Material Insights · Understanding the Material World

Structures and materials: stress, stiffness, strength and why things break: Why things break: cracks, fatigue, corrosion, creep and the margin that keeps them standing

All courses / Understanding Engineering / Module 4: Structures and materials: stress, stiffness, strength and why things break Part 5 of 5

Why things break: cracks, fatigue, corrosion, creep and the margin that keeps them standing

Structures rarely fail because the load was too big for the nominal strength. They fail because of a flaw, a repeated load, rust, or a margin that was too thin. This part sets out the mechanics of those breakages with real cases and numbers, and ends with the factor of safety, the margin that covers what the calculation did not know.

5.1 A crack concentrates stress at its tip, so brittle fracture can run at loads far below yield

Cut a small nick in the edge of a plastic carrier bag and the bag tears easily, even though the unnicked film would take a much bigger pull. The nick acts as a stress concentrator. Force flowing through a plate must squeeze past a hole or a notch, crowding in at the edge. A round hole in a wide plate raises the local stress to three times the average. A sharp crack raises it by far more, because the tighter the tip, the higher the crowding.

If the material is ductile, the metal at the tip yields, blunts the crack and spreads the load. If it is brittle, or cold, the tip cannot yield and the crack advances through the material at hundreds of metres per second. The criterion is a property called fracture toughness, written K, in MPa times the square root of metres. A crack of size a becomes unstable when stress x square root of (pi x a) reaches K. Rearranged, the critical crack size is a = (K / stress) squared / pi. The worked estimate shows how small this is in glass and how large in tough steel.

Many ships of the 1940s, built by welding in large numbers during a war, cracked across their decks. The steel was tolerable at room temperature but turned brittle in cold seas, and the cracks started at defective welds in sharp corners of the structure, where stress crowds. A US board of investigation found fractures in 970 of the 4,694 welded ships it reviewed, about one in five. The lessons were to test steel for toughness at service temperature, round the corners, and weld carefully. Rounded corners on every aircraft window are the same lesson.

The point for a layperson: strength and toughness are different, and a notch makes a tough material act brittle. Any material that cannot yield at the crack tip needs inspection for flaws, since the flaw size decides the failure load. Glass and ceramics, which tolerate no flaws, are therefore used in compression where cracks stay closed.

Data Worked estimate 10: the crack size at which fracture begins
Material and stateK (MPa times root m)Stress (MPa)Critical crack = (K / stress) squared / pi
Tough structural steel, warm50200(0.25) squared / 3.14 = 0.020 m, so 20 mm
The same steel grade, cold and brittle20200(0.10) squared / 3.14 = 0.0032 m, so 3 mm
Window glass0.730(0.0233) squared / 3.14 = 0.00017 m, so 0.17 mm
Hole in a platestress concentration factor 3100 averageLocal stress 300 MPa

Crack sizes calculated. The formula treats a small edge crack in a large plate and ignores geometry factors near 1.1, so use these as rough orders of magnitude. Check: a glass cutter works by scoring a scratch a fraction of a millimetre deep, which matches a flaw of about 0.2 mm, while a 20 mm crack in a steel plate would be easily seen on inspection.

Source: Petroski, To Engineer Is Human (Macmillan, 1985), ch. 10, 'When Cracks Become Breakthroughs'
  • Petroski, To Engineer Is Human (Macmillan, 1985) · ch. 10, 'When Cracks Become Breakthroughs'
  • Petroski, To Engineer Is Human (Macmillan, 1985) · ch. 11, 'Of Bus Frames and Knife Blades'
  • Kosky et al., Exploring Engineering (Academic Press, 2013) · ch. 11, 'Materials Engineering'
  • TWI Ltd, 'Schenectady T2 tanker' (read November 2026), summarising the 1946 US Board of Investigation · 4,694 welded ships, 970 with fractures; notch-sensitive steel; https://www.twi-global.com/media-and-events/insights/schenectady-t2-tanker

5.2 Fatigue breaks parts by repeated loads that are each far below the strength

A phone charging cable usually fails at the neck of the plug, after months of small bends that each did no visible harm. The copper strands inside have cracked one by one. This is fatigue. Each cycle of stress opens a microscopic crack a little further. The crack is too tiny to see for most of the life of the part. When the crack has grown to the critical size of the previous section, or the uncracked part can no longer carry the load, the part fails suddenly. Fatigue strikes bridges, cranes, railway axles, aircraft skins and bicycle frames.

Three rules guide engineers. First, what matters is the range of stress in each cycle, not only its peak. Second, the life depends steeply on that range. For welded steel details, life is roughly proportional to 1 / range cubed, so cutting the stress range by a fifth about doubles the life. Third, cracks start at stress concentrators: the toe of a weld, a sharp corner, a bolt hole, a scratch. Smooth, polished shapes live longer than welded ones. Steel often has a stress range below which cracks do not grow, the endurance limit. Aluminium does not, so it eventually fails at any stress if the cycles are enough.

