Why a Joint Fails at the Interface and Not in the Metal

Why this matters

Almost nobody breaks a pipe. They pull an anchor out of a ceiling, strip a thread, crack a solder fillet, or push a gasket out of a flange. The member itself is usually the strongest thing in the assembly, and it is the one everybody sizes, because it is the one with a published number on it. The connection is what fails, and it fails at the interface: the small area where two things actually touch, or the thin film between them. Size a connection against the strength of the parts it joins and you will be wrong in the direction that drops a filled pipe run on somebody.

Before opening any fastened joint: a bolted joint holds energy even when nothing is flowing through it. Rig or block the member so it is supported independently before you release a hanger, a bracket, or the last two bolts, and stand out of the arc it will swing along. Where the joint contains fluid, gas, or refrigerant, isolate the source and confirm zero on a gauge at the joint itself rather than upstream of a valve that may not be holding. 29 CFR 1910.147 requires stored energy to be relieved, disconnected, or restrained before servicing, and a pressurized or loaded joint is stored energy. Where the connection is an electrical termination, that is a different standard: 1910.147(a)(1)(ii)(C) excludes electric utilization work, so de-energize and lock out under 29 CFR 1910.333(b)(2) and prove dead live-dead-live per NFPA 70E-2021, 120.5, which reaches you through your employer's electrical safety program rather than on its own authority.

A connection is a chain, and the chain runs into something nobody sized

Draw the load path for any connection as an ordered list of elements the force passes through, from the thing being held to the thing doing the holding. For a supported pipe: pipe wall, clamp, rod, nut, and then the anchor, and then the base material the anchor sits in. For a bolted flange: gasket, flange face, bolt shank, nut, and the flange neck behind it. For a soldered fitting: tube wall, the filler in the annular gap, cup wall.

The capacity of that chain is its smallest element, and in field work the smallest element is reliably the one at the far end, past where the catalog stops. Rods, clamps, bolts and pipe all come with load tables. Concrete, hollow block, wood decking, sheet metal ductwork and thirty-year-old plaster do not, and neither does the quality of the film in a soldered gap. So the elements that get checked are the ones that were never going to fail.

Three structural reasons the interface, not the metal, is the small link:

• Contact area is a fraction of the member section. A tube has its whole wall in tension. The soldered joint holding it carries the same load through a thin annulus of filler. The filler is weaker per unit area and there is less of it, and both effects run the same direction.
• The load turns a corner. Tension in a rod becomes tension plus prying in an anchor, and prying multiplies the force at the anchor head above the force in the rod. A connection that changes the direction of the load has amplified it somewhere.
• Capacity is set by something invisible after assembly. Preload in a bolt, wetting in a brazed gap, embedment depth in a drilled hole, cure in an adhesive. All four are decided during assembly, none can be read afterwards from the outside, and all four are the actual capacity.

The three interface families

The middle row is the uncomfortable one. On a bonded joint the only thing visible after assembly is the outer fillet, which is the part that did the least work. A sibling article covers reading that fillet in detail; here it is enough to know that visual acceptance of a bonded joint accepts the geometry, not the bond.

Worked example: a load-path ledger for one hanger

A run of 4-inch steel pipe, water filled, on 10-foot hanger spacing. From the pipe schedule table and the inside diameter, that pipe runs about 16 pounds per foot filled, so a plain-run support carries roughly 160 pounds. One hanger sits under a valve and its manifold, and by the layout it takes about half again as much. Call that hanger's load 1.0 unit. Every number below is that hanger's load expressed as a multiple of 1.0, which is the only form that lets four different components with four different rating conventions be compared on one page.

Each capacity below comes from a published table for that specific component, not from judgment, and the point of the exercise is which of them turned out to have no table at all.

Read only that far and the anchor looks like the strongest element in the assembly. It is the weakest, and the four corrections that get it there are all conditions the test slab did not have.

