CH NEO-ZÜRICH EDITION
WEATHER · ACID RAIN 5°C
BLEND OF THE DAY · 07/ROGUE
EST. 2027
THE AEC CYBER MORNING NEWS

PAZ Kaffi

DESIGN · DEMOLITION · CAFFEINE · DISPATCH
EDITION 1122 · 22 November 2026
BROADCAST 04:42 CET
2,400 BROADSHEETS PRINTED
READ TIME · 47 MIN
Wöhler's S-N Curve: Half the Stress Range Buys Eight Times the Life
SYSTEMS
FRAME · 07:00
22-11-2026

Wöhler's S-N Curve: Half the Stress Range Buys Eight Times the Life

Why steel that carries a load once cracks after millions of smaller cycles: Wöhler's axle tests, the S-N curve and the cube law in EN 1993-1-9.

Take a paper clip and bend it open. Nothing happens. Bend it back and forth, gently, never as hard as that first bend, and somewhere around the fortieth swing it goes warm at the elbow and snaps. You never exceeded what the wire could carry once. You exceeded how often it could carry it. That is metal fatigue. August Wöhler measured it on railway axles in Frankfurt (Oder) between the 1850s and 1870, and today EN 1993-1-9 still asks every welded steel detail one question in his terms: what stress range can it survive for 2 million cycles?

What it is: Fatigue is the slow cracking of a metal under repeated loads that are each well below its static strength. Every cycle opens a microscopic flaw a little further: at a scratch, a weld toe, a bolt hole, a sharp re-entrant corner. Then the cycle closes it again. For millions of cycles the part looks perfect. Then the remaining section can no longer carry even an ordinary load, and it fails in one cycle that looks, to anyone who was not counting, like a surprise. A bridge, a crane runway, a façade bracket in a gusty corner or a stadium roof under crowd and wind is therefore checked against two questions: how much, and how often.

Why it works: The mechanism is crack growth, not overload. A crack advances a tiny distance with every cycle, and how far it advances depends mainly on how far the stress swings, the stress range Δσ, rather than on the peak it reaches. This is the sentence to own. As the Wikipedia entry on material fatigue summarises Wöhler’s 1870 conclusion, cyclic stress range matters more than peak stress. Plot stress range against cycles to failure on log-log axes and the test points fall on a straight line, the S-N curve, also called the Wöhler curve. A straight line on log-log paper is a power law: N · Δσm = constant. For welded steel details, Eurocode fixes the slope at m = 3. Life therefore falls with the cube of the range.

Put a number on it. EN 1993-1-9, as the IDEA StatiCa knowledge base lays it out, gives every detail a detail category Δσc: the stress range that detail survives for 2 million cycles. Life at any other range is NR = 2·106 · (Δσc/Δσ)3. Take a detail rated at, say, 71 MPa (an illustrative figure, not a table value). At a 71 MPa range it lasts 2 million cycles. Double the range to 142 MPa and life drops to 2·106 × (1/2)3 = 250,000 cycles, one eighth. Raise the range only 26 %, to 89.5 MPa, and life halves to about one million. Read it the other way and it becomes a design gift: halve the range and the detail lasts eight times as long.

The line does not run forever. Below a knee, the constant-amplitude limit ΔσD = (2/5)1/3 · Δσc, about 52 MPa for our 71 detail, the code switches to a shallower m = 5 branch. Below a further cut-off ΔσL, stress ranges are treated as not contributing to damage at all. So the cube law is a tool for the band above the knee; extrapolating it downward overstates the damage and misreads the code. The physics underneath is still debated. Steel and titanium show an apparent endurance limit, yet very-high-cycle studies suggest that true fatigue limits may not exist for any metal. The code’s branches are an engineering settlement with that uncertainty, written down so it can be checked.

←TODAY: EN 1993-1-9 reduces each welded steel detail to one number at 2 million cycles and a slope of 3. →3012: Zurich’s oldest standing steel will be the steel whose details were drawn to keep the range low, not the steel that was strongest. Fulcrum: Strength is a property of the material; fatigue life is a property of the drawing.

Origins: In the 1840s, railway axles began breaking under loads they had carried easily for months. Macquorn Rankine recognised in 1842 that stress concentrates at abrupt changes of section. In 1843 Joseph Glynn traced a tender-axle crack to its keyway, the sharp notch cut for a key. Back in May 1842, the Meudon accident near Versailles followed a broken locomotive axle. The question was no longer academic.

