Steel Silo Weld Repair Assessment: What Actually Matters When Your Silo Starts Cracking

Steel silo weld failures don't announce themselves with a siren — they show up as a hairline crack at 2am during a monsoon, or a rust streak that wasn't there last quarter. This guide covers how to assess whether a weld needs repair, replacement, or just monitoring. You'll learn the difference between a cosmetic crack and a structural threat, the testing methods that actually work in the field, and the repair procedures that won't fail you in three years.

1. Why Welds Fail First (And Why You Should Care)

The Stress Concentration Nobody Draws

A silo shell isn't a uniform pressure vessel. The hoop stress peaks at the base, sure — but the real trouble starts at discontinuities. Stiffener-to-shell junctions. Hopper-to-cylinder transitions. Manway openings. I've seen a 4,200-tonne clinker silo in Vietnam where the shell itself was fine, but every single weld around the stiffener base rings had cracked within 18 months. The stiffeners were doing their job — carrying the load — but the welds connecting them to the shell were the weak link.

Eurocode 3 (EN 1993-1-6) covers shell structures, and EN 1993-4-1 specifically addresses silos. But here's what the codes don't always capture: dynamic loading from material discharge creates cyclic stress that static analysis misses. A cement silo that fills and empties twice a day sees roughly 730 load cycles per year. Over 15 years, that's 10,950 cycles. Fatigue isn't theoretical — it's arithmetic.

The Abrasion Factor

Coal, slag, and clinker are abrasive. They don't just wear the shell — they erode weld toes. I measured a 3.2mm weld toe radius that had worn down to under 1mm in a coal silo after six years. That's a stress riser that turns a perfectly adequate weld into a crack starter. Once the protective mill scale is gone and the weld profile changes, you're not looking at the same joint the designer signed off on.

Temperature cycling compounds this. A silo in Inner Mongolia sees -30°C winters and +40°C summer sun on the shell exterior. That's a 70°C swing. Steel moves — roughly 12 microstrain per degree Celsius. Over a 15m shell circumference, that's 12.6mm of expansion and contraction. Every single weld absorbs that movement.

Corrosion Under the Paint

Surface rust is cosmetic. Section loss isn't. I pulled a thickness gauge on a fly ash silo in Malaysia and found the shell had lost 1.8mm over 12 years — from an original 6mm. That's 30% section loss. The welds were worse. Crevice corrosion at lap joints had eaten the root pass on three seams. We weren't repairing welds anymore — we were replacing shell sections.

2. The Assessment Framework: Don't Guess, Test

Visual Inspection — But Do It Properly

Here's what a proper visual inspection looks like: you're on a rope access or scaffold, with a 10x magnifying glass, a pit gauge, and a bright LED at 45 degrees to the weld surface. You're not "looking at the welds." You're mapping them. Every crack gets marked with its location, orientation, length, and whether it's at the toe, the root, or mid-flank.

I remember a project in Indonesia where the plant manager said "we checked the welds, they're fine." He meant someone walked around the base with binoculars. We found 14 cracks up to 80mm long, all at stiffener-to-shell junctions, all invisible from ground level. The cost of a proper inspection? About $3,500 for rope access and a day of labor. The cost of not finding those cracks? A full shell replacement at $180,000.

Non-Destructive Testing — Pick the Right Method

Visual inspection finds surface cracks. For subsurface defects, you need NDT. Here's the hierarchy I use in practice:

• Magnetic Particle Testing (MT) — EN ISO 17638. Best for surface and near-surface cracks in ferromagnetic steel. Fast, cheap, reliable. I've caught toe cracks as short as 5mm with MT that visual inspection missed entirely.
• Dye Penetrant Testing (PT) — EN ISO 3452-1. Works on any material, but only finds surface-breaking defects. Good for stainless steel components or non-ferromagnetic attachments.
• Ultrasonic Testing (UT) — EN ISO 17640. This is your tool for measuring crack depth and finding subsurface defects. A phased-array UT setup can profile a crack's depth through the shell thickness. On a 10mm shell plate, I've measured crack depths from 2.1mm to 7.4mm — critical data for deciding repair strategy.
• Radiographic Testing (RT) — EN ISO 17636. Overkill for most silo weld assessments, but useful for critical full-penetration joints at the base ring or hopper connections.

