What Cracking Tells You About When It Happened

Why this matters

Weld cracks are named for the temperature they formed at, not the shape they ended up as, and that is the single most useful thing to know about them. A crack that opened while the last of the pool was still liquid, a crack that opened during a later heating, and a crack that opened cold two days after everyone went home are three different failures with three different causes and three different fixes. They can all end up looking like a line in a weld. So the first question at a cracked joint is not where it cracked, it is when, and the joint carries enough evidence to answer that without a laboratory.

Before touching a cracked load-carrying member: a crack in a member under load is a live structural hazard, so stop work under it, keep people out from under it, and get the member unloaded or independently supported by the party responsible for the structure before any grinding, gouging or welding begins. Removing metal from a cracked member while it carries load reduces the section that is already compromised. Once the member is secured, gouging and grinding are their own exposures: air carbon arc gouging puts the work inside a hearing conservation program under 29 CFR 1910.95 and generates a heavier fume load than welding, and grinding releases base metal and coatings as respirable dust, both inhalation routes needing capture at the tool or local exhaust at the arc plus a respirator issued under a written program meeting 29 CFR 1910.134. Coatings decide the constituent: hexavalent chromium from stainless and chromate primers under 29 CFR 1910.1026 or 29 CFR 1926.1126 in construction, lead from old paint under 29 CFR 1910.1025 or 29 CFR 1926.62. The repair weld carries the arc's ultraviolet at the minimum protective shade from 29 CFR 1910.133(a)(5), and its fire prevention and fire watch belong to the hot work permit under 29 CFR 1910.252(a) or 29 CFR 1926.352.

The regimes, and the evidence that separates them

Solidification cracking, at the end of freezing. While a weld is freezing, the last liquid to solidify sits as a thin film along the boundaries between growing grains, and it is enriched in whatever melts at the lowest temperature - sulfur and phosphorus above all. If the joint is pulling itself apart faster than that film can feed, the film tears. It is a hot tear in a casting, happening in the middle of a structural member. It is favoured by a deep narrow bead, by high restraint, by a crater filled too quickly, and by dilution from a parent metal carrying free-machining sulfur, which the article on what a weld actually is covers as a dilution problem.

Hydrogen-assisted cracking, cold and on a delay. Below roughly 300 degrees F, and most actively near ambient, dissolved hydrogen migrates to the highest-stress region of a hard transformed structure and cracks it. It needs three things at once - hydrogen, a susceptible hard microstructure and tensile stress - and removing any one of them stops it. That is what the preheat card is about. Its defining property is the delay: hours to days after the arc is out, commonly inside the first two days.

Reheat cracking, during a later heating. It is absent from the table below on purpose, it is separated by the event that caused it rather than by evidence on the joint, so a reheat step in the history is the cue. Certain low-alloy steels crack in the coarse-grained heat-affected zone when they are heated again, either in a post-weld heat treatment or in high-temperature service. It is intergranular and it happens at temperature, not at the arc and not cold.

Lamellar tearing, which is keyed to direction rather than to temperature. Rolled plate carries flattened non-metallic inclusions parallel to its surface. Put a heavy tee or corner joint on it with enough restraint to strain the plate through its thickness, and the plate delaminates along those planes in a stepped, terraced tear below the weld. The crack is in the parent plate, not in the weld or its heat-affected zone.

Reading the joint

The two cues that do the most work are timing and fracture face colour. A face that opened at high temperature oxidises immediately and comes out dark; a face that opened cold in a shop stays bright. Photograph the face before anyone wire-brushes it, with capture at the tool if a powered brush is used, because that colour is evidence and it does not survive cleaning.

The crater is its own case

A crater is a small, deep, heavily restrained pool with the worst possible shape for feeding, and it freezes last with nothing left to feed it. A star-shaped crack in a crater is solidification cracking every time, it is visible on the day, and it is prevented by filling the crater rather than by any change to the rest of the weld. Shops that treat crater cracks as a general cracking problem go looking for hydrogen and find nothing.

A bracket that passed on Friday and was cracked on Monday

A shop welded four heavy brackets onto a 1 inch plate frame on a Friday. Fillet welds both sides, plain carbon structural steel, covered electrodes. Shop temperature 60 degrees F. Visual inspection was carried out about 4 hours after the last weld and everything passed. On Monday morning, one bracket carried a crack running along the toe on the flange side.

The timing eliminated the first candidate on its own. Solidification cracking is visible on the day. The welder and the inspector both looked at that toe at roughly 4 hours and nothing was there. Something that appears between hour 4 and hour 62 did not form while the metal was liquid.

The fracture face confirmed it. The crack was ground open enough to expose the face, and the face was bright and flat. A face that opened while the joint was hot would have oxidised in the first minutes. Bright means cold.

Position eliminated the third candidate. The crack ran along the weld toe in the transformed band beside the deposit, not below the joint in the plate and not stepped. Lamellar tearing sits in the parent plate parallel to its surface, and this was not that.

