Shell buckling — thin-walled cylinder under axial load: Question bank

Definition Symbol glossary (read first — every symbol used below)

So this page stands on its own, here is every symbol before you meet it in a question:

$E$ — Young's modulus: how stiff the wall material is (Pa). Bigger $E$ = harder to stretch.
$ν$ — Poisson's ratio: how much the material bulges sideways when squeezed (≈0.3 for metals).
$t$ — wall thickness (m). "Thin" means $t ≪ R$ .
$R$ — cylinder radius (m).
$L$ — cylinder length (m); the ratio $L / R$ decides which buckle mode wins (see the mode box below).
$w$ — the small radial dimple displacement of the wall (in/out) during buckling.
$σ_{cr}$ — the classical critical stress: the axial squeeze a perfect cylinder can take before buckling, $σ_{cr} = \frac{E}{3 ( 1 - ν ^{2} )} \frac{t}{R} \approx 0.605 E \frac{t}{R}$ .
$P_{cr}$ — the classical critical load (in newtons): the total axial force at buckling, $P_{cr} = σ_{cr} \cdot 2 π R t$ (stress $\times$ load-bearing wall area).
$σ_{Y}$ — yield stress: the squeeze at which the material itself permanently deforms.
$D = \frac{E t ^{3}}{12 ( 1 - ν ^{2} )}$ — plate flexural rigidity: how hard it is to bend the wall (see Plate bending and flexural rigidity D).
$K = \frac{E t}{R ^{2}}$ — membrane (hoop) stiffness: how hard it is to radially dimple the curved wall.
$k$ — the wavenumber of a buckle pattern: how many wrinkles fit around/along the shell. Small $k$ = long stretched-out waves; large $k$ = short crinkly waves. Wavelength $= 2 π / k$ .
$γ$ — knockdown factor: the fraction ( $0.2$ – $0.7$ ) of $σ_{cr}$ a real dented cylinder actually survives; $γ = 1 - 0.901 (1 - e^{- ϕ})$ .
$ϕ = \frac{1}{16} R / t$ — a dimensionless thinness number that grows with $R / t$ ; NASA SP-8007 fits the observed test scatter to it (see the box below).

Definition Two competing buckle modes — and what selects them

A perfect cylinder can buckle in (at least) two rival patterns:

Axisymmetric (ring) mode — the wall folds into circumferential accordion rings, like a bellows; $w$ varies only along the axis.
Asymmetric (diamond/chessboard) mode — the wall dimples into a field of diamonds; $w$ varies both around and along.

For the classic long, thin, axially compressed cylinder these two have nearly the same classical stress $\approx 0.605 E t / R$ (this near-degeneracy is exactly why the shell is so imperfection-sensitive). As $L / R$ shrinks toward a short, stubby shell, boundary effects favour the axisymmetric mode and the load creeps above the local formula; as $L / R$ grows long, the diamond mode with many circumferential waves dominates. This mode competition is the geometric backdrop for several edge cases below.

The three figures below make the "geometric mean", " $L$ cancels", and "empirical knockdown" arguments visual before you meet them in the questions.

Figure — Shell buckling — thin-walled cylinder under axial load

Figure 1 — the energy of a buckle mode is $D k^{4} + K / k^{4}$ . Bending (black dashed) rises for short crinkly waves; membrane (grey dashed) rises for long stretched waves. Their sum (red) has a bowl-shaped minimum exactly where the two curves cross — that crossing is the geometric mean $2 D K$ .

Figure 2 — the buckle pattern that wins is a diamond field of dimples $w$ . Because the shell can pick how many diamonds fit, the length $L$ drops out of the answer for a long cylinder.

Figure 3 — the NASA SP-8007 knockdown curve $γ = 1 - 0.901 (1 - e^{- ϕ})$ drawn as a safe lower bound (red) under a scatter of real test points that fall well below the perfect-cylinder ceiling $γ = 1$ .

True or false — justify

Doubling the wall thickness

t

doubles the classical critical stress

σ_{cr}

True.

σ_{cr} = 0.605 E t / R

is linear in

t

, so twice the thickness gives twice the critical stress — and because the load-bearing area

2 π R t

also doubles, the critical load

P_{cr} \propto t^{2}

quadruples.

A shell that survives the classical stress

0.605 E t / R

is safe.

False. Real cylinders buckle at only

20

–

70%

of the classical value due to imperfection sensitivity; you must design to

γ σ_{cr}

, never the bare classical stress.

Buckling is a form of material yield that happens early.

False. Buckling is a stiffness/stability failure (geometry-driven loss of equilibrium), not a strength failure. The material is still elastic and below yield when the wall folds — see Yield vs stability failure modes.

The

(1 - ν^{2})

term in the flexural rigidity

D

makes a plate less stiff than a beam.

False. Since

ν < 1

, dividing by

(1 - ν^{2}) < 1

makes

D

larger: Poisson coupling suppresses anticlastic curvature, so a wide plate is stiffer than a narrow beam of the same thickness — see Plate bending and flexural rigidity D.

A flat plate of the same thickness has the same axial buckling resistance mechanism as the cylinder.