The man who turned the question into a method was August Wöhler (1819–1914), born in Soltau, for 23 years Obermaschinenmeister of the Niederschlesisch-Märkische Bahn in Frankfurt (Oder), where the railway’s main repair works stood. Tasked by the Prussian Ministry of Commerce with the cause of axle fractures, he built apparatus that loaded railway axles over and over, through the 1850s to 1870. Behind those rigs stood workshop hands whose names the record does not keep: the fitters who mounted axle after axle and kept the machines turning until something cracked. Every S-N curve drawn since carries their patience. Wöhler showed the work at the Paris Exhibition of 1867, and that brought it to an international audience. He published it in Zeitschrift für Bauwesen in 1870, with an English abstract a year later. He showed that fatigue grows as cracks from surface defects, that a fracture surface can be read like a diary of its loading, and he pressed for state testing of iron and steel. The first Materialprüfungsanstalt opened in Berlin-Gross-Lichterfelde in 1904; it is the forerunner of today’s BAM. William Fairbairn in Britain was testing systematically in the same years.

The curve became a rule in stages. O. H. Basquin, in 1910, put Wöhler’s data on a straight log-log line, which gave us the power law. Arvid Palmgren in 1924 and Milton A. Miner in 1945 supplied the damage sum: add up n/N for every stress range a part sees, and call it failure when the sum reaches 1. Miner’s rule is how a real, messy load history gets counted against a clean curve. The aircraft world learned to count in 1954. At RAE Farnborough, under Director Sir Arnold Hall, a complete Comet fuselage was pressure-cycled in a water tank until it burst open after 3,057 cycles. In 1981 the investigation into the 1980 loss of the Alexander L. Kielland platform traced a fatigue crack back to a small 6 mm fillet weld. Both are footnotes to Wöhler’s lesson: the crack begins at the detail.

PAZ has drawn the static side of this story before. Our graphic statics piece on Culmann’s Zürich showed equilibrium as a closed force polygon. Fatigue adds the clock. The same force diagram, swinging between two load cases a few million times, is a different design problem.

In practice: Fatigue lives in details, and details are things you draw. The weld toe, the cope, the bolt hole, the cut-out corner of a stiffener: each one sets a detail category, and the category sets how many cycles the joint survives at a given range. That moves part of the fatigue decision onto the architect’s and the façade designer’s screen in Rhino. A softened corner radius instead of a sharp re-entrant notch. A weld moved away from the stress hotspot. A bracket shaped so the live-load swing it carries drops by a fifth. That fifth is worth roughly double the life, by the cube. A footbridge deck, a cantilevered balcony bracket on a busy façade, a crane runway in a timber-and-steel hall, a cable-net anchor in the wind: these are the places where how often outweighs how much. The trade-off is plain: a fatigue-friendly detail often costs more fabrication time and more steel at the joint, and that cost is paid up front while the benefit arrives decades later.

Atelier: An office working in Grasshopper already knows where every repeated connection in its model sits; the engineer does not always see it. The shift is to stop treating a connection as a point and start treating it as a detail with a fatigue identity that travels with the geometry. Monday move: add a detail_category user-text attribute to every repeated steel connection in your Rhino model, leave it blank where unknown, and colour the blanks red so the engineer sees exactly where the counting still has to happen.

Hack: Turn the cube law into a three-line life check you can drop into a GhPython component, with a guard that refuses to apply it below the knee. Feed it the detail category from your attribute and the stress range from the engineer’s load case. The assert is the point: the moment a range falls under about 52 MPa for a 71 detail, the script stops you instead of quietly extrapolating. Run it and you get the knee, then 2,000,000, 250,000 and about 1,000,000 cycles. Those are the worked numbers above, now attached to geometry.

dsc, m = 71.0, 3                    # detail category [MPa] (illustrative), Eurocode slope
knee = (2/5)**(1/3) * dsc           # ~52.3 MPa: the m = 3 line ends here
def life(ds): assert ds >= knee, 'below knee: use m = 5 branch'; return 2e6 * (dsc/ds)**m
print(round(knee, 1), [round(life(s)) for s in (71, 142, 89.5)])

From the late 2070s, the steel I trust is not the steel that was strongest on delivery. It is the steel whose drawings still say which detail category each joint was given, and why. Write the range down, name the category, keep the derivation next to the geometry. Open your model today and tag the first repeated connection.

Sources & Further Reading

FILED FROM
CO-SIGNERS
PAZ Academy
CONFIDENCE
HIGH
REPRINTS
© PAZ - PARAMETRIC ACADEMY ZURICH · ALL RIGHTS RESERVED

PAZ Kaffi · multidisciplinary editorial, led by PAZ Academy

⚑ REPORT AN ERROR · SUBMIT A CORRECTION
◂ BACK TO FRONT PAGE · PAZ KAFFI

© 2026 PAZ Academy.