Steel Silo Fillet Weld Quality: What Holds the Whole Structure Together

A fillet weld on a steel silo carries the same load whether the silo holds 500 tonnes or 15,000 tonnes — but the consequences of failure scale with every tonne. The answer? Design fillet welds to the actual load path, not the minimum code requirement. And inspect them with UT and MT, not just visual. That's the difference between a silo that lasts 25 years and one that cracks at the hopper seam in year three.

Why Fillet Welds Are the Hidden Backbone of Silo Integrity

The Load Path Nobody Draws

When you look at a structural drawing for a 6,000-tonne cement silo, the big numbers go to the shell plate thickness and the anchor bolt pattern. The fillet welds — hopper-to-shell, stiffener-to-shell, roof-to-shell — get a generic callout: "6 mm fillet weld, continuous." That's a problem.

Here's what actually happens. A full silo of clinker exerts a horizontal pressure of roughly 35–55 kPa at the shell wall, depending on the diameter and the material's bulk density (typically 1.2–1.5 t/m³ for clinker). That pressure transfers through the hopper ring, which is almost always connected to the bottom shell course with a fillet weld. If that weld is undersized by just 2 mm on the leg length, the effective throat drops by about 30%. At 40°C ambient in a plant outside Karachi, I watched a hopper ring weld develop a 140 mm crack after two filling cycles. The original weld was specified at 8 mm leg length. What we measured in the field: 5.2 mm average.

Where the Real Stress Lives

Fillet welds on silos don't just carry static weight. They're subjected to:

• Cyclic bending from filling and discharge — a silo cycling twice a day racks those welds 730 times per year.
• Thermal differential — a silo in Mongolia sees shell temperature swings from -30°C to +45°C. That's 75°C of differential expansion between the shell and the concrete foundation ring.
• Discharge impact — when 800 tonnes of coal drops onto a vibrating feeder, the transient load on the hopper welds can spike to 1.8× the static design load.

I remember a project in Vietnam where the client wanted to save money by reducing all fillet welds to the Eurocode minimum. We ran the numbers: at the hopper-to-shell junction on a 12 m diameter silo, the minimum 5 mm fillet would have a fatigue life of under 4 years at the plant's cycling rate. We specified 8 mm instead. That was 2019. The silo's still running with zero weld defects reported.

Design Rules That Actually Work in the Field

Throat Thickness vs. Leg Length — Stop Confusing the Two

This is the single most common mistake I see on drawings. A "6 mm fillet weld" means a 6 mm leg length. The effective throat — the part that actually resists shear — is 0.7 × leg length, so 4.2 mm. But here's the catch: most field welds don't achieve the theoretical throat. Undercut, uneven leg lengths, and convex profiles can reduce the effective throat by 10–25%.

Eurocode 3 (EN 1993-1-8) and AWS D1.1 both recognize this. The practical fix? Specify leg length with a minimum that accounts for field variability. For primary structural connections on silos carrying more than 1,000 tonnes, I don't go below 7 mm leg length. For hoppers handling abrasive materials like slag or alumina, I push to 8–10 mm. The cost difference in filler metal is maybe $0.30 per linear meter. The cost of a weld repair at height, with scaffolding and a crew of four, starts at $2,800 per day.

Welding Process Selection for Silo Construction

Three processes dominate silo field welding, and each has a fillet weld quality profile you need to understand:

FCAW (Flux-Cored Arc Welding): Most common on silo projects. Deposition rates of 3–6 kg/hr. Good penetration on dirty or slightly rusted plate — which is what you get on a construction site where shell plates have sat outside for three weeks. The downside: slag inclusions if the interpass temperature isn't controlled. I've seen rejection rates hit 12% on windy sites when FCAW gas shielding got disrupted.

SMAW (Stick Welding): Slower — 1.5–3 kg/hr — but more forgiving outdoors. Wind speeds up to 8 m/s don't kill the weld pool the way they do with gas-shielded processes. For vertical-up welds on shell stiffeners, SMAW with 7018 rods is still my go-to. The weld quality is more consistent because the welder has better control over the pool.

GMAW (MIG): Rarely used for structural fillet welds on silos in the field. Too sensitive to wind. I've seen it used successfully inside enclosed silos for internal stiffener attachment, where the environment is controlled. Deposition rate is high — 4–7 kg/hr — but porosity rates outdoors can exceed 8%.

Inspection: What Catches Problems Before They Become Failures

Visual Inspection Is the Floor, Not the Ceiling

Every fillet weld on a silo should get a visual inspection per AWS D1.1 or EN ISO 5817. That catches the obvious stuff: undercut deeper than 0.5 mm, visible porosity, cracks, insufficient leg length. But visual inspection misses 60–70% of the defects that actually cause failures. Lack of fusion at the root, internal slag inclusions, micro-cracking in the heat-affected zone — none of that shows on the surface.

On a 4,500-tonne fly ash silo we commissioned in Shandong, visual inspection passed 100% of the hopper fillet welds. Then we ran ultrasonic testing (UT) on a 20% sample per EN ISO 17640. Three welds had root lack-of-fusion defects spanning 40–80 mm. All three were at the same stiffener location — a fit-up issue where the stiffener was clamped too tight against the shell, preventing root access during welding. We re-welded all 12 stiffener connections at that elevation. Total cost: $14,000. The alternative — discovering it after the silo was filled — would have been a structural repair at height with a full crew, probably $80,000 minimum.

Magnetic Particle Testing for Surface-Breaking Defects

MT (per EN ISO 17638 or ASTM E1444) is the right tool for fillet welds on carbon steel silos. It finds surface and near-surface cracks that visual inspection misses — especially toe cracks, which are the most common fatigue-initiation point on silo welds.

