Cracks between holes and in knuckles

Some cracks are easy to explain once mapped: a fatigue crack at the toe of a poorly supported attachment, for instance. Two kinds are harder, and the 1919 Airco booklet spends pages on both. The first runs between holes, across the narrow ligaments of plate that separate them. The second runs along a knuckle, the curved flange where a flat plate turns to meet another. Both are signs that a region of plate is being strained, cycle after cycle, more than it can tolerate.

This chapter looks at what those cracks mean, how the booklet closed them, and why a modern repair has to ask a structural question the booklet barely raised.

Ligaments concentrate strain

Chapter 6 introduced ligament efficiency. A row of holes is a row of narrow necks in the plate, and every neck has a stress concentration at its edges. When the plate is strained repeatedly, by pressure, by temperature or by flexing, the necks see the highest local strain. A crack that starts at one hole’s edge has only a short way to travel to reach the next hole. Once a ligament has cracked, its neighbours take more of the load and crack sooner.

That is why the booklet was cautious. It warned that cracks between staybolt holes are not easily welded, because proper allowance for expansion is hard to make, and that a side sheet with many such cracks would be better replaced. A run of cracked ligaments is a region failing, not a collection of separate small defects.

The booklet’s method for closing holes

Where welding was used, the booklet described a careful sequence for a crack running through holes that were to be closed:

1. Bevel the sides of each hole and the crack between them.
2. Prepare a soft steel disc for each hole, a little smaller than the hole and bevelled at its edge, with a length of rod melted onto it as a handle.
3. Heat the plate between the first two holes and keep it red hot while tacking and welding the first disc into its hole.
4. Let that heated section cool together with the weld, so the shrinkage is shared.
5. Weld the next disc in the same way, then weld the crack between the holes, and reinforce the finished weld.

For a plate with several cracked rows, it set out an order: preheat, weld one row, reheat to relieve the shrinkage, let the whole job cool, then move to the next row, leaving the longest continuous crack until last.

Figure 1Closing a cracked row of holes, as described in 1919

1. 1Bevel each hole and the crack
2. 2Prepare bevelled discs on rod handles
3. 3Heat the plate between holes
4. 4Weld one disc, then cool together
5. 5Weld the next disc, then the crack
6. 6Reinforce, reheat, cool before the next row

The question the method skips: what were the holes for?

In a firebox side sheet, the holes being closed were staybolt holes. Staybolts tie the inner firebox sheet to the outer wrapper so that the flat plates, which could not resist boiler pressure on their own, are supported at close intervals. Close a staybolt hole with a disc and you remove a support. The plate around it now spans further between the remaining stays.

For a flat plate supported at regular points and loaded by pressure, bending stress rises roughly with the square of the span between supports relative to its thickness. That proportionality is enough to see the consequence. If the stays sit at a 100 mm pitch and one is removed, the local span between remaining supports becomes about 200 mm. Twice the span means roughly four times the bending stress in that region, before any weld effects are counted.

Figure 2Relative plate bending stress as the support span grows

100 mm span, all supports present1

141 mm, the diagonal of the grid2

200 mm, one support removed4

The numbers are proportions, not a design calculation, and real stayed plates are analysed with methods from the applicable boiler standard. The point is qualitative and decisive: closing holes that carried load changes the structure. A modern repair that removes stays, bolts or rivets must be assessed by a competent engineer against the governing standard before it is done, not justified afterwards by how good the weld looks.

Plugging a hole is a design change

Closing a hole by welding is sometimes the right answer, for example when an unused penetration is no longer needed. It should still be treated as a design change rather than a routine repair. Before plugging any hole, ask:

1. What did the hole do? A staybolt, rivet, bolt or tube hole carried load or held something in place. Removing that function changes how the plate is supported.
2. What will carry that load instead? If nothing will, the surrounding supports and plate must be checked for the new span, as the proportional example above showed.
3. How will the plug be joined? A plug or disc welded into a hole is a closed-loop weld under heavy restraint, with all the cracking risks described earlier in this book. Full fusion through the thickness must be achievable and verifiable.
4. Can the plug be inspected? A plug weld that cannot be examined through its thickness leaves uncertainty in exactly the place where the plate was weakest.

The booklet’s soft steel discs, tacked by a rod handle and welded into bevelled holes, answered the third question with ingenuity and the fourth not at all. A modern repair answers all four, documents the answers and, for regulated equipment, obtains approval before the first disc is cut.

Knuckles flex

A knuckle is the curved flange where a flat plate, such as a firebox door sheet or tube sheet, turns to meet another plate. In service, flat plates under changing pressure and temperature flex slightly, and the knuckle is where that movement concentrates. Fatigue cracks along the radius are the result.

The booklet treated knuckle cracks with its usual tools: preheat one end, tack, back-weld towards the start, reheat the far end, and weld from the water side where possible. It asked for knuckle welds to be reinforced about 1/8 inch and widened to about 2 inches at the top, adding stiffness where the plate had been moving.

A modern repair asks why the knuckle cracked. A knuckle radius smaller than the design called for, thinning at the bend, corrosion, scale build-up on the water side that let the plate overheat, or restraint from nearby stays can all contribute. If the cause remains, a weld across the radius may simply move the crack to the edge of the repair. Sometimes the right answer is a new flanged section with the correct radius, welded in away from the bend.

Stop drilling is not a repair

A common workshop response to a crack is to drill a small hole at its tip, removing the sharp crack front so it stops growing. A crack-arrest hole can buy time and may be a sensible temporary measure while a proper repair is planned. It does not remove the cause, the crack may continue from the hole’s edge, and in pressure equipment any such measure falls under the same rules as other repairs. Record it as temporary, inspect it, and plan the permanent repair.

