How anchor bolts fail
An anchor bolt is only half of the anchorage. The other half is the concrete it sits in, and in most failures it is the concrete that lets go, not the steel. Press play to see the three ways an anchorage ends: the bolt breaks, a cone of concrete pulls out, or the bolt slides out of a cone that never forms. Only the first is ductile, and only the first gives any warning.
What to give us when you order
- The embedment depth the engineer has designed to — not the bolt length.
- Grade, and whether the anchorage is required to be ductile, which decides whether steel must fail first.
- Plate washer or hook, size and thickness, or tell us to size it.
- Edge distance and spacing if either is tight, so the drawing can carry a note.
The four modes, and which one you want
The bolt yields, necks and breaks. It stretches first, so the structure deflects and the failure announces itself. This is the mode design codes try to make govern.
A cone of concrete tears out around the anchor, idealised at about 35\u00B0 from the embedded end. No warning, and the capacity depends on embedment depth, concrete strength, edge distance and spacing.
The bolt slides out of the hole. Nothing at the buried end was big enough to mobilise a cone, so bond alone carried it until it didn't. The classic result of leaving the plate washer off.
Close to an edge with deep embedment, the concrete bursts sideways off the face rather than coning upward. An edge-distance problem, not a bolt problem.
What the plate washer is actually for
A plate washer on the buried end of a straight anchor bolt is not a spacer and not a convenience. It is the bearing area that converts tension in the bar into compression in the concrete. Without it you are relying on bond along a smooth bar, which is why a plain straight rod with no end detail pulls out at a fraction of its steel capacity.
The same job can be done by a hook — the L and J shapes — but a hook mobilises less concrete than a plate of the same embedment and is more sensitive to how it is placed. Where the load matters, a plate is the more predictable detail, and it is what most project specifications now call for. You can size either one in Build Your Bolt.
Edge distance and spacing
Cone capacity assumes a full cone can form. Put the anchor near a free edge and part of that cone is missing, so the capacity drops. Put two anchors close together and their cones overlap, so they cannot both develop fully — the group is worth less than the sum of its bolts. Both effects are handled explicitly in ACI 318 Chapter 17 and EN 1992-4, and both are why a bolt that was fine on the calculation sheet fails on a corner of a base plate.
Questions we are asked
Why does the code want the steel to fail first?
Because steel failure is ductile. The bolt yields and stretches well before it breaks, so the connection deforms visibly and load can redistribute. Concrete failure modes are brittle: they give no warning and release suddenly. Designing so that the steel governs is how a code buys warning time into the structure.
Can I just use a longer bolt instead of a bigger plate?
Deeper embedment does increase cone capacity, and it is often the right answer. But length alone does nothing for pull-out if the buried end has nothing to bear on, and deep embedment near an edge can move you into side-face blowout rather than away from failure. Embedment, end detail and edge distance are one decision, not three.
Does galvanizing change any of this?
Not the failure mechanism. It changes the bolt's thread fit, which is its own subject, and on high-strength grades the coater has an embrittlement risk to manage. The concrete side is unaffected.
Who is responsible for the embedment depth?
The engineer of record. MF manufactures the bolt to the length, grade, end detail and finish on the drawing, and will tell you if what is drawn looks inconsistent — but the anchorage design, the concrete strength assumed and the edge distances are the engineer's, and nothing on this page substitutes for that.
Related
Why bolts fail · Installation methods · Torque is not preload · Back to the Bolt Lab