Failure that takes time
Most structures that fail do not fail on their first day. A railway axle carries a load it could bear a million times over, and breaks on the hundred-millionth turn. A turbine blade glowing at over 1,000 °C stretches by a fraction of a millimetre a year until it touches its casing. A steel pipeline buried in damp soil loses a little metal every day until it leaks. Fatigue, creep and corrosion are the slow ways of failing, and together they decide the lives of aircraft, power plants, bridges, ships, implants and electronics.
Because they are slow, they can be measured, predicted and managed, and that is the theme of this lesson. Lumenary ME's failure and fatigue gives the designer's view of fatigue (S–N curves, mean-stress corrections and Miner's rule), and Lumenary Chemistry's electrochemistry explains corrosion cells, electrode potentials and Faraday's law. Here we look inside: how cyclic slip starts a crack and how fast the crack grows, how hot metals flow by dislocation climb and diffusion, and how metals corrode at a mixed potential, with what materials scientists do about each.
How a fatigue crack starts
Plastic slip is not perfectly reversible. When a metal is loaded back and forth, even at a stress amplitude well below its yield strength, a few favourably oriented grains near the surface slip a little in each half-cycle. The dislocations organise themselves into persistent slip bands, thin lamellae that carry nearly all the cyclic strain; they were named in the 1950s because they reappeared in the same places after the surface was polished. Where a band meets the surface, the to-and-fro slip does not exactly cancel: it pushes out thin ribbons of metal, extrusions, and leaves sharp slots, intrusions. An intrusion a few micrometres deep is already a crack.
The crack first grows along the slip band, at an angle to the load (stage I), through a grain or two. Then it turns to run perpendicular to the largest principal stress (stage II), opening and closing with each cycle. Each cycle blunts the tip and re-sharpens it, advancing the crack by a small step and often leaving a ripple, a striation, on the fracture surface. Under variable loading, pauses and changes in the load leave the larger beach marks that can be seen with the eye. Failure comes when the remaining section can no longer carry the peak load, and it is the ordinary fracture of Fracture and toughness.
In engineering parts the slip-band stage is usually skipped: cracks start at whatever concentrates stress at or near the surface, such as machining marks, the toe of a weld, a corrosion pit, an inclusion or a pore. So fatigue strength depends heavily on processing. Cleaner steels, smoother surfaces and compressive residual stress all help: shot peening blasts the surface with small hard balls, and carburising and nitriding harden it, both leaving the surface in compression that a crack must overcome. Metal parts made by additive manufacturing are especially sensitive, because pores and unfused powder act as ready-made cracks, so critical printed parts are usually hot isostatically pressed to close the pores, and machined.
In high-cycle fatigue the stress stays mostly elastic and lives run to millions of cycles (the S–N curves of ME's lesson). In low-cycle fatigue, such as the thermal cycles of a turbine disc at each start-up, every cycle yields plastically and lives are thousands of cycles; Louis Coffin and Samuel Manson found in the 1950s that the plastic strain amplitude then follows , with typically between and .
How fast a fatigue crack grows
Once a crack is long enough for fracture mechanics to apply, its growth per cycle depends on the range of the stress intensity factor during the cycle, .
Definition— The Paris law
Over the middle range of , the crack growth per cycle follows a power law,
with about 2–4 for metals in air and much higher (10 or more) for ceramics. Paul Paris and his colleagues proposed it in 1961–1963, against some resistance from journal referees.
The full curve has three regions. Below a threshold , typically 2–8 MPa·m for steels, a long crack does not grow measurably; this is fracture mechanics' version of the endurance limit. In the Paris region, rates run from about a nanometre to a micrometre per cycle. Near the top, approaches the fracture toughness and the crack accelerates to final fracture. The curve also depends on the stress ratio : Wolf Elber found in 1970 that the faces of a fatigue crack touch before the load reaches its minimum (crack closure), so part of each cycle does no damage, and higher mean stresses, which hold the crack open, grow it faster. Very short cracks, a grain or two long, break the rules and can grow below the threshold, which is why the fatigue limit of a smooth specimen and the threshold of a long crack are linked only through careful models.
Computing it: how long does a crack take to grow?
