What a Guardrail Has to Withstand, and Where It Has to Be

Why this matters

A guardrail is the only fall protection that works on a worker who did nothing right that morning. Nobody has to inspect it, don it, clip it, or measure a clearance. It stands there and it protects the tech, the apprentice, the electrician who wandered over, and the customer's maintenance man who was never in your toolbox talk. That is worth a great deal, and it is only true when the rail meets a specific set of dimensions and a specific load applied in a specific direction at the worst point on the run. This card is about those numbers, what they are measured against, and what the rail is bolted to, which is the part nobody sizes.

What passive buys you, and what it costs

Every active system has a human step in front of it. Restraint needs the right anchor and the right tether length for where you are standing. Arrest needs a clearance calculation, a rescue plan and someone who clipped in. Each of those has to go right every time, per person, per position.

A guardrail has one human step and it happens once, at installation. After that its protection is a property of the building rather than a property of anyone's discipline. The cost is that a guardrail is fixed, so it cannot follow an edge that moves, and it can obstruct the work itself, which is the pressure that produces a rail with a section missing.

The dimensions, and which Part sets them

Construction work is 29 CFR 1926.502(b). General industry is 29 CFR 1910.29(b). On the core dimensions they agree, and that agreement is worth knowing so you stop hunting for a difference that is not there.

Element Requirement Construction General industry
Top edge height 42 in plus or minus 3 in above the walking surface 1926.502(b) 1910.29(b)
Top rail strength Withstand at least 200 lb applied within 2 in of the top edge, in any outward or downward direction, at any point 1926.502(b) 1910.29(b)
Deflection limit With 200 lb applied downward, the top edge must not deflect below 39 in 1926.502(b) 1910.29(b)
Midrail Midway between the top edge and the walking surface, withstanding at least 150 lb in any downward or outward direction 1926.502(b) 1910.29(b)
Toeboard At least 3.5 in vertical height, withstanding at least 50 lb in any downward or outward direction, with no more than 1/4 in clearance above the surface 1926.502(j) 1910.29(k)

Two details in the construction text that people miss. Surfaces must be smooth enough not to cause lacerations, punctures or snagged clothing, and steel and plastic banding are not permitted as top rails or midrails. And a wire rope top rail has to be flagged at intervals of not more than 6 ft with high-visibility material, because a taut cable at chest height is invisible from the side.

The tolerance band is not symmetric in what it costs you. A top rail installed at 39 in is inside the 42 plus-or-minus-3 band and legal on the day it is installed, and it has zero deflection allowance left, because any downward deflection at all puts it under the 39 in floor. A rail installed at 45 in is equally legal and has 6 in of allowance. Install toward the top of the band. That is the conservative direction and it costs nothing.

What the 200 lb means, and where it goes

The load is not a design load you compare against a catalog number. It is a performance requirement applied at any point along the top edge, in any outward or downward direction. Two different points on the run govern two different checks, and they are not the same point.

For deflection and rail bending, the worst point is mid-span between two posts, because that is where the rail is least stiff and where the downward 200 lb produces the most drop toward the 39 in floor.

For the post and its base connection, the worst point is at a post, because a load applied directly over a post is carried entirely by that post rather than shared with its neighbour.

That second one produces a number the tech with a tape measure cannot check and should not pretend to. A 200 lb outward force applied at a 42 in nominal top rail acts on a lever arm of 3.5 ft, so the moment the base connection has to develop is 200 lb multiplied by 3.5 ft, which is 700 ft-lb, at a post carrying the load alone. That figure holds only for a load applied at the post; the same 200 lb applied mid-span splits roughly in half between two posts, so each base sees closer to 350 ft-lb while the rail itself carries the bending.

Who owns whether the base can develop 700 ft-lb is not you. It is the guardrail manufacturer's published base detail, or a registered professional engineer for a fabricated or site-built system. Your job is to know that the number exists and to notice when a rail is bolted to something nobody sized: a wooden curb, a deck plate, a stanchion set in a bucket.

Where a guardrail has to be

The trigger height forks by Part and the numbers genuinely differ.

  • Construction, 29 CFR 1926.501(b)(1): unprotected sides and edges at 6 ft or more above a lower level, protected by a guardrail system, a safety net system, or a personal fall arrest system.
  • General industry, 29 CFR 1910.28(b)(1): unprotected sides and edges at 4 ft or more above a lower level, protected by a guardrail, a safety net, a travel restraint system or a personal fall arrest system.

Beyond the plain edge, both Parts reach specific features. Holes in a walking surface, at 1926.501(b)(4) and 1910.28(b)(3), which includes skylights and floor openings. Ramps, runways and elevated walkways. Hoist areas, where a guardrail section removed for a lift has to be replaced by a chain, a gate or a personal fall arrest system while it is out, not by attention. Wall openings tall and wide enough to fall through. Wells, pits and shafts at 1926.501(b)(7), and note the excavation work itself falls under 29 CFR 1926 Subpart P, which has no general-industry counterpart at all. And 1926.501(b)(8) reaches falls into dangerous equipment regardless of height, which is the paragraph people forget when the drop is short and the thing at the bottom is not.