Load was eliminated by the record. Nothing was hung on the frame over the weekend and the shop crane was locked out from Friday afternoon. There was no service stress event to attribute it to.

What was true, counted from the shop's own records. Three conditions lined up, and hydrogen-assisted cracking needs exactly the three:

• Hydrogen present. The covered electrodes had come out of the holding oven on Wednesday morning and were logged back in Friday evening, which is about 59 hours of continuous atmospheric exposure, counted as elapsed hours out of the oven, not shift hours. Low-hydrogen coverings absorb moisture from shop air, and the exposure limits and any rebake belong to the electrode's own AWS A5 classification and to the exposure table in AWS D1.1 in the adopted edition. 59 hours is far outside any of them.
• A hard microstructure available. 1 inch plate at a 60 degree F starting temperature is a substantial heat sink, and the heat-affected-zone card explains why joint mass and starting temperature, rather than technique, set how fast the band cools.
• Tensile stress. Four brackets welded both sides on a stiff frame is a restrained joint, and weld shrinkage puts the toe in tension without any external load at all.

No preheat was used, and no preheat had been specified. The preheat card owns that determination, which needs the steel's chemistry, the combined thickness at the joint and the consumable hydrogen level together, and none of the three had been looked at.

Correction, printed: the 4-hour inspection was not early against a rule, it was early against the mechanism. AWS D1.1's delayed visual inspection requirement, in the edition the specification invokes, applies to certain quenched and tempered steels and requires a wait of at least 48 hours. This plate was not one of those grades, so the 4-hour inspection breached nothing. That is the trap. The 48-hour figure exists because the mechanism takes that long, and the mechanism does not check which grade triggered the code requirement. Inspecting at 4 hours against a mechanism running out to about 48 saw roughly the first twelfth of the window in which the crack could form.

Check against the sibling rules, with the figures printed. The porosity card states that a clean inspection for pores is not evidence of low hydrogen; this case had no reported porosity and the hydrogen finding rests on the 59-hour electrode exposure, the 1 inch section at 60 degrees F and the restraint, not on the absence of pores. The preheat card states that a preheat value requires chemistry, combined thickness and hydrogen level together; this case names the thickness (1 inch), names the hydrogen condition (59 hours out of oven) and does not name the chemistry, so it specifies no temperature and routes the determination. The heat-input card states that arc energy compares two procedures at constant thickness and geometry; no energy figure is quoted here and none is used to explain the crack.

How the repair was confirmed. The two affected brackets were gouged out and rewelded to a procedure that specified a preheat and interpass temperature set by the engineer against the actual plate chemistry, using electrodes drawn from the oven that shift. Each repair was re-inspected at 48 hours and again at 7 days, and neither showed a crack at either check. Inspecting once at 48 hours would have been the minimum defensible check; the 7 day look was cheap and it is the one that makes the result an observation rather than a hope.

What would flip the conclusion. If the crack had been present on Friday at the 4-hour look and had simply grown over the weekend, the timing evidence collapses and the answer moves toward solidification cracking with the fracture face as the tiebreaker. If the plate had been a thin section in a warm shop, the hardening condition largely disappears and the same 59 hours of electrode exposure would very likely have produced nothing. Hydrogen alone does not crack a joint; it is one of three conditions, and a shop can break the chain wherever it is cheapest to break it.

How to verify you read the timing right

• Establish when the joint was last confirmed sound, with a name and a time. "It looked fine" without a time makes the timing evidence unusable, and timing is the strongest evidence available.
• Photograph the fracture face before any cleaning. Colour is the second-strongest cue and a wire brush destroys it in one pass.
• Write down which of the three hydrogen conditions you can prove and which you are assuming. A finding that names hydrogen without naming a hydrogen source is a guess, and consumable exposure logs are usually the easiest of the three to evidence.
• Set your own re-inspection at 48 hours on any restrained, hardenable, heavy-section joint, whatever grade it is and whether or not the code requires a delay for it, and record that the check happened.
• Do not accept a crater crack as evidence of a shop-wide hydrogen problem. Look at the crater technique first; it is a different regime with a different fix.

References

• AWS D1.1 structural welding code for steel, in the edition invoked by your project specification or the adopted building code, for its delayed visual inspection provision for certain quenched and tempered steels and its consumable atmospheric exposure tables
• AWS A5 series filler metal specifications, which set the hydrogen designator, storage condition and rebake requirements for the electrode in use
• 29 CFR 1910.95 for noise during gouging; 29 CFR 1910.134 for respiratory protection; 29 CFR 1910.1026 and 29 CFR 1926.1126 for hexavalent chromium; 29 CFR 1910.1025 and 29 CFR 1926.62 for lead
• 29 CFR 1910.133(a)(5) for filter shade; 29 CFR 1910.252(a) and 29 CFR 1926.352 for hot work fire prevention on the repair
• See related: Why Preheat Exists and What It Is Controlling; What the Heat-Affected Zone Is and Why It Governs; Why Porosity Appears and What It Tells You