Integrating the Paris law from an initial crack to the critical crack gives the number of cycles,
With a constant it has a closed form; for a real geometry, where changes as the crack grows, it is integrated numerically. The program takes a 40 mm wide steel plate with an edge crack, loaded from 0 to 120 MPa in each cycle, with a fracture toughness of 80 MPa·m and Paris constants of the size often used as an upper bound for ferritic–pearlitic steels in air. The geometry factor is a standard handbook fit for an edge crack. It starts from a crack of 0.5 mm, roughly the smallest that routine inspection can reliably find.
import numpy as np
# Fatigue crack growth from an edge crack in a steel plate, by the Paris law
C, m = 6.9e-12, 3.0 # m/cycle with dK in MPa m^0.5: a common upper bound for ferrite-pearlite steels in air
W = 0.040 # m, plate width
s_max, s_min = 120.0, 0.0 # MPa, constant-amplitude cycle (R = 0)
K_Ic = 80.0 # MPa m^0.5
a0 = 0.5e-3 # m, the smallest crack inspection can be relied on to find
def Y(a):
"""Geometry factor for a single edge crack in a plate in tension (handbook fit, a/W <= 0.6)."""
x = a / W
return 1.12 - 0.231 * x + 10.55 * x**2 - 21.72 * x**3 + 30.39 * x**4
def K(a, s):
return Y(a) * s * np.sqrt(np.pi * a)
lo, hi = a0, 0.6 * W # critical crack: K_max = K_Ic, found by bisection
for _ in range(60):
mid = (lo + hi) / 2
lo, hi = (mid, hi) if K(mid, s_max) < K_Ic else (lo, mid)
a_c = lo
print(f"critical crack length a_c = {a_c*1e3:.2f} mm (a/W = {a_c/W:.2f})")
a = np.geomspace(a0, a_c, 20001)
dNda = 1.0 / (C * K(a, s_max - s_min) ** m) # cycles per metre of growth
N = np.concatenate([[0.0], np.cumsum(0.5 * (dNda[1:] + dNda[:-1]) * np.diff(a))])
print(f"cycles from {a0*1e3:.1f} mm to failure: {N[-1]:,.0f}")
# Closed form with a constant Y = 1.12 (a long plate, small crack), for comparison
k = C * (1.12 * (s_max - s_min) * np.sqrt(np.pi)) ** m
N_const = (a0 ** (1 - m / 2) - a_c ** (1 - m / 2)) / (k * (m / 2 - 1))
print(f"same calculation with constant Y = 1.12: {N_const:,.0f}")
print("\n crack (mm) dK (MPa m^0.5) da/dN (um/cycle) share of life used")
for target in [0.5e-3, 1e-3, 2e-3, 4e-3, 8e-3, 16e-3, a_c]:
i = np.argmin(abs(a - target))
print(f" {a[i]*1e3:6.2f} {K(a[i], s_max - s_min):6.1f} {C * K(a[i], s_max - s_min)**m * 1e6:8.4f}"
f" {N[i] / N[-1]:6.1%}")The plate fails when the crack reaches about 19 mm, after some 706,000 cycles; ignoring the growth of with crack length would overestimate the life by about 14%. The striking column is the last one. Forty per cent of the life is spent growing the crack from 0.5 to 1 mm, and 86% before it reaches 4 mm, while the last 11 mm take only 4%. A fatigue crack spends nearly all its life small, which has two consequences. Life is extremely sensitive to the initial flaw size, so cleaner material and better surfaces pay off handsomely. And inspection must find cracks early: by the time a crack is big enough to see easily, little life remains.
This is the basis of damage-tolerant design, required for aircraft since the 1970s. The engineer assumes a crack just below the detectable size is present, calculates the cycles to grow it to critical, and sets the inspection interval at a fraction of that (a half or a third), so that every crack is found at least twice before it becomes dangerous.
Creep: when metals flow
At room temperature a steel bolt under a steady load stretches elastically and stops. At 600 °C it keeps stretching, slowly and permanently, for as long as the load stays: it creeps. Creep matters when the homologous temperature (both absolute) exceeds about 0.4, because then atoms diffuse fast enough for dislocations and grain boundaries to rearrange under load: above roughly 400 °C for steel. Lead, melting at 327 °C, is already at at room temperature, which is why old lead roofs sag; tin-based solders live at about , so the solder joints of every circuit board creep, and on–off thermal cycles make them fail by creep–fatigue. Glaciers flow by the creep of ice.