What takes a guardrail off the table

Four conditions, and only one of them is about cost.

The edge moves. Decking, forming and any operation where the edge advances as material is placed cannot be guarded by a fixed rail that is always in the wrong place. That is the leading-edge case and it is covered in a sibling card.

The work is on the rail line. Installing or removing the guardrail itself is exposure at an unprotected edge, and the protection during that work is a separate decision.

The structure cannot take the reaction. A parapet that will not develop the base moment, a membrane roof with no penetration allowance, a deck plate over an unknown structure. This is an engineering answer, not a judgement.

The rail would be climbed. A rail placed where crews have to get over or through it to work produces a rail with a section removed and nobody noticing. That is not a reason to skip the rail; it is a reason to design the access into it.

Worked example: the acceptance record for one mezzanine edge

A warehouse mezzanine, 9 ft above the finished floor, in a facility doing general industry work. The 9 ft drop is over the 4 ft trigger in 1910.28(b)(1). Welded steel post-and-rail, posts at 8 ft on centre, bolted base plates.

How the load was applied, before any reading is worth anything. The outward 200 lb was applied with a strap and a hand-held spring scale from a stance more than the tester's own height back from the edge, so that a rail failure could not put the tester in the fall path. Where a point could not be reached that way, the load was applied with a come-along anchored to a fixed column on the protected side, with nobody standing between the rail and the edge. Nobody on this test was tied off to the guardrail, because using the thing under test as the anchorage for the person testing it is the one connection that cannot be permitted.

Field                              Reading      Check
Location                           Mezz east edge, bays 4-6
Drop to lower level                9 ft over 4 ft trigger, 1910.28(b)(1)
Top edge height, measured          43.5 in within 42 +/- 3
Deflection allowance remaining     4.5 in       43.5 - 39
Midrail nominal (43.5 / 2)         21.75 in     "midway"
Midrail measured                   21.0 in      0.75 in low, within "midway"
Toeboard height                    4.0 in over the 3.5 in minimum
Toeboard clearance at floor        3/16 in under the 1/4 in maximum
Toeboard openings none solid plate

200 lb outward, at a post held no visible movement at base
200 lb outward, at mid-span held no separation at rail splice
200 lb downward, at mid-span top edge fell 43.5 to 41.75 in
  deflection                       1.75 in
  final height                     41.75 in above 39, pass by 2.75 in
150 lb outward, midrail mid-span held

Base moment demand, re-based to
  the measured rail height:
  200 lb x (43.5 in = 3.625 ft)    725 ft-lb at a post, not mid-span
  Owner of this check manufacturer's base detail / engineer

Two lines need their reasoning said out loud.

The base moment is a re-basing, not an addition. The general section computed 700 ft-lb at a 42 in nominal rail. This rail is at 43.5 in, so the lever arm changed and the figure becomes 725 ft-lb. It does not get added to the 700; the 700 was never a separate load. Getting that backwards inflates a demand by a factor of two and sends someone chasing a base detail that was never in question.

The deflection result is the one that would have failed on a different day. This rail passed with 2.75 in of margin because it was installed at 43.5 in. The identical rail, identically built, installed at 39.5 in, would drop to 37.75 in under the same 200 lb and fail, having been perfectly legal on the tape measure. The height reading and the deflection reading are not two independent checks; the first one sets how much of the second you can afford.

The failure mode this record catches is the rail that everyone assumed was fine because it looks like a guardrail. The most common real defects are a splice that separates under outward load, a toeboard sitting on the floor with a gap that lets a dropped fitting through, and a base plate anchored to a deck nobody sized. None of those are visible from a walk-by, and all three show up the moment 200 lb is applied at the right point.

How to verify you got this right

Verify at the worst point, not the convenient one. Walk the run and pick mid-span between the two most widely spaced posts for the deflection check, and the post with the least substantial base for the outward check. A test at a corner post next to a column proves nothing about the 8 ft bay in the middle.

Measure the top edge before you load it and again after, at the same point, with the tape held to the same surface. The deflection number is a difference between two readings from one tape, so a consistent offset in how you hold it cancels out; what does not cancel is measuring from the deck in one reading and from the toeboard in the other.

Write the acceptance record with the measured height on it, not the nominal one. A record saying "42 in guardrail, passed" tells the next person nothing about how much deflection allowance is left, and that allowance is the thing that changes as a rail ages, a splice loosens, or a post base works in its holes.

References

  • 29 CFR 1926.502(b) for construction guardrail systems and 29 CFR 1926.502(j) for toeboards.
  • 29 CFR 1910.29(b) for general industry guardrail systems and 29 CFR 1910.29(k) for toeboards.
  • 29 CFR 1926.501(b)(1), (b)(4), (b)(7) and (b)(8) for where guardrails are required on construction work; 29 CFR 1910.28(b)(1) and (b)(3) for general industry.
  • See related: universal-why-a-leading-edge-is-treated-differently and universal-what-a-hole-cover-has-to-do-and-what-marks-it.