The estimate counts cycles. The numbers are large. A bridge sees 90 million truck passes in 50 years. A railway axle turns 350 million times in a million kilometres. Seen this way, fatigue is an accounting job: count the cycles of a part's life, then design its details so that the stress range is below what that count allows.

In 1954 two Comet jet airliners broke up in flight, killing 35 and 21 people. Wreckage of the first showed a fatigue crack at a cut-out in the cabin roof. A complete fuselage, pressurised and relaxed in a water tank, split after about 3,060 cycles counting its earlier flights, from the corner of a hatch. British rules then required fatigue tests on a complete pressure cabin, and later airliners were designed to hold together with a crack present. How that lesson was turned into a standing process is told in the module Design, failure and safety.

Data Worked estimate 11: counting the cycles in a part's life
PartWorkingCycles in its life
Highway bridge girder detail5,000 heavy trucks a day x 365 x 50 years9.1 x 10 to the 7th (91 million)
Railway axle, wheel 0.92 m acrosscircumference 2.89 m; 1,000,000 km = 10 to the 9th m; 10 to the 9th / 2.893.5 x 10 to the 8th (350 million)
Wind-turbine blade, 15 rpm, 20 years15 x 60 x 24 x 365 x 201.6 x 10 to the 8th (160 million)
Aircraft fuselage, 4 flights a day, 25 years4 x 365 x 2536,500 (about 4 x 10 to the 4th)
Effect of cutting stress range by 20 per centlife varies as 1 / range cubed: 1 / 0.8 cubed = 1 / 0.512life about 1.95 times longer

All results calculated. Check: the aircraft has a few thousand times fewer cycles than the axle, but each is a big one, because the cabin is pressurised by a little over half an atmosphere. Fatigue depends on both the count and the size of the cycles.

Source: Petroski, To Engineer Is Human (Macmillan, 1985), ch. 3, 'Lessons from Play; Lessons from Life'
  • Petroski, To Engineer Is Human (Macmillan, 1985) · ch. 3, 'Lessons from Play; Lessons from Life'
  • Petroski, To Engineer Is Human (Macmillan, 1985) · ch. 8, 'Accidents Waiting to Happen'
  • Kosky et al., Exploring Engineering (Academic Press, 2013) · ch. 11, 'Materials Engineering'
  • Aerossurance, 'Common Comet Misconceptions and Collaborative Contribution to Safety' (read November 2026), summarising the Court of Inquiry report CAP 127 (1955) · G-ALYP 35 dead, G-ALYY 21 dead; ADF cut-out origin; G-ALYU tank test failure after 3,060 cycles; 1956 BCAR fatigue-test rule; https://aerossurance.com/safety-management/comet-misconceptions/

5.3 Corrosion thins the steel quietly, and creep lets loads deform what strength alone would hold

Iron is mined as oxide and wants to return to it. In moist air or water the steel gives up electrons to oxygen and turns to rust. Rust takes up several times the volume of the metal it came from, so rust on steel inside concrete pushes the cover off, opens the surface and lets in more moisture. Chloride from sea air or de-icing salt breaks the protective film in concrete, and carbon dioxide slowly lowers the alkalinity, so cover thickness is specified generously, from 20 mm in a mild setting to 75 mm in extreme ones. Galvanising, paint and stainless steel are other defences. The site covers the chemistry under Passivation and corrosion resistance and Stainless steel.

Corrosion is dangerous because it is hidden and it cuts the load-carrying area. In October 2022 the old suspension bridge at Morbi in Gujarat, built in 1887 and reopened after repairs only four days earlier, collapsed with a crowd on it, and 135 people died. The Special Investigation Team reported that one of the two main cables, made of seven strands of seven wires, had 22 of its 49 wires corroded and probably broken before the day. Old and new hangers had been welded together, a rigid aluminium deck had replaced timber planks, no load test was done before reopening, and nearly 300 people were on it, far more than it could carry. The main failure of the cable is plain physics, and the estimate works it out.

Creep is the other slow enemy: the steady growth of strain under a constant load. Creep adds roughly one and a half times the first deflection of a concrete slab, so a slab that sags 8 mm when the shuttering is removed can reach 20 mm after years, enough to crack partitions and spoil drainage. Timber beams sag for decades. Plastic pipes and chairs creep within months, faster when hot. Steel creeps only when hot, above roughly 400 degrees Celsius, which matters in boilers and furnaces.

The design response is the same for both: allow for the loss in advance, and look often. A cable's rating is calculated for its new section. After thirty years of weather, that number is no longer true.

Data Worked estimate 12: a 49-wire cable with 22 wires lost, and a creeping slab
StepWorkingResult
Wires in the cable7 strands x 7 wires49
Wires lost22 corroded and probably broken22
Area left27 / 4955 per cent
Stress in the remaining wires, same load1 / 0.551.8 times higher
Factor of safety if it was 3.0 when new (an illustrative value)3.0 x 0.551.65
Crowd adds 50 per cent to the load1.65 / 1.51.1, almost no margin left
Slow-creep slab: 8 mm sag, long-term factor about 2.68 x 2.6about 21 mm, above a span/250 limit of 18 mm for a 4.5 m slab

All results calculated; the creep factor follows the IS 456 creep coefficient of 1.6 for loading at 28 days, and span/250 is its final deflection limit. Check: wires that are already broken cannot share load, so the real picture is worse than the average, because stress piles onto the survivors and each snapped wire sheds load to its neighbours. The 3.0 margin is assumed for illustration, not the bridge's actual figure.