In the primary stage the material work-hardens and the rate falls. In the secondary stage hardening is balanced by recovery, and the strain grows at a steady minimum creep rate, the number design is usually based on. In the tertiary stage the section thins, voids grow on grain boundaries and the microstructure coarsens, until the part ruptures.
Definition— Creep mechanisms
- Dislocation (power-law) creep: dislocations blocked by obstacles escape by climbing out of their slip planes, which needs vacancies to diffuse to or from them. The minimum creep rate follows , with typically 3–8 and close to the activation energy of self-diffusion (Diffusion).
- Diffusional creep: at low stress, atoms diffuse from grain boundaries under compression to boundaries under tension, and the grains lengthen. Through the grains (Nabarro–Herring creep) the rate is proportional to ; along the boundaries (Coble creep) to , where is the grain size.
- Grain-boundary sliding accompanies diffusional creep and opens cavities where boundaries meet.
Harold Frost and Michael Ashby charted which mechanism wins at each stress, temperature and grain size, for dozens of materials, in deformation-mechanism maps. They explain the rules for creep resistance, almost the opposite of those for strength at room temperature: large grains, or none (the hottest turbine blades are single crystals, with no boundaries to slide or carry Coble creep); slow diffusion (a high-melting base metal plus heavy elements such as rhenium and tungsten); and stable obstacles (a high fraction of coherent precipitates that do not coarsen). Single-crystal blades, developed in the United States in the 1970s, run at metal temperatures above 1,000 °C in gas hotter than their melting point, protected by internal air cooling and ceramic thermal-barrier coatings of yttria-stabilised zirconia. In power stations, 9–12% chromium steels strengthened by fine carbonitrides allow steam at 600–620 °C, and nickel alloys are needed for the 700 °C plants that would raise efficiency further (ME's power plants and energy conversion).
Creep tests at service conditions would take decades, so engineers test hotter or at higher stress and extrapolate. The most common tool is the Larson–Miller parameter: tests at one stress that rupture after hours at absolute temperature are assumed to share the value
with about 20 for many steels. A plot of stress against from short tests then gives the stress for any combination of temperature and life.
Computing it: creep rates and rupture lives
The program first shows how sensitive power-law creep is to temperature and stress, with an activation energy and stress exponent typical of dislocation creep in steels, at 600 °C. It then fits a Larson–Miller constant to two illustrative rupture tests at the same stress, uses it to extrapolate, and asks how much a modest scatter in one test changes the answer.
import numpy as np
R = 8.314 # J/(mol K)
# 1. Power-law creep: rate = A * sigma^n * exp(-Q / RT). How sensitive is it?
Q, n = 300e3, 5.0 # J/mol and stress exponent, typical of dislocation creep in steels
T = 873.15 # K (600 C)
print("change in conditions creep rate multiplied by")
for dT in [10, 25, 50]:
print(f" {dT:3d} K hotter {np.exp(Q / R * (1 / T - 1 / (T + dT))):6.2f}")
for f in [1.1, 1.25, 1.5]:
print(f" stress x {f:<4} {f**n:6.2f}")
# 2. Larson-Miller: P = T (C + log10 t_r) is the same for all tests at one stress
tests = [(700.0, 100.0), (650.0, 1550.0)] # (temperature C, hours to rupture) at one stress: illustrative
(T1, t1), (T2, t2) = [(Tc + 273.15, t) for Tc, t in tests]
C = (T2 * np.log10(t2) - T1 * np.log10(t1)) / (T1 - T2)
P = T1 * (C + np.log10(t1))
print(f"\nfitted constant C = {C:.1f}, Larson-Miller parameter P = {P:,.0f}")
for Tc in [600, 575, 550]:
life = 10 ** (P / (Tc + 273.15) - C)
print(f" predicted life at {Tc} C: {life:12,.0f} h ({life / 8766:6.1f} years)")
T_100k = P / (C + 5) - 273.15
print(f" temperature for 100,000 h: {T_100k:.0f} C")
# 3. The same extrapolation if the second test had lasted 10% less
t2b = 0.9 * t2
Cb = (T2 * np.log10(t2b) - T1 * np.log10(t1)) / (T1 - T2)
Pb = T1 * (Cb + np.log10(t1))
print(f"\nwith the second test at {t2b:,.0f} h: C = {Cb:.1f}; life at 550 C = {10 ** (Pb / 823.15 - Cb):,.0f} h")Ten degrees hotter makes this steel creep 60% faster, and 50 degrees more than nine times faster; a 10% rise in stress has about the same effect as 10 degrees. That is why the temperature of a boiler tube or turbine blade is monitored so closely, and why a blocked cooling passage in a blade is serious. The Larson–Miller extrapolation from two tests of 100 and 1,550 hours predicts a life of about 33,000 hours at 600 °C and over a million at 550 °C, and says that 100,000 hours needs a temperature below 583 °C.