The procedure matters. You need a minimum current of 3 A/mm of prod spacing for AC yoke method. On a cold morning in Inner Mongolia (ambient -12°C), we had to pre-heat the weld zone to 15°C before MT because the magnetic particle suspension wouldn't flow properly below 10°C. The inspection took three hours longer than planned. Found two toe cracks, each about 25 mm long, on a roof-to-shell fillet weld. Both were within 50 mm of a roof vent penetration — a stress concentration point the original designer hadn't accounted for.

Common Defects and What They Tell You About Your Process

Undercut: The Silent Fatigue Killer

Undercut is a groove melted into the base metal adjacent to the weld toe, left unfilled. On a silo shell plate that's 8–12 mm thick, even 0.5 mm of undercut creates a stress concentration factor of 2.5–3.0. In a cyclic loading environment, that's where fatigue cracks start.

Undercut is almost always a technique problem, not a procedure problem. Excessive travel speed, wrong electrode angle (holding the rod more than 45° from vertical on a horizontal fillet), or too high amperage. On one project in Algeria, we had a crew producing undercut on 30% of their fillet welds. Root cause? They were running 10% above the WPS amperage to go faster. We pulled the WPS cards, re-qualified the welders, and the rejection rate dropped to under 4% in one week.

Slag Inclusions in FCAW Welds

Slag inclusions in flux-cored welds are a fit-up and technique issue. The most common cause on silos: welding over a gap wider than 2 mm between the stiffener and the shell plate. The slag flows into the gap and gets trapped at the root.

EN ISO 5817 Quality Level B (the standard for structural silos) permits slag inclusions up to 2 mm in length for fillet welds. But here's the thing — a 2 mm slag inclusion at the root of a hopper fillet weld on a 5,000-tonne silo is a crack initiation point. I've seen plants reject welds that technically met the standard because the location was wrong. That's not being overly conservative. That's understanding that a silo isn't a building frame — you can't easily access the hopper weld from inside once the silo is operational.

Lack of Fusion at the Root

This is the defect that scares me most. Root lack of fusion on a fillet weld means the weld never bonded to the base material at the joint root. The weld looks fine on the surface. It passes visual inspection. It might even pass UT if the probe angle isn't right. But under cyclic loading, that unfused root acts like a pre-existing crack.

The fix is procedural: back-gouging and re-welding from the root side, or using a backing bar. For critical hopper connections, we specify back-gouge and re-weld on the root pass. It adds about 15% to the welding time for that joint. On a silo with 24 hopper support connections, that's roughly 12 extra labor hours. Against the cost of a structural failure, it's nothing.

Case Study: 8,000-Tonne Clinker Silo — Weld Quality Recovery

Project: 8,000-tonne clinker silo, 22 m diameter, 38 m height to roof. Cement plant, Southeast Asia.

Problem: During construction, the EPC contractor's QC team found that 18% of fillet welds on the hopper-to-support-ring connection failed UT inspection. The original WPS specified FCAW with E71T-1C wire, 1.2 mm diameter, 220–260 amps, 24–28 volts. Field conditions: 34°C average, 70% humidity, wind gusts to 6 m/s.

Root Cause Analysis: The welders were running at the upper end of the amperage range (255–260A) to compensate for the wind disruption of gas shielding. Higher amperage + gas disruption = porosity and lack of sidewall fusion. Additionally, fit-up gaps between the hopper flange and support ring averaged 3.5 mm — well above the 2 mm maximum in the WPS.

Corrective Actions:

• Installed windbreaks (tarps on frames) around all hopper weld stations, reducing local wind speed to under 2 m/s.
• Re-qualified WPS with reduced amperage range: 210–230A, increased voltage to 26–28V, and reduced travel speed by 15%.
• Implemented a fit-up check before welding — any gap over 2 mm required a buildup pass on the root side before the main fillet weld.
• Increased UT sampling from 10% to 100% on hopper connections, 20% on shell stiffeners.

Result: Rejection rate dropped from 18% to 3.2% over the next 400 linear meters of welding. The silo was commissioned in November 2021. At the 24-month inspection in late 2023, zero weld defects were found on any fillet weld connection. Total cost of the corrective actions during construction: approximately $22,000. Estimated cost of post-commissioning repair: $150,000–$200,000 (scaffolding, material removal, re-welding, re-inspection, production downtime).

FAQ

Q: What's the minimum fillet weld size for a steel silo hopper connection?

A: It depends on the load, but as a practical minimum for industrial silos over 1,000 tonnes capacity, I specify 7 mm leg length. Eurocode 3 permits smaller welds based on calculation, but field variability means a 5 mm specified weld often measures 3.5–4 mm in practice. For abrasive materials (slag, coal, alumina), go to 8 mm minimum. The filler metal cost is negligible compared to the risk.

Q: Should fillet welds on silos be 100% UT inspected?

A: Not all of them. 100% UT on hopper-to-shell and hopper-to-support-ring connections — yes, those are non-redundant structural joints where a failure is catastrophic. For shell stiffener fillet welds, 20–30% random sampling per EN ISO 17640 is sufficient, provided the WPS is qualified and the welders are certified. Roof structure fillet welds: visual + 10% MT. The key is matching inspection intensity to consequence of failure.

Q: Can you repair a cracked fillet weld on an operating silo?

A: Yes, but it's expensive and risky. You need to empty the silo to at least 30% below the crack location to reduce stress on the repair zone. Grind out the defect completely — don't just weld over it. Verify removal with MT. Then re-weld with a qualified WPS, typically using SMAW for better control in repair conditions. Post-repair, UT the full repair length plus 150 mm on each side. I've done this twice. Both times, the repair cost 5–8× what it would have cost to get it right during construction. Prevention isn't just better — it's cheaper by an order of magnitude.