How a Thread Fails and What the Failure Tells You

Why this matters

A broken fastener is the best-documented part of any failed joint. The surfaces record how much load it saw, whether the load was steady or cycling, where the crack started, and whether the metal was already compromised before the machine ever ran. Most of that gets thrown away, because the standard field response is to fit a stronger bolt and move on. A stronger bolt is the right answer to exactly one of the five modes below, and it makes two of them worse.

Before removing the remaining fasteners from a joint that has already lost one: the load path is now carrying on fewer members than it was designed for, so put an independent support under the assembly before anything else comes out and stay clear of the piece rather than beside it. Isolate and relieve stored mechanical energy at the joint under 29 CFR 1910.147 in general industry, or 29 CFR 1910.333(b)(2) where a conductor is involved, since 1910.147 excludes exposure to electrical hazards from work on conductors and equipment in electric utilization installations at (a)(1)(ii)(C). Handle the broken piece with gloves, since fracture edges are as sharp as anything on the truck, and put it in a bag rather than a pocket.

Do not clean a fracture surface you intend to read. Wire-brushing, scraping or solvent-scrubbing removes the evidence and can add marks that look like the ones you are hunting. If you must abrade a plated fastener for any reason, understand the route: grinding or heating a cadmium-plated fastener releases cadmium oxide fume, which is an inhalation hazard and needs a respirator selected and fit-tested under a written program meeting 29 CFR 1910.134 plus local exhaust, not a glove and safety glasses.

The five modes and their signatures

Tensile overload, single event. The fastener necked down before it parted, so the diameter near the break is visibly smaller. The surface is rough and fibrous, often a cup on one side and a cone on the other, with a slanted shear lip around the edge. The threads along the failed length may measure a longer pitch than the same bolt's undamaged threads, because the whole length stretched. This says the joint saw a load beyond the fastener's capacity in one go: an impact, a jammed mechanism, a lift that went wrong, or gross over-tightening at assembly.

Fatigue. The surface is largely flat and smooth over one region, sometimes with fine curved lines like ripples on sand, and then a smaller rough zone where it finally tore through. No necking. The starting point is nearly always a stress concentration: the root of the first engaged thread, the thread runout where the thread meets the shank, or the fillet under the head. Fatigue says the fastener was carrying a cycling load, and the whole point of preload is that it should not have been.

Thread strip. No fracture across the fastener at all. Instead, thread flanks are rolled over, torn, or gone, and there is swarf. Whichever side stripped was the weaker of the pair, which is either a material mismatch or too little engagement. The sibling article on thread engagement owns the arithmetic for that.

Galling, also called cold welding. The fastener seizes part way through run-down at a torque far below its value, and the flanks show smeared and torn metal, often transferred from one part onto the other. The defining fact is timing: galling fails during assembly, not in service. It is common where both parts are the same alloy family and the oxide layer is thin enough to break through, which is why stainless into stainless is the classic case.

Hydrogen embrittlement and stress corrosion cracking. A flat, brittle fracture with no necking and no fatigue markings, at an apparent load well below capacity, and it typically shows up hours to days after installation rather than on the day. Susceptibility is commonly cited as beginning somewhere in the region of the low-to-mid 30s on the Rockwell C hardness scale for electroplated carbon and alloy steel fasteners, but the specific alloy, the plating process and whether a post-plating bake was performed all govern it, so the fastener manufacturer and the plating specification own that answer rather than a remembered number.

The evidence table

Reading a joint: the mount that lost one bolt

A four-bolt steel bracket into an aluminum housing. One bolt snapped, one hole stripped, two bolts still in place. Everything the shop needed was on the parts.

The snapped bolt. Flat, beach-marked over roughly 70 percent of the cross section, with a rough final tear across the remaining 30 percent, initiating at the root of the first engaged thread. No necking anywhere. Thread pitch along its length measured the same as an unfailed bolt from the same box.

The stripped hole. Aluminum swarf around the opening and rolled thread flanks. The bolt that came out of it was undamaged.