Corrosion as electrochemistry
Metals are found in nature as oxides, sulfides or carbonates, and smelting them takes energy; corrosion gives it back. As Lumenary Chemistry's electrochemistry explains, wet corrosion is a short-circuited electrochemical cell: at anodic sites the metal dissolves (), at cathodic sites the electrons reduce something, usually dissolved oxygen () or, in acids, hydrogen ions, and the metal itself carries the current between them.
Thermodynamics: can it corrode? Marcel Pourbaix, in Belgium, mapped for each metal the potentials and pH values at which it is stable (immunity), dissolves (corrosion) or forms a solid oxide or hydroxide that may protect it (passivity). A Pourbaix diagram says what is possible, not how fast it happens. Aluminium, titanium and chromium are thermodynamically eager to corrode, yet in neutral water they are protected by oxide films a few nanometres thick.
Kinetics: how fast? On a single piece of freely corroding metal both reactions happen at once, and they must run at the same total rate, because the metal cannot store charge. Carl Wagner and Wilhelm Traud put this as mixed-potential theory in 1938: the metal settles at the corrosion potential at which the anodic and cathodic currents are equal, and that common current is the corrosion current . Plotting each reaction's potential against the logarithm of its current gives an Evans diagram, named after the Cambridge corrosion scientist Ulick Evans; the two lines cross at the corrosion point.
The program models steel in still, aerated water with Tafel kinetics for both reactions (illustrative constants) and a limit on the oxygen supply, then connects the steel to a noble metal of various areas.
import numpy as np
# Mixed-potential model of steel in aerated, near-neutral water (illustrative kinetic constants)
E_Fe, i0_Fe, b_Fe = -0.62, 1e-8, 0.060 # V vs SHE, A/cm^2, V/decade: iron dissolution (Tafel)
E_O2, i0_O2, b_O2 = 0.80, 1e-10, 0.120 # oxygen reduction on the metal surface
i_L = 1e-5 # A/cm^2: oxygen-diffusion limit in still water
def i_anodic(E):
return i0_Fe * 10 ** ((E - E_Fe) / b_Fe)
def i_cathodic(E):
i_act = i0_O2 * 10 ** ((E_O2 - E) / b_O2)
return 1 / (1 / i_act + 1 / i_L) # activation and diffusion resistances in series
def corrosion_potential(area_ratio=0.0):
"""Potential at which iron's anodic current equals oxygen reduction on iron plus a cathode of
area_ratio times the iron's area (for example copper), per unit area of iron."""