Source: Press Trust of India, 'Corroded wires, welded suspenders behind Morbi bridge tragedy: SIT report', Business Standard (19 February 2023), https://www.business-standard.com/article/current-affairs/corroded-wires-welded-suspenders-behind-morbi-bridge-tragedy-sit-report-123021900854_1.html

5.4 A factor of safety is the margin that pays for what the calculation does not know

The factor of safety is the load that would cause failure divided by the load expected in service. A factor of 2 means the structure could carry double its expected load. Why not use a factor of 1.0 and save material? Because every number in a design is uncertain. Real loads vary and sometimes exceed the code values. Steel and concrete come in batches with scatter. The calculation ignores connections, residual stress and local effects. Builders make mistakes, and structures age. The factor is the margin that covers the sum of these. Its size is not derived from theory: code committees revise it in the light of experience, after failures and after long records of designs that worked. The module What engineers do introduced the safety factor as a rule of thumb, and here it gets numbers.

Modern Indian codes split the factor into separate parts, called partial factors. IS 456 for concrete multiplies loads by 1.5, divides concrete strength by 1.5 and steel strength by 1.15. Each uncertain item gets a margin sized to how uncertain it is: loads and concrete are more variable than factory-made steel. The estimate stacks them for the beam of Part 4 and finds that the beam in effect yields at about twice the service load.

Safety factors differ across fields because the consequences and the knowledge differ. Stay cables are limited to about 45 per cent of breaking load, a factor above 2. Wire-rope slings on a crane are commonly used at 5. Lift suspension ropes, which carry people, use factors of 12 or more under the European lift standard EN 81-20. Airliners must carry 1.5 times the biggest load expected in service under US rules, a small figure because every kilogram of margin is paid for in fuel, so they depend on tests, inspection and redundancy instead.

A factor of safety does not cover blunders or abuse. It does not rescue a structure whose users double the crowd or whose cable has lost half its wire. It is a margin against honest uncertainty, and it must be maintained, as the cost of safety is paid over the life of the structure, a trade-off taken up in the module Cost, trade-offs and resources.

Data Worked estimate 13: how the margins stack in the Part 4 beam
MarginWorkingValue
Load factor in the codeservice moment 36.6 kN m to design moment 55 kN m1.5
Steel material factor500 MPa to 435 MPa1.15
Rounding the steel up to whole bars402 mm2 chosen against 351 mm2 needed1.15
Product1.5 x 1.15 x 1.15about 2.0
Direct check of yield moment500 N/mm2 x 402 mm2 x 360 mm = 72 kN m, against service moment 36.6 kN m1.98
Strand stay in Part 3working at 45 per cent of breaking strength1 / 0.45 = 2.2

All results calculated. Check: both ways of reckoning the beam give about 2. This is the margin against first yielding of the bars. Actual collapse needs a further rise of load, since the steel strain-hardens and the concrete can redistribute load, so the true factor against collapse is somewhat higher.

Source: Bureau of Indian Standards, IS 456:2000, Plain and reinforced concrete, cl. 5.6.3, 6.2.2, 6.2.5.1, 23.2, 26.4 (table 16), 26.5.1.1, 36.4.2, 38.1 and table 18; https://law.resource.org/pub/in/bis/S03/is.456.2000.pdf
  • Petroski, To Engineer Is Human (Macmillan, 1985) · ch. 9, 'Safety in Numbers'
  • Kosky et al., Exploring Engineering (Academic Press, 2013) · ch. 11, 'Materials Engineering'
  • Petroski, To Engineer Is Human (Macmillan, 1985) · ch. 13, 'The Ups and Downs of Bridges'

See it in the table

H₂O WaterCu CopperAl AluminiumSt SteelNaCl SaltTm TimberO2 OxygenCO2 Carbon dioxide

Check your understanding

A steel plate with a hole has an average stress of 100 MPa. Roughly what is the stress at the edge of the hole?

A round hole in a wide plate under tension raises the local stress to about three times the average, so 300 MPa. This is why holes and sharp corners are where cracks start.

Why does cutting the stress range by 20 per cent nearly double the fatigue life of a welded steel detail?

For welded steel, the number of cycles to failure falls steeply, as the cube of the stress range. A small cut in the range therefore gives a large gain in life.

If a cable loses 45 per cent of its wire area to corrosion, with the load unchanged, what happens to the stress in the remaining wires?

Stress is force / area. With 55 per cent of the area left, stress rises by 1 / 0.55 = 1.8 times, and the factor of safety falls by the same ratio.
← Part 4: Concrete, timber, plastics and composites: building with what each does wellMark read · Wrap-up →