The two survivors. Assembly marks put on at the last service were rotated by roughly 30 degrees each. The mating faces showed a reddish-brown powder between them.

Now sequence it. A fatigue break whose beach-marked area is large and whose final tear is small means the nominal stress was low and the cycle count was high. Low static load on a bolt in a clamped joint means the bolt was not carrying much preload, and a bolt with little preload in a joint that keeps moving is the picture of a joint that has already separated. So the fatigue break is late in the story, not the start of it.

The rotated marks and the reddish powder date the start. That powder is fretting debris, which only forms where the mating faces are sliding against each other, and sliding faces mean clamp force had already dropped below what the joint needed. Rotation of the marks says the fasteners were unwinding under that slip.

So the order runs: clamp force fell and the faces began to slip, the slip rotated the fasteners and dropped clamp further, the joint opened and started handing the full cycling load to the bolts, one bolt accumulated enough cycles to fatigue through, its share redistributed to the three that were left, and the aluminum thread, which was the weakest element in the remaining set, stripped.

The counter-check that confirms it. If the bolt had simply been overloaded in a single event, the surface would show necking and a fibrous cup-and-cone, the thread pitch on the failed length would measure long, and the surviving bolts' marks would still line up because nothing would have been rotating. Two independent pieces of evidence, the pitch measurement and the marks, both say that is not what happened.

What the finding changes. Nothing here calls for a stronger bolt, and a stronger bolt would have made it worse: the aluminum thread was already the weakest element, and raising the male grade raises the engagement the female side needs. The fix is on the clamp side, which means restoring preload with matched clean hardware, taking the paint and soft material out of the grip, and putting a steel insert in the housing so the female thread stops being the limiting member. Marks go back on, and the re-check is timed after a full thermal cycle.

Two modes that fail during assembly, and how to stop them there

Galling is prevented in the run-down, not in the design. Slow the run-down, because heat builds with speed and heat is what welds the flanks. Keep threads clean, since debris breaks the oxide film. Use an anti-seize compound qualified for the material pair, and remember that adding compound changes the torque-to-preload relationship, so use a value derived for the lubricated condition rather than the dry one. Where the pair is prone to it, the durable answer is to change the pairing, for example a different alloy or a different hardness on one side.

Embrittlement is prevented at procurement. Where a high-strength fastener is electroplated, the relief bake after plating is part of the specification, and it is not something a field shop can verify by looking. If a joint has failed this way, ordering another of the same part number from the same source is repeating the experiment. Route it to the fastener manufacturer with the failed part and the timing, and consider whether a mechanically-applied coating or a different corrosion strategy fits the environment better.

Checking you got this right

• Match the mode to the timing before anything else. Failed during install points at galling; failed within days points at embrittlement; failed after a long run points at fatigue; failed during a specific event points at overload.
• Measure the thread pitch of the failed part against an unfailed one from the same box. It is a two-minute check and it separates overload from everything else.
• Find the initiation site. A fatigue crack that started at the head fillet rather than the first engaged thread points at bending, which points at a bearing face that is not square or members being pulled into alignment by the fasteners.
• Look at the mating faces and not only at the fastener. Fretting debris and witness marks date the loss of clamp, which is usually earlier than the fracture.
• Keep the failed parts. A bagged fastener with the joint identified on the tag is the only thing that lets a manufacturer or a metallurgist tell you something the field cannot, and it costs nothing to keep.

References

• 29 CFR 1910.147, control of hazardous energy, and 29 CFR 1910.333(b)(2) for the electrical case that 1910.147 excludes at (a)(1)(ii)(C)
• 29 CFR 1910.134, respiratory protection program requirements, where a plated fastener is heated or abraded and metal fume is released
• Fastener manufacturer documentation for plating, post-plating bake requirements, and anti-seize compounds qualified for a given material pair
• Published fastener engineering and failure-analysis references for fracture-surface interpretation and fatigue initiation sites
• See related: Thread Engagement and Why Two Threads Are Not Enough; Why a Bolted Joint Loosens; What a Bolted Joint Is Actually Holding