lo, hi = -0.62, 0.30
for _ in range(100):
E = (lo + hi) / 2
if i_anodic(E) < i_cathodic(E) * (1 + area_ratio):
lo = E
else:
hi = E
return E
E = corrosion_potential()
i_corr = i_anodic(E)
um_per_year = 11.6 * i_corr * 1e6 # iron: 1 uA/cm^2 removes 11.6 um per year
print(f"bare steel: E_corr = {E:.3f} V vs SHE, i_corr = {i_corr*1e6:.1f} uA/cm^2, about {um_per_year:.0f} um per year")
print("\nsteel coupled to a noble metal (oxygen reduced on both):")
print(" cathode/anode area E_couple (V) steel corrosion rate")
for r in [0.1, 1, 10, 100]:
Ec = corrosion_potential(r)
print(f" {r:7.1f} {Ec:6.3f} {11.6 * i_anodic(Ec) * 1e6:7.0f} um/year (x{i_anodic(Ec) / i_corr:.1f})")Bare steel corrodes at 10 µA/cm², about 0.12 mm a year, the right order for steel in still seawater, and the rate is set by the oxygen supply, not by the iron: that is why flow, aeration and temperature matter so much. Coupled to a noble metal, the steel must supply the electrons for oxygen reduction over the whole cathode, so its corrosion rate rises as : double for equal areas, eleven times for a cathode ten times larger. A small anode connected to a large cathode is the worst combination in corrosion engineering: steel rivets in a copper sheet are eaten away quickly, while copper rivets in a steel sheet do little harm. The model lets the whole cathode deliver current; in reality the electrolyte's resistance limits how far the galvanic current reaches, only millimetres in a film of condensation but much farther in seawater, so damage concentrates near the junction.
The galvanic series. The order of the standard electrode potentials is not a reliable guide to galvanic corrosion, because real metals in real waters carry films and alloying. Engineers use instead a galvanic series: corrosion potentials measured in a given environment, usually flowing seawater (ASTM G82 explains how to build and use one). From most active to most noble it runs roughly: magnesium (about −1.6 V against a saturated calomel electrode), zinc (about −1.0 V), aluminium alloys, carbon steel and cast iron (−0.6 to −0.7 V), stainless steels in the active state, lead and tin, brasses and bronzes, copper (about −0.3 V), copper–nickel alloys, nickel alloys, passive stainless steels and titanium (near 0 V), and finally graphite (about +0.25 V). Graphite's place at the noble end matters in aircraft: carbon-fibre composites are galvanically noble, so aluminium in contact with them is insulated or replaced by titanium.
Example— The Statue of Liberty
The statue raised in New York harbour in 1886 has a thin copper skin on a framework of iron bars, originally insulated from each other. Over a century the insulation broke down in the salty air, and the iron, small in area and anodic to the large copper skin, corroded and swelled, pulling rivets out of the copper. In the restoration of 1984–86 the iron bars were replaced with stainless steel, newly insulated from the copper.
The program also gives the conversion between current and metal loss, , and the charge a kilogram of each anode metal can deliver.
F = 96485.0 # C/mol
YEAR = 365.25 * 24 * 3600 # s
# metal: molar mass g/mol, electrons per atom, density g/cm^3
metals = {"iron": (55.845, 2, 7.874), "copper": (63.546, 2, 8.96), "zinc": (65.38, 2, 7.14),
"aluminium": (26.982, 3, 2.70), "magnesium": (24.305, 2, 1.738)}
anodes = {"zinc", "aluminium", "magnesium"} # metals used as sacrificial anodes
print("metal thickness lost per year theoretical charge")
print(" at 1 uA/cm^2 (um) per kg of anode (A h/kg)")
for name, (M, z, rho) in metals.items():
rate = 1e-6 * M / (z * F * rho) * YEAR * 1e4 # cm/s -> um per year
capacity = z * F / M * 1000 / 3600 # C/g -> A h per kg
cap = f"{capacity:6.0f}" if name in anodes else " -"
print(f"{name:10} {rate:6.2f} {cap}")
# A rough anode budget: a structure needs an average 2.0 A of protection current for 25 years
I, years = 2.0, 25
charge = I * years * YEAR / 3600 # A h
for name, practical in [("zinc", 780.0), ("aluminium-zinc-indium alloy", 2000.0)]:
print(f"{name:28}: {charge / practical:6.0f} kg of anode at a design capacity of {practical:.0f} A h/kg")Aluminium delivers the most charge per kilogram, nearly 3,000 ampere-hours in theory, which is why aluminium–zinc–indium alloys have largely replaced zinc on offshore structures. Practical capacities are lower because part of each anode corrodes without protecting anything (magnesium delivers only about half its theoretical charge); the program's design values are of the size used in offshore practice, and a real design also allows for the part of each anode that cannot be used up.