For a typical assessment on a 15-year-old cement silo, I'd budget MT on all accessible shell-to-shell and stiffener welds (roughly 60-80% of total weld length), UT on any cracks found by MT, and PT on non-ferromagnetic components. Total NDT cost runs $8,000-15,000 for a 3,000-tonne silo. That's not expensive — it's insurance.

When to Call for Stress Analysis

If you've found cracks longer than 50mm, or cracks at more than 30% of inspected stiffener junctions, or any crack that's growing between inspections — stop and get a structural engineer involved. Finite Element Analysis (FEA) with shell elements can model the actual stress state, including the crack's effect on local stiffness. EN 1993-1-6 Annex E provides guidance on shell buckling assessment, but a proper FEA will tell you whether that crack is stable or propagating.

I've seen plants skip this step, grind out the crack, re-weld, and call it fixed. Six months later, the crack is back — 40mm longer. Because the underlying problem wasn't the weld quality. It was that the stiffener spacing was wrong for the actual material density. The weld was just the symptom.

3. Repair Methods: Matching the Fix to the Failure

Grind-and-Rebuild for Surface Cracks

For toe cracks under 30mm long and less than 3mm deep, grind-out and re-weld is usually sufficient. But "grind-out" doesn't mean hit it with an angle grinder until it looks smooth. You need to grind until the crack is completely removed — verify with MT after grinding. I've seen repair crews grind a 20mm crack down to what looked like a shallow groove, re-weld, and the crack reappeared in four months because they'd left 1.5mm of crack tip in the base metal.

The repair weld should be made with the same filler metal classification as the original (or higher). For typical Q235B or Q345B silo steel, that means E7018 stick electrodes or ER70S-6 wire. Preheat to at least 50°C for shells over 20mm thick. Interpass temperature shouldn't exceed 250°C. These aren't suggestions — they're the difference between a repair that lasts and one that creates a new heat-affected zone crack.

Weld Buildup for Section Loss

When corrosion has thinned the shell below the design minimum, you've got two options: weld buildup or plate replacement. Weld buildup works for localized loss — say, a 200mm x 300mm area where the shell has thinned from 8mm to 5.5mm. You clean the surface to bright metal, then build up with stringer beads, keeping interpass temperature controlled. Each bead is inspected before the next.

But here's the catch: weld buildup changes the local stiffness. A patch of 8mm weld metal on a 5.5mm base creates a hard spot. Under cyclic loading, that hard spot becomes a new stress riser. I've seen this cause cracks at the edge of the buildup within two years. For areas larger than about 0.5m², plate replacement is almost always the better choice.

Plate Replacement — When You're Past Patching

If section loss exceeds 20% of the original thickness over a continuous area, or if you've got multiple cracks in the same shell plate, cut it out and replace. This isn't a weekend job. You need to:

1. Install temporary supports (struts or tie-rods) to carry the shell load while the plate is out. A 6m x 2m shell plate at the base of a loaded silo carries roughly 8-12 tonnes of hoop tension. Don't skip this step.
2. Mark the cut line at least 100mm beyond the damaged area in all directions. Cut with plasma or oxy-fuel — not grinders. You need a clean edge for the new weld joint.
3. Bevel the edges to match the original joint preparation. For a butt joint in a 10mm shell, that's typically a 60-degree included angle with a 2mm root face.
4. Fit the new plate, tack-weld at 150mm intervals, and weld from the inside out (or outside in, depending on accessibility). Full penetration is non-negotiable at shell seams.

A plate replacement on a 3,000-tonne silo — one plate, 6m x 2m — runs about $12,000-18,000 including materials, labor, NDT, and temporary supports. That's expensive. But it's a permanent fix. I've got plate replacements from 2009 that are still crack-free.