• Ultimate becomes a usable capacity. The 4.0 is an ultimate tested to failure, and it cannot be used as a working load until it is converted to the basis your design uses. Under ACI 318 Chapter 17 the report gives a nominal strength, reduced by a strength-reduction factor and compared against factored loads; where a report instead publishes an allowable-stress value it states its own conversion factor, 4 in this case. Whichever it is, name it, because the whole point of the ledger is that a number's basis travels with it. On this report the conversion lands at 1.00.
• The concrete is cracked, not uncracked. Hangers went into the underside of a suspended slab, which is the tension face at midspan, so the report's cracked-concrete row applies, not the uncracked one. That row is about 0.7 of the uncracked value here. 1.00 x 0.7 = 0.70.
• The embedment actually drilled is the shallow one. The deck was thinner than the detail assumed, so the crew drilled to the report's shallow embedment rather than its deep one. The shallow row is roughly 0.55 of the deep row. 0.70 x 0.55 = 0.385.
• One anchor landed close to a slab edge. The report's edge-distance reduction for that spacing is about 0.8. 0.385 x 0.8 = 0.31.

The anchor delivers about 0.31 of the load it is carrying, against 2.5 for the next weakest element. The weakest element is about eight times weaker than the one above it, and it is the only one nobody looked up.

Now the part that explains the timeline. An allowable of 0.31 with a factor of safety of 4 behind it means an ultimate of about 1.24 times the static load. The run does hold. It held through fill, through the hydrostatic test, and through eleven months of service, because 1.24 is greater than 1.0 and nothing was asking it for more. It let go the first time the system was drained for a repair and refilled with a fast-opening valve, because a surge asks for more than static, and the margin between 1.24 and the static load is not enough to absorb one.

That is the general shape of an interface failure. It does not fail when you build it, so the build feels validated. It fails at the first event that is not steady state.

What would move the weak element somewhere else

If the base material were solid grouted masonry rather than a cracked thin slab, the cracked and shallow-embedment corrections both come off and the anchor lands near 1.0, which is still the weak element but no longer the reckless one. If the support were a beam clamp on structural steel, the base material stops being the limit and the clamp's own table becomes the smallest number, which you can then read. If the detail called for a through-bolt with a plate on the far side of the deck, the concrete stops carrying the load in breakout at all and the ledger's last row becomes the bolt. And adding a second anchor next to the first does not simply double the number: below the report's minimum spacing the two anchors share one breakout cone and each one loses capacity, which is the specific way a well-intentioned field fix makes the ledger worse.

Do not drill the test holes for this exercise without dust control. Drilling or cutting concrete and masonry releases respirable crystalline silica; use a shrouded bit with dust collection or wet methods per 29 CFR 1926.1153 on a construction site or 1910.1053 in general industry, with respiratory protection supplied under a written program meeting 29 CFR 1910.134 where the control alone does not get you there. A glove does nothing for this one, because the route is inhalation.

How to verify you got this right

Do this on a support you have already built, not on a drawing.

1. Write the elements in order, from the load to the structure, and do not stop at the last component with a part number. If your list ends at the anchor rather than at the base material, you have not finished the list.
2. Normalize. Put the applied load at 1.0 and convert every published capacity to a multiple. Mixing an ultimate, an allowable, a working load limit and a design load on one page in their native units is how a 4.0 gets compared to a 2.5 and wins.
3. For every number, name where it came from and what conditions it was measured under. A capacity that came out of a catalog page carries the geometry of the test that produced it: crack state, embedment, edge distance, temperature. If you cannot name those conditions, that number is not usable yet.
4. Flag the element with no source. There will be one. It is the base material, the bond, or the preload, and it is your answer. Route it to whoever owns it: the anchor manufacturer's evaluation report for a base-material capacity, the engineer of record for the structure's ability to accept the load at that point, or the adopted code for spacing and support intervals. Those three own different questions and one cannot answer for another.
5. Re-read the ledger bottom-up. The bottom of a support list is where the shortcuts get taken, because that is the end that is drilled last, in the worst position, by the person on the ladder.

References

• 29 CFR 1910.147, control of hazardous energy, for relieving and restraining stored energy before opening a loaded or pressurized joint; 29 CFR 1910.333(b)(2) for electrical terminations, which 1910.147(a)(1)(ii)(C) excludes
• 29 CFR 1926.1153 (construction) and 29 CFR 1910.1053 (general industry), respirable crystalline silica, and 29 CFR 1910.134 for the respiratory protection program
• MSS SP-58, pipe hangers and supports, in the edition adopted by the mechanical code your authority having jurisdiction enforces or by your project specification
• ACI 318 Chapter 17, anchoring to concrete, as adopted through the building code in force locally, together with the anchor's own evaluation report for crack state, embedment and edge-distance factors
• See related: What Torque Actually Controls and What It Does Not; Thermal Expansion Mismatch Inside a Joint