Forms of corrosion and how to prevent them
Uniform corrosion, the even thinning of a surface, is the easiest to live with: measure the rate and add a corrosion allowance to the thickness. The dangerous forms are localised:
- Galvanic corrosion, above.
- Pitting: where chloride ions break a passive film locally, the small pit becomes an anode under a large passive cathode, the solution in it turns acidic, and it grows inwards quickly. A stainless-steel tank can perforate while most of its surface looks perfect.
- Crevice corrosion: under gaskets, washers and deposits, oxygen is used up, the crevice becomes anodic, and the same acidification follows.
- Intergranular corrosion: if austenitic stainless steel is held between about 450 and 850 °C, as in the heat-affected zone of a weld, chromium carbides form on the grain boundaries and leave chromium-depleted zones alongside them that corrode ("weld decay"). The remedies are low-carbon "L" grades and grades stabilised with titanium or niobium, which tie up the carbon.
- Environment-assisted cracking: a tensile stress and a specific environment together crack a material that neither would harm alone. Austenitic stainless steels suffer chloride stress-corrosion cracking in warm chloride solutions; high-strength steels suffer hydrogen embrittlement, which matters increasingly as pipelines and tanks are designed for hydrogen; and cyclic stress in a corrosive environment (corrosion fatigue) removes the endurance limit altogether.
Prevention attacks one part of the cell or another:
- Barriers: paints and polymer coatings, metallic coatings, anodised films on aluminium. Chromate conversion coatings, long standard on aircraft aluminium, are being replaced because hexavalent chromium is toxic and its use is now restricted in many countries.
- Sacrificial protection: zinc coatings (galvanising) and anodes of zinc, aluminium or magnesium corrode instead of the steel, as in Lumenary Chemistry's lesson.
- Impressed-current cathodic protection: a direct-current supply drives the structure's potential down until it no longer corrodes. A classic criterion for buried steel is a potential of −0.85 V or lower against a copper/copper sulfate reference electrode.
- Changing the environment: removing oxygen from boiler water, raising the pH, adding inhibitors.
- Alloy design. Stainless steels contain at least 10.5% chromium, enough to form a continuous chromium-rich oxide film, a few nanometres thick, that heals itself when scratched. Molybdenum and nitrogen strengthen the film against chlorides, which is summarised in the pitting resistance equivalent number, : about 19 for the common 304 grade, about 24 for 316 with its 2% molybdenum, about 35 for 2205 duplex steel, and 40 or more for the super-duplex and super-austenitic grades chosen for seawater. Weathering steels, with a little copper, chromium, nickel and phosphorus, form a dense rust layer that slows further attack, and are used unpainted for bridges.
- Design: avoid crevices and water traps, insulate dissimilar metals, never pair a small anode with a large cathode, and allow access for inspection.
In professional practice
Fatigue. Crack-growth curves are measured to ASTM E647 in servo-hydraulic machines, and crack-growth software (NASGRO, developed with NASA, is widely used) integrates them over measured load spectra to set inspection intervals. As of 2026, qualifying additively manufactured parts for fatigue-critical use, by controlling their porosity and inspecting it with X-ray computed tomography, is one of the most active areas of the field.
Creep. Creep and rupture tests (ASTM E139, ISO 204) run for years in banks of furnaces. For plant in service, engineers estimate remaining life from operating temperatures and from replicas of the surface microstructure that reveal cavities on grain boundaries. Machine-learning models trained on decades of creep data now suggest new compositions, but tests that take years must still confirm them.
Corrosion. Pipelines, refineries, bridges and ships are managed by risk-based inspection, cathodic-protection surveys and coating maintenance, under rules from ISO, ASTM, AMPP and classification societies. New challenges include materials for hydrogen, for carbon capture and storage, and for the molten salts and liquid metals of advanced nuclear and solar plants.
Common misconceptions
Misconception
"A part that has survived a million cycles has proved itself." Most of a fatigue life is spent growing a crack too small to see, so an inspection that finds nothing says little about how much life remains unless it could have found a crack below the critical size with margin. Variable loads and corrosion can also end a life that constant-amplitude tests suggested was infinite.
Misconception
"Creep is a high-temperature problem." It is a high-homologous-temperature problem. Solder, lead, ice and many polymers creep at room temperature, and concrete creeps for years under its own weight.
