Read the damage before choosing a repair

Most weld repairs that fail were decided too quickly. A crack is visible, a welder is available and the job looks small. The repair is completed, the part returns to service and the crack reappears beside the new weld, because the conditions that made the first crack were never changed.

The 1919 Airco booklet on locomotive fireboxes is a useful reminder that even enthusiastic advocates of welding knew this. Discussing cracks between staybolt holes, it advised that a side sheet carrying many such cracks would be better replaced with a new sheet than welded. A process vendor, writing to sell welding, still drew a line where repair stopped making sense. This chapter sets out how to draw that line deliberately.

Separate the symptom from the damage mechanism

A leak, a crack or a thin patch is a symptom. Before any repair is designed, name the mechanism that produced it. Common mechanisms in steel plate include:

• Thermal fatigue: repeated heating and cooling that strains the plate where it is restrained, typical of furnace and firebox plates and of parts that cycle between hot and cold.
• Mechanical fatigue: repeated loading that grows cracks from holes, weld toes, sharp corners and other stress raisers.
• Corrosion and erosion: general thinning, pitting and grooving from water, chemicals or escaping fluid.
• Overload: a single event that exceeded the part’s strength, usually leaving bending, tearing or necking.
• Manufacturing or earlier repair defects: lack of fusion, porosity or poor fit-up that gave a crack a place to start.

The firebox shows several at once. Its inner plates sat against fire on one side and water on the other, were held by hundreds of staybolts and were heated and cooled every time the locomotive was fired or allowed to cool. The booklet’s foreword attributes the frequent leaks at riveted joints to the abnormal heat strains those plates endured. Caulking leaking seams over and over then wore the plate edges, which the booklet treats as a separate repair.

Map the extent, not just the visible part

Cracks are usually longer than they look. Their tips are tight and may be hidden by scale, paint or corrosion product. Before a repair is planned, clean the area and establish the full extent of the damage with a suitable method, such as dye penetrant or magnetic particle inspection for surface cracks, and ultrasonic thickness measurement for thinning. Record the findings on a sketch or photograph with dimensions, so the repair can be planned against the real damage rather than the visible part of it.

A mapped crack also tells a story. A crack that runs between holes in a row suggests the row is a line of weakness under cyclic strain. A crack at a weld toe suggests a stress concentration or a weld defect. Several parallel cracks across a region suggest the region as a whole is being strained beyond what it can tolerate, which is exactly the situation in which the 1919 booklet preferred replacement.

Figure 1Repair, insert or replace?

?What does the mapped damage show?

One short crack, cause understoodLocal weld repairafter full excavation

A damaged region, sound plate around itFlush insert or patchedges in sound metal

Many cracks or widespread thinningReplace the part or sheetrepair would chase damage

Cause unknown or still activeStop and investigatebefore any welding

Ask what the part must still do

The repair’s acceptance depends on the part’s job. Write down the loads, temperatures, pressures, environment and service life the repaired part must survive, and the consequence if it fails. A crack in a bracket on a garden trailer and a crack in a pressure boundary are not the same decision, even if they look alike.

This is where regulated equipment separates from everything else. If the part contains pressure, lifts people or loads, carries structural load in a building, or forms part of a rail vehicle, the repair is not purely the workshop’s decision. The owner’s obligations, the applicable standards and the relevant regulator or inspection body determine what must be done and who must approve it. Identify that before any planning, because it changes everything that follows.

Compare repair, insert and replacement honestly

Three options usually exist:

1. A local weld repair removes the crack and replaces the excavated metal with weld. It suits short, isolated cracks in material that is otherwise sound.
2. An insert or patch cuts out a damaged region and welds in new plate. It suits a defined area of damage surrounded by sound material.
3. Replacement of the part or a complete sheet suits widespread damage, uncertain material condition or a part whose design invites the same failure again.

Compare them on more than cost. A repair adds heat, residual stress and new inspection obligations. It may also remove evidence of the original failure, so record the damage first. Replacement may allow a design change that removes the cause, such as a better corner radius or a different staying arrangement. Sometimes the repair that looks cheapest is the one that returns soonest.

The tank case: the first assessment

This is an illustration. A restoration workshop is repairing a riveted and welded atmospheric water tank. One side plate is nominally 10 mm thick carbon steel. Inspection after cleaning finds a 140 mm crack running from the toe of an old fillet weld that attached an outlet nozzle. Dye penetrant shows the crack is 30 mm longer than it first appeared, at 170 mm in total. Ultrasonic thickness readings around it range from 9.4 to 9.8 mm, so the surrounding plate is only lightly thinned.

The workshop records the crack’s position and length, photographs the old weld and notes that the outlet pipe has no support for a long run of attached pipework. Vibration and the weight of that pipework are a plausible cause of fatigue at the weld toe. The repair decision is therefore in two parts: remove and repair the crack, and support the pipework so the cause is removed. Because the tank is atmospheric and not pressure equipment, the repair remains the owner’s engineering decision, though the workshop still documents it.

Figure 2Before planning any weld repair

• Avoid: Clean the area and map the full extent of the damage
• Avoid: Name the likely damage mechanism
• Avoid: Record the loads, temperatures and environment in service
• Avoid: Identify whether the part is regulated equipment
• Avoid: Compare repair, insert and replacement
• Do: x Decide how the cause will be removed or controlled

When the history helps and when it misleads

The 1919 booklet makes strong claims for welding: stronger joints than riveting, no leakage and lower cost, depending on the welder’s skill. Some of those claims reflected real advantages of a well-made weld over a leaking riveted seam. Others reflected the enthusiasm of a company selling a new process before modern inspection, metallurgy and pressure-equipment rules had developed. Fusion welding of boilers went on to need decades of codes, qualification and testing before it was trusted for pressure parts.