The tank case: cracks along a riveted seam

This is an illustration. The second crack in the tank, introduced in Chapter 1’s exercise, ran from a rivet hole along the rivet row, in a region thinned to 6.8 to 7.4 mm. Penetrant inspection now finds two more short cracks at neighbouring rivet holes in the same row.

The workshop treats this as a failing region, not as three small cracks. The seam is a row of holes in thinned plate carrying the tank’s water load; welding up the cracks and closing holes would change how the seam is held together and would still leave thin plate. The decision is to remove the riveted joint region and fit a new welded insert that extends beyond the thinned and cracked area, reviewed by the owner’s engineer because it changes the joint type of the seam.

Find every crack, not just the one that leaked

A crack that leaks is the one somebody noticed. Cracks between holes and in knuckles rarely appear alone, because the conditions that produced one are present across the whole region. Before deciding on any repair, inspect the full area around the known crack, and the corresponding areas elsewhere on the part, such as the other side of a symmetrical plate or the matching knuckle on the opposite flange.

Practical points:

• Clean thoroughly. Scale, corrosion product and deposits hide tight cracks. Inspection on a dirty surface understates the damage.
• Choose the method for the material and access. Magnetic particle testing suits ferritic steels and finds surface and near-surface cracks well; penetrant testing works on most materials but needs a clean, open surface; ultrasonic methods can find cracks below the surface and measure remaining thickness.
• Inspect from both sides where possible. In stayed plates and flanged knuckles, cracks may start on the side that is harder to reach.
• Map and photograph everything, including indications that turn out to be harmless, so the next inspection can compare like with like.

The count of cracked ligaments from this chapter’s exercise is only meaningful if every ligament was inspected.

Why stayed plates crack where they do

The firebox shows clearly how service conditions focus strain at holes and knuckles. The inner sheets faced the fire; the outer wrapper sat in cooler surroundings. As the boiler was fired, the inner sheets grew more than the outer ones, and the staybolts tying them together had to bend to accommodate the difference. Each firing and cooling flexed the plate around every stay. Plates near the fire and near the corners, where movement was greatest, took the largest share.

Scale and deposits on the water side made matters worse. A layer of scale insulates the plate from the water that should be cooling it, so the metal runs hotter, expands more and loses strength. Cracks between staybolt holes and along knuckles were, in large part, the visible record of those thermal cycles.

The same pattern appears in modern equipment: heat exchanger tube sheets, flanged heads on vessels that cycle in temperature, perforated plates in furnaces and dryers. Wherever a perforated or flanged plate is heated and cooled while restrained, ligaments and radii are where the strain concentrates.

Figure 3How thermal cycling cracks a stayed plate

1. 1Plate is heated and expands
2. 2Stays and edges restrain it
3. 3Ligaments and radii take the strain
4. 4A crack starts at a hole edge or radius
5. 5Neighbouring ligaments take more load
6. 6Plate cools and the cycle repeats

A repair weld resets the clock, but not the cause

Fatigue cracks grow in stages: a crack starts at a stress concentration, grows a little with each load cycle, and eventually the remaining section fails. A good repair removes the existing crack, but it also introduces new features that can start the next one: weld toes, residual tensile stress, possible small defects and changes in stiffness. If the cycling that produced the first crack continues unchanged, the repaired area is often where the next crack starts.

That is why the response to cracks between holes and in knuckles should look beyond the weld:

• Reduce the cycling. In fired equipment, slower warm-up and cool-down, and avoiding rapid temperature changes, reduce the strain per cycle.
• Remove the aggravating condition. Water treatment that prevents scale, cleaning that removes deposits, or support that removes vibration can matter more than the repair itself.
• Change the geometry. A larger knuckle radius, a different hole pattern or added support may be possible when a section is replaced, though such changes to regulated equipment need engineering review and approval.
• Monitor the repair. Inspect repaired areas at shorter intervals at first, so that any new crack is found while it is small.

Choosing between repair and replacement for a failing region

When a region of ligaments or a knuckle has several cracks, a structured comparison helps. Consider the number and length of cracks, the remaining thickness, the cause and whether it can be controlled, the consequence of failure, and the evidence each option would leave behind. A replacement section removes all of the damaged, cycled material at once and may allow the geometry to be improved; several local repairs leave fatigued plate between them and add residual stress at each. The booklet reached the same conclusion with less analysis when it advised that a badly cracked sheet would be better replaced.

The tank case: checking the other seams

This is an illustration. Having found three cracked holes in one riveted seam, the workshop inspects every other riveted seam on the tank by penetrant testing and thickness measurement. One more seam shows thinning to about 8.6 mm but no cracks. The workshop records it, schedules a re-inspection at the next shutdown, and notes that both affected seams sit on the side of the tank exposed to standing water, which links back to the drainage change chosen in Chapter 1.

Exercise: count the cracked ligaments

A row of 12 holes has 11 ligaments between them. Inspection finds cracks in ligaments 3, 4, 5 and 8. What proportion of the ligaments are cracked, and what does the pattern suggest?

Four of eleven ligaments are cracked, about 36%. Three adjacent cracked ligaments, 3 to 5, suggest a region where strain is concentrated, possibly by a local cause such as a nearby restraint, a heat concentration or thinning. The isolated crack at 8 may be an early instance of the same mechanism. The pattern points towards investigating the region and considering an insert or replacement, rather than welding four separate cracks.