Misconception
"Stainless steel does not corrode." It corrodes very slowly while its passive film is intact. Chlorides, crevices, warm water, sensitisation by welding and contact with more noble materials can all break the film, and then it pits, cracks or corrodes along grain boundaries. The grade must suit the environment.
Misconception
"The more active metal in the electrochemical series always corrodes." The galvanic series in the actual environment matters, not the table of standard potentials: passive films can make aluminium or titanium behave far more nobly than their standard potentials suggest. And how fast the anode corrodes depends on the area ratio and the conductivity of the electrolyte as much as on the potential difference.
Exercises
1. Fatigue, creep and corrosion
Warm-upHumanIn the lesson's crack-growth calculation, a crack grew from 0.5 mm to a critical 19 mm. Roughly what share of the total life was spent while the crack was shorter than 1 mm?
2. Fatigue, creep and corrosion
CoreComputationalWrite paris_cycles(a0, a1, dsigma, C, m, Y=1.12), which returns the number of cycles for a crack to grow from length a0 to a1 (metres) under a stress range dsigma (MPa), with the Paris law and a constant geometry factor , so that . Integrate in closed form, and remember that is a special case.
Check: from 1 mm to 10 mm at MPa with and , a crack takes about 237,000 cycles.
3. Fatigue, creep and corrosion
CoreHumanA creep-rupture test at 700 °C fails after 1,000 hours. Using the Larson–Miller parameter with , estimate the rupture life at the same stress and 650 °C. Give the answer in thousands of hours, rounded to the nearest thousand.
4. Fatigue, creep and corrosion
Warm-upHumanA tin–silver–copper solder melts at about 217 °C. What is its homologous temperature in a phone at 25 °C? Give 100 times the ratio, rounded to the nearest whole number.
5. Fatigue, creep and corrosion
Warm-upHumanA duplex stainless steel contains 22% chromium, 3.1% molybdenum and 0.17% nitrogen by mass. Using , what is its pitting resistance equivalent number, rounded to the nearest whole number?
6. Fatigue, creep and corrosion
CoreHumanA boat builder must join copper sheet with rivets and has only steel and copper rivets. Which statement is right?
7. Fatigue, creep and corrosion
CoreComputationalAt high temperature, metals oxidise in dry air, and whether the oxide protects them depends partly on its volume. The Pilling–Bedworth ratio (1923) is the volume of oxide formed divided by the volume of metal consumed. For an oxide containing metal_atoms metal atoms per formula unit,
Write pilling_bedworth(M_oxide, rho_oxide, M_metal, rho_metal, metal_atoms) and classify(pbr), which returns "porous" below 1, "protective" from 1 to 2 inclusive, and "spalling" above 2. Data: Al₂O₃ 101.96 g/mol, 3.95 g/cm³; Al 26.98 g/mol, 2.70 g/cm³; MgO 40.30 g/mol, 3.58 g/cm³; Mg 24.31 g/mol, 1.74 g/cm³.
Further reading and resources
- A critical analysis of crack propagation laws · P. C. Paris and F. Erdogan · 1963Journal of Basic Engineering, volume 85; the paper that established the power law for fatigue crack growth.
- Deformation-Mechanism Maps · H. J. Frost and M. F. Ashby · 1982Pergamon Press. The classic book of maps showing which creep mechanism dominates at each stress and temperature, for dozens of materials.
- A time-temperature relationship for rupture and creep stresses · F. R. Larson and J. Miller · 1952Transactions of the ASME, volume 74; the origin of the Larson–Miller parameter.
- AMPP (Association for Materials Protection and Performance)The professional society for corrosion control and protective coatings, formed from NACE International and SSPC; its IMPACT study estimated the global cost of corrosion.
- ASTM InternationalStandards for fatigue crack growth (E647), creep and rupture testing (E139), galvanic series (G82) and pitting tests (G48).
- NIST (National Institute of Standards and Technology)Measurement science and reference data for materials, including corrosion, fatigue and long-term reliability.
- DoITPoMS teaching and learning packagesFree packages from the University of Cambridge on creep, fatigue and Pourbaix diagrams.