify, and the verification is a tape measure rather than a chart. This card is about the one gate that decides whether a rig is genuinely restraint, and what to do on the day it fails.

Where restraint sits in the order

Four tiers, and the harness is in the last one.

  1. Eliminate the exposure. Do the work from the ground, from inside, from a lift, or by bringing the equipment down. No exposure, no system to get right.
  2. Guard the edge passively. A compliant guardrail protects everyone on the surface without anyone doing anything correct that morning. Covered in a sibling card.
  3. Restrain. Tether the worker so the body physically cannot reach the edge.
  4. Arrest. Let the fall happen and stop it inside the available clearance, then rescue before the suspension clock runs out.

Tiers 3 and 4 use similar-looking hardware, which is why they get confused. They are not variations of one system. They have different acceptance criteria, different forces, different failure modes, and only one of them needs a rescue plan.

What restraint is, in equipment terms

General industry defines it. Under 29 CFR 1910.140, a travel restraint system is an anchorage, an anchorage connector, a lanyard or other means of connection, and body support, used to eliminate the possibility of the worker going over the edge. Three details follow from that definition and each one changes what you buy.

The connection must be fixed in length. An energy-absorbing lanyard gets longer when the pack tears, and a self-retracting device pays out. Either one can turn a rig that measured short into a rig that reaches. Restraint uses a fixed-length lanyard or a positioning device with a locked adjustment.

Body support may be a belt in general industry. 29 CFR 1910.140 prohibits a body belt as part of a personal fall arrest system and permits a body belt as body support in a travel restraint system, because a restraint belt never has to distribute an arresting force. Construction work is the tighter case: 29 CFR 1926.502(d) prohibits body belts for fall arrest, and 29 CFR 1926.502 does not define a travel restraint system under that name, so a construction crew running restraint should expect to justify the configuration against the arrest criteria and should equip accordingly.

An anchorage is not a fixture anchor. This library carries several articles on concrete and hollow-wall anchors for hanging equipment. Those are rated for a static hanging load in a defined base material, in a defined embedment, and they carry no fall-protection rating whatsoever. A fall protection anchorage is a different subject with a different basis: 29 CFR 1926.502(d) sets 5,000 lb per employee attached for construction, or a design with a safety factor of at least two under the supervision of a qualified person, and 29 CFR 1910.140 sets the general-industry counterpart. Never tie a lifeline to a fixture anchor because the label said anchor.

Anchorage rating for restraint, in practice. Rate it to the arrest criterion anyway. Restraint theoretically sees only the pull of a worker leaning against the tether, which is a small fraction of an arrest load, but a restraint anchor is a piece of steel on a roof that someone will clip a shock-absorbing lanyard to on a Tuesday you are not there. Rating it to the arrest criterion costs nothing at the design stage and removes the whole failure path.

The gate

With the system rigged exactly as it will be used, can any part of the worker's body reach any unprotected edge?

That is the whole test. If the answer is no with margin, it is restraint and there is no clearance calculation, no arrest force, no swing fall and no rescue clock. If the answer is yes, or you cannot demonstrate no, it is an arrest system whether you called it one or not, and it has to meet every arrest criterion.

Two words in the gate do the work. Any part of the body, which is not the same as the feet under the D-ring. And any unprotected edge, which means the shortest distance to the nearest one in any direction, not the distance to the edge you happen to be facing. Skylights, floor openings, hatches and shaft openings are edges.

Measuring the reach, term by term

Every term gets measured and every term rounds up, because rounding a reach down is the direction that puts a worker past the edge.

  • Anchorage connector length. The sling, strap or beam clamp between the anchorage and the connecting device. Measure it in its installed loop, not flat.
  • Connecting length. The labeled length of the fixed lanyard.
  • Body lead. How far the body can get ahead of the dorsal D-ring. A worker facing away from the anchor with the tether trailing puts their boots forward of their back. Measure it on your own crew, in gear, and round up.
  • Anchorage travel. If the anchor connector can slide along a beam, a pipe or a cable before it bears on something, that slide is added reach. A horizontal lifeline adds its own sag and its own travel and is a separate design question for a qualified person.
         plan view, looking down at the roof

  ================== unprotected edge ==========

              distance to nearest edge
                        |
                        v
              . . . . . . . . . .
            .                     .
           .      anchorage         .
          .           point          .
           .                        .
            .                     .
              . . . . . . . . . .

     reach circle radius = anchorage connector
     + lanyard + body lead + anchorage travel

     the gate is the gap between the circle and
     the NEAREST edge, in any direction

One gate, two roofs

Same crew, same rig, same afternoon, opposite answers.