Read the booklet as evidence of how practitioners reasoned, not as a source of approved methods. Its most durable lesson is the one in this chapter: understand the damage and the plate before welding anything.

Collect the evidence before anything is cut

The first cut destroys evidence. Grinding out a crack removes its surface, its tip and the corrosion product inside it, all of which can tell you how it grew. Before excavation, record what you can:

• Photographs of the damage, the surrounding area and the whole part, with a scale and a reference to its position on the part.
• Dimensions: crack length and orientation, the extent of thinning, and the distance to holes, welds and edges.
• Inspection results: which method was used, by whom, when and with what result, including any indications that later prove false.
• Service history: operating hours or cycles, loads, temperatures, fluids, previous repairs and any events such as an impact, an overload or a period out of service.
• Material information: drawings, specifications, mill certificates and, if those are missing, a decision about whether to test a sample.

For a significant failure, consider keeping a section of the damaged material. A short length of cracked plate removed during an insert repair can be examined later if the repair does not behave as expected. Label it, record where it came from and store it where it will not be thrown out in the next clean-up.

Read what the fracture surface is saying

Where a part has cracked through, or where a section is removed during the repair, the fracture surface is one of the best clues to its cause. A few broad patterns are worth recognising, while remembering that confirming a mechanism may need laboratory examination:

• Fatigue surfaces are usually flat and relatively smooth, often with curved bands, sometimes called beach marks, spreading out from the origin. The origin is commonly at a stress raiser: a hole edge, a weld toe, a corrosion pit or a sharp corner. There is often little overall deformation.
• Overload in a ductile steel tends to show distortion, necking or a rough, fibrous surface, sometimes with sloping shear lips at the edges.
• Brittle fracture surfaces are often bright and granular with little deformation, and may show chevron-like markings pointing back towards the origin.
• Corrosion-assisted cracking may leave crack faces covered in corrosion product, branching cracks or pits along the crack path.

Even a quick look changes the repair decision. A fatigue crack means the part has been cycling at a stress it cannot sustain, so something about the loading, the geometry or the support must change. An overload means a single event exceeded the design, and the question becomes whether it can happen again. A brittle fracture in old plate raises questions about the material itself that welding cannot answer.

Weigh likelihood and consequence

Two parts with the same crack can deserve very different responses. A simple way to keep the decision honest is to consider two questions separately: how likely is the repaired part to fail again, and what happens if it does?

Figure 3Triage by likelihood and consequence

Consequence of failure: low → high

Severe but unlikelyengineered repair with inspection

Severe and likelyreplace or redesign; seek approval

Minor and unlikelysimple repair, recorded

Minor but likelyfix the cause before repairing again

Likelihood of repeat failure: low → high

A crack in a non-structural cover that cracked once in many years of service sits in the low corner: repair it, record it, move on. A crack in a pressure boundary, a lifting point or a structural member sits in the high-consequence half regardless of how likely a repeat seems, and the repair must be engineered, inspected and approved accordingly. A crack that keeps returning in a low-consequence part is a signal that the cause has not been found, even if each failure is cheap.

Write the repair decision down

A short decision record, written before any welding, keeps the reasoning visible to the welder, the inspector, the owner and whoever inspects the part next. It should state:

1. what was found, with references to the photographs and inspection results;
2. the damage mechanism, and how confident the assessment is;
3. the options considered: repair, insert, replacement or redesign;
4. the option chosen and why;
5. how the cause will be removed or controlled;
6. whether the part is regulated equipment, and what approvals are required;
7. who made the decision, and who must review it.

The record does not need to be long. A page with clear headings and references to photographs is often enough for workshop equipment. For pressure equipment and structural repairs, the format may be set by the owner’s quality system or the inspection body. Chapter 10 builds this record into a full repair dossier.

Common traps in damage assessment

A few mistakes recur:

• Measuring the crack you can see. Visible length understates the real crack, often by a meaningful margin. Map it with an appropriate method first.
• Assuming the cause. “It cracked because it is old” is not a mechanism. Age explains nothing until it is linked to fatigue, corrosion or deterioration of a specific kind.
• Repairing over corrosion. Welding onto thinned or pitted plate without measuring the remaining thickness can leave a sound weld surrounded by plate too thin to carry the load.
• Removing evidence too early. Grinding, painting or blasting before recording the damage can make the cause impossible to establish later.
• Treating repeat failures as bad luck. A second crack in the same place is evidence that the first repair did not address the cause.

The booklet’s preference for a new sheet over a sheet covered in cracks reflects the same discipline: look at the whole region, decide what it is telling you, and choose the repair that matches the damage rather than the one that is quickest to start.

Exercise: assess a second crack

A second crack in the same tank runs 60 mm from a rivet hole in a riveted seam, roughly along the line of the rivet row. Thickness readings around it are 6.8 to 7.4 mm on a nominal 10 mm plate, and two neighbouring rivets show weeping. List the questions you would answer before choosing a repair.

A sound answer asks why the plate has lost up to 3.2 mm, roughly a third of its nominal thickness, near the seam, whether corrosion is concentrated at the seam, how far the thinning extends, whether the rivet row is a line of weakness under load, and whether other holes in the row are cracking too. It also asks whether a local weld is wise in a riveted seam at all. With widespread thinning and a crack along a row of holes, an insert extending well beyond the seam, or replacement of the plate, deserves serious consideration before a local weld.