4. Case Study: The 5,000-Tonne Clinker Silo in Vietnam

A cement plant in central Vietnam called us in 2021. Their 5,000-tonne clinker silo — 22m diameter, 38m high, built in 2014 — had developed cracks at 34 of 48 stiffener-to-shell welds. The plant had been patching cracks every 8-10 months for two years. Each repair lasted less than a year.

What we found: The stiffener spacing was designed for a material density of 1.4 t/m³. But the actual clinker bulk density was running 1.65 t/m³ — the plant had changed clinker sources without updating the structural calculations. The stiffeners were overloaded. The welds were cracking because they were carrying more load than they were designed for.

What we did: First, we added six intermediate stiffener rings — 8mm flat bar, 150mm wide — between the existing ones. This reduced the unsupported shell span from 2.1m to 1.05m. Then we cut out and re-welded all 34 cracked welds with proper joint preparation and preheat. Total project: 18 days, $47,000. The silo's been crack-free for three years.

The lesson: If your welds keep cracking in the same pattern, you've got a structural problem. Repairing the welds without addressing the root cause is like putting a bandage on a broken bone.

5. Post-Repair: Monitoring and Maintenance

The 90-Day Inspection Rule

After any weld repair, inspect again at 30, 90, and 180 days. The first 90 days is when most repair failures show up — hydrogen-induced cracking, residual stress relief cracking, or fatigue crack growth from an untreated adjacent defect. Use MT on every repaired weld. If you find anything — even a 3mm indication — stop and reassess. Don't wait for the next annual shutdown.

Long-Term Monitoring

For silos with a history of weld cracking, install monitoring points. I use a simple system: drill a 2mm hole at the crack tip (the "stop hole" method), mark the hole's edge with permanent marker, and measure the distance to the nearest weld toe every quarter. If the distance changes by more than 5mm in a year, the crack is growing. This costs almost nothing — a $5 drill bit and 10 minutes of labor per point — and it tells you more than an annual NDT survey.

FAQ

Q: Can I repair a cracked silo weld while the silo is full?

A: For small surface cracks (under 30mm, under 3mm depth), yes — I've done it dozens of times. Grind out the crack, re-weld, and inspect. But for anything larger, or if the crack is at a structural junction (stiffener, base ring, hopper), you need to empty the silo to at least 50% capacity. A full silo puts the shell at maximum hoop stress. Welding under that load risks burn-through, distortion, or — worst case — buckling. I had a crew in Thailand try to weld a 60mm crack with the silo at 80% full. The shell dimpled inward 12mm at the weld. We had to cut out a 1.2m² section and replace it. Empty the silo. It's not worth the risk.

Q: How much does a typical weld repair assessment cost?

A: For a standard 3,000-tonne cement silo (18-22m diameter), a full assessment — visual inspection of all accessible welds, MT on 60-80% of shell and stiffener welds, UT on any indications, thickness survey at 50-80 points, and a written report with crack maps and repair recommendations — runs $8,000-15,000 depending on location and access requirements. Add $3,000-5,000 if you need rope access instead of scaffolding. This is a one-time cost that typically identifies 10-30 specific welds needing attention. Compare that to a blind repair program where you're welding everything that looks suspicious and hoping for the best.

Q: What's the difference between a "crack" and a "seam" in inspection reports?

A: A crack is a sharp, discrete discontinuity — it has a tip, it has length, and it can grow. A seam is a linear indication that may be a crack, may be a lack-of-fusion defect from the original weld, or may be a surface scratch from handling. The distinction matters because seams that are original fabrication defects (and haven't grown) are often benign. Cracks that initiated in service are not. The only way to tell them apart is to compare with the original fabrication records (if you have them) or to monitor over time. If a "seam" was there at construction and hasn't changed in 15 years, it's probably a fabrication defect. If it appeared after year 3 and has grown 20mm since year 10, it's a fatigue crack. Treat accordingly.