The rig, measured once:

Anchorage connector (sling, measured in loop)   2.0 ft
Fixed-length lanyard (labeled)                  6.0 ft
Body lead (measured on this crew, rounded up)   1.0 ft
Anchorage travel (sling can slide to the
  beam fitting before it bears)                 0.5 ft
                                                ------
Effective reach                                 9.5 ft
Margin policy (this shop's number, not a
  published one)                                2.0 ft
Reach plus margin                              11.5 ft

Roof one. The anchor is a permanent roof anchor 18 ft from the nearest roof edge, with no openings closer. 18 ft against 11.5 ft leaves 6.5 ft of gap. The gate passes. There is no clearance calculation to do, no rescue plan to time, and the acceptance record is a tape measure reading and a photograph of the rig.

Roof two. Two hours later the same tech services a unit on the far side, clips to a different roof anchor, and starts work. That anchor is 22 ft from the roof perimeter, which sounds better than the first one. It is 7 ft from a skylight.

A skylight is a hole in the walking surface, and 29 CFR 1926.501(b)(4) treats holes as fall hazards in construction work, with the general-industry counterpart at 29 CFR 1910.28(b)(3). Skylight glazing is not a walking surface and a screen or dome is not a cover unless it meets the cover criteria. So the nearest unprotected edge is 7 ft, not 22.

9.5 ft of reach against 7 ft to the edge means the worker can get 2.5 ft past it. The gate fails, and it fails before the margin is even applied.

What the failure produces, which is the point of running a gate rather than making a judgement. Solve for the lanyard that would fix it. Reach plus margin must land at or under 7.0 ft. The terms that are not the lanyard total 2.0 plus 1.0 plus 0.5, which is 3.5 ft, plus the 2.0 ft margin, which is 5.5 ft. That leaves 1.5 ft of lanyard, and a 1.5 ft tether is not a working length for someone servicing a unit. Restraint at that anchor is not achievable, and no amount of care makes it achievable.

So the answer is one of three engineering moves, in order: relocate or add an anchor so the nearest edge is farther than 11.5 ft, cover the skylight with a cover that meets the criteria in the related card so it stops being an unprotected edge, or convert to arrest with equipment selected for the anchorage height and a clearance calculation that closes.

What the third option costs, stated so nobody picks it casually. The fixed-length lanyard that was correct for restraint is the wrong equipment class for arrest. It has essentially no energy absorption, so a fall on it produces an arresting force set by the stretch of the webbing and the body, which is well above the 1,800 lb maximum arresting force that 29 CFR 1926.502(d) allows with a body harness. The manufacturer publishes the tested arrest force for their product and that number governs. Converting to arrest means different hardware, a clearance stack that the anchorage height controls, and a rescue plan with a measured time on it.

The failure mode in the field is not that someone reasons through this badly. It is that nobody re-runs the gate when the anchor changes. The rig was verified once, in the morning, at the first anchor, and the verification was treated as a property of the equipment rather than a property of the position. Restraint is a statement about where you are standing.

How to verify you got this right

Verify at the anchor, not at the truck, and verify by walking rather than by reasoning.

Clip in, then walk the tether taut in a full circle around the anchor with the tether trailing behind you, the way you would if you turned your back on it. Watch your boots, not the tether. If your boots come within the margin of any edge, opening, skylight or hatch, the rig is not restraint at that anchor. This takes under a minute and it catches the two things a tape measure misses: an edge you did not know about, and the extra reach you get from turning around.

Record the reading. A restraint acceptance record is one line: anchor identifier, measured reach, measured distance to the nearest edge in any direction, date, name. If the anchor changes, the line is void and a new one gets written. An acceptance record with no anchor identifier on it is a record of nothing.

References

  • 29 CFR 1910.140, personal fall protection systems for general industry, which defines a travel restraint system, prohibits body belts as part of a personal fall arrest system, and sets the general-industry anchorage criteria.
  • 29 CFR 1926.502(d), personal fall arrest systems for construction: 5,000 lb per employee attached or a design with a safety factor of at least two under a qualified person's supervision, and the 1,800 lb maximum arresting force with a body harness.
  • 29 CFR 1926.501(b)(4) for holes on construction work and 29 CFR 1910.28(b)(3) for the general-industry counterpart; a skylight opening is a hole.
  • See related: universal-what-a-hole-cover-has-to-do-and-what-marks-it, universal-what-a-guardrail-has-to-withstand-and-where-it-has-to-be, and universal-why-suspension-trauma-makes-rescue-a-timed-event for the clock restraint avoids.