Issue No. 37 · Autumn Quarter
Metallurgy · Weld Integrity · Failure Analysis
Bench notice: sectioning saw down for bearing replacement Nov 3–5. Hardness rig and metallography unaffected. Expedited jobs quoted case by case.

Cover Story / Autumn · File 37-A

When a Bridge Gusset Lies to You.

Three field samples, two hardness traverses, and a macro-etch that doesn't match the paperwork. What a cracked K-brace on a pedestrian overpass can teach anyone still pulling a stinger for a living.

Read the piece ↓
01We test. We photograph. We write it up in plain sentences.
02No marketing department. Whoever ran the test writes the column.
03Print run of 900. Mailed through the shops that asked.

The Lying Gusset

A county bridge inspector pulled a cracked plate off a pedestrian overpass in late May and shipped it to us in a five-gallon bucket with a torn shipping tag. The paperwork said likely fatigue. The plate said something else, and said it loud enough that we opened a new case file before the coffee was done.

Everything about the sample looked textbook at first. A quarter-inch gusset, A572-50, fillet welds on three sides, roughly nine years in service. The crack ran clean along the toe of the weld, bright on the fresh fracture face, scaled brown on the older portion. You'd call it fatigue from across a parking lot, and you'd write it up that way if you were billing by the hour.

We didn't. Here's why.

What the macro-etch showed

Pris pulled a transverse slice at the crack plane, mounted it in Bakelite, polished to 1µm, and hit it with 2% nital. On a clean repair the etch shows three zones — base, heat-affected, weld — and the boundaries are roughly parallel to the original fusion line. This one showed two HAZs. Overlapping. Offset by about 4mm.

Somebody had rewelded this gusset in the field, on top of the original fillet, and whoever did it never ground the old cap back. The second weld sat on top of a first one the paperwork didn't mention.

What the hardness map showed

Vickers traverse across the HAZ came back ugly. We expected a smooth climb from base metal into weld and back down — call it HV 160 up to HV 220, normal mild-structural behavior. What we got was a spike. HV 340 in a narrow band about 1.5mm off the fusion line, roughly 32 HRC equivalent. That number has no business appearing on a plain-carbon structural plate.

Martensite. Full stop. Somebody had rewelded this gusset cold, in the field, in what was probably a January wind, and walked away without preheat. The untempered martensite that formed in that narrow band had been cracking quietly for nine years before the live load finally found it and finished the job.

"Fatigue was the symptom. The cause was a repair nobody logged."

What to do about it on Monday

If you're the one running the gun on field repairs: preheat the plate. 150°F minimum for anything over 3/8" in A572-50, higher if the ambient is below 40°F or the member is restrained. A rosebud and a surface pyrometer cost less than one deposition.

If you're the one signing off: ask for the repair log and don't accept "it's fine" from a foreman who wasn't there when the arc struck. A field repair without a written WPS addendum is, legally speaking, an unknown.

If you're the one writing the spec: require post-weld hardness verification on any field repair over 3/8". One portable Equotip traverse, five points, ten minutes. It would have caught this in 2015.

The samples

Three gussets in total. One from the failed bent, two from adjacent members pulled for comparison during the same maintenance window. The adjacent pieces were clean — normal HAZ profile, normal grain, peak hardness HV 215. Only the failed plate carried the anomaly. That localization is what convinced us this wasn't a bad heat of steel. It was a bad day on a ladder nine years ago.

Full micrograph set — 50×, 200×, 500× — runs on the back cover of the print edition. Photos are worth more than our opinions.

— R.V., bench log entry 37-A · signed off by D.H.

More From This Issue

Field Notes · p. 11

Tack Welds vs. Skip Welds: Stop Arguing About It

A tack holds position. A skip carries load. If the print doesn't tell you which you're making, the print is wrong — and the shop that lets that print leave the office is, politely, also wrong. Six paragraphs on why this matters when a QC inspector shows up with a set of gauges and an opinion.

6 min · Dell H.

Bench Report · p. 17

7018 Moisture: What 48 Hours in a Pickup Truck Does

We left a sealed can of 7018 and an opened can in the back seat of a work truck through a humid July week in Eau Claire. Then we ran guided-bend tests to AWS D1.1 on coupons welded from each. The opened can failed — root cracks on three of four coupons. The sealed can passed clean. Full numbers and bend photos inside.

9 min · Pris K.

Q&A · p. 22

Porosity, Plain and Cold

Reader question from a structural shop in Duluth: "My short-arc MIG welds look like swiss cheese every time we join galvanized. Why?" Answer, short version: zinc boils at 907°C. You cannot out-amp chemistry. Grind the coating back two inches either side of the joint and the problem goes away. Long version has the spectrograph.

4 min · Dell H.

Back Page · p. 28

The Shop Dog, and Other Quality Systems

Margo is a fourteen-year-old shepherd mix who sleeps under the optical microscope bench next to the compressor vent because it's warm. She has, by our informal count, never once been wrong about whether a sample is going to fail a bend test. A short essay on intuition, experience, and the limits of both.

5 min · R.V.

The Quarterly is mailed, not emailed.

Nine hundred copies, four times a year, to the shops and inspectors who asked to be on the list. The list is currently closed. When a spot opens — and it does, roughly a dozen a year — we hear about it through the trade. A working welder, a fabrication super, or an inspector you already know can put your shop's name forward at the next printing.

Printed and trimmed at a family press two counties over. Saddle-stitched, 32 pages, uncoated cover stock. Issue 37 ships November 15th.

Lab Work

A bridge inspector, a fab-shop super, or an engineer we've signed a report for can put a sample on the intake shelf. Strangers get a polite call and a referral to one of the two NDE shops within a half-day drive. Winter bench load is heavier than summer — plan PQR runs around the freeze.

See what we test →