False. A flat plate has only bending stiffness

D

; the cylinder adds the membrane (hoop) term

K = E t / R^{2}

, which exists purely because of the curvature

1/ R

. That extra spring is what produces the geometric-mean formula.

Increasing the radius

R

(keeping

t

fixed) makes the cylinder harder to buckle.

False.

σ_{cr} \propto 1/ R

, so a bigger radius gives a lower critical stress — a wider can is weaker per unit wall. This is the "Rockets" (bigger

R

⇒ weaker) part of the mnemonic.

The knockdown factor

γ

approaches

1

for very thin walls (large

R / t

False. As

R / t

grows,

ϕ = \frac{1}{16} R / t

grows,

e^{- ϕ} \to 0

, and

γ \to 1 - 0.901 = 0.099

. Thinner walls are more imperfection-sensitive, so

γ

shrinks toward its floor, not toward

1

For a given material, whether the shell fails by buckling or yield depends only on

R / t

True (for the axial-compression comparison). Both classical

σ_{cr} \approx 0.605 E t / R

and

γ

depend on the single ratio

R / t

; comparing

γ σ_{cr}

against the material's

σ_{Y}

decides the governing mode — see Yield vs stability failure modes.

The axisymmetric ring mode and the diamond mode always buckle at very different loads.

False. For the long, thin, axially compressed cylinder they are nearly degenerate (both

\approx 0.605 E t / R

); that near-equal energy is precisely what makes the shell imperfection-sensitive. Only when

L / R

is small do boundary effects clearly separate them.

Spot the error

"The critical stress is

2 D K

, and since the geometric mean is always bigger than either term, buckling is stronger than each individual effect."

The dimensional/units reasoning is wrong.

σ_{cr} t = 2 D K

balances the two restoring springs

D

(bending) and

K = E t / R^{2}

(membrane); it is not comparing

σ_{cr}

D

K

(different units). As Figure 1 shows, the geometric mean is the balanced optimum of a trade-off, not "stronger than both."

"A dent that is tiny compared to the radius is far too small to matter."

The dangerous scale is the wall thickness

t

, not the radius

R

. An imperfection of order

t

(which is

≪ R

) already slashes the load, because many buckle modes are near-degenerate and the cylinder sits on a knife-edge — see Imperfection sensitivity and knockdown factors.

"Since the classical formula was derived rigorously for a perfect cylinder, it is the safest number to design with."

Rigor for the idealised case does not make it safe for the real case. Perfection is the unattainable upper bound; real hardware always has imperfections, so the rigorous perfect-cylinder stress is an unsafe overestimate.

"We can ignore the membrane term because bending stiffness

D

dominates."

Wrong for the optimum mode. Bending dominates only for short-wavelength (large

k

, many-wave) dimples; the buckle mode selects itself to balance bending and membrane equally (the bottom of the red bowl in Figure 1), so neither term can be dropped — both appear inside the geometric mean.

"Poisson's ratio

ν

only matters for how the material squishes sideways; it does not affect buckling."

It enters

σ_{cr}

through

3 (1 - ν^{2})

in the denominator (via the plate rigidity

D

). For

ν = 0.3

this fixes the famous

0.605

coefficient; a different

ν

changes it.

"The buckling load

P_{cr}

scales linearly with

t

just like the stress does."

No.

σ_{cr} \propto t

and the wall area

2 π R t \propto t

, so

P_{cr} = σ_{cr} \cdot 2 π R t \propto t^{2}

. The load scales quadratically, not linearly.

"Because ring and diamond modes give the same classical stress, it doesn't matter which one we analyse."

Their equal classical stress is exactly the problem, not a convenience: the near-degeneracy is why a tiny imperfection can trigger whichever mode is nudged first, so both must be tracked in imperfection-sensitivity and knockdown assessment.

Why questions

Why does the length

L

vanish from the classical buckling stress?

Minimising over the buckle wavenumber

k

lets the shell pick an optimal number of waves (Figure 2); for a sufficiently long cylinder that optimum is set by the local

D

-vs-hoop balance, so

L

cancels out — length only matters for short/stubby shells or global Euler column buckling.

Why is the optimum a geometric mean

2 D K

and not an average?

The energy per unit area of a buckle of wavenumber

k

D k^{4} + K / k^{4}

— bending grows for short waves (large

k

), membrane grows for long waves (small

k

). Minimising this sum (setting its derivative to zero) makes the two terms equal, and the minimised sum is

2 D K

— exactly the AM–GM statement

a + b \geq 2 ab

. Figure 1 shows this as the crossing point at the bottom of the red bowl.

Why does the hoop/membrane term

K

contain

1/ R^{2}

(curvature) while a flat plate lacks it entirely?

Pushing the wall radially by

w

on a curved surface stretches the circumference; the stored energy scales as

E t (w / R)^{2}

, i.e.

K = E t / R^{2}

. A flat plate (

1/ R = 0

) has no circumference to stretch, so this restoring spring simply doesn't exist for it.

Why do engineers use NASA SP-8007's exponential

γ = 1 - 0.901 (1 - e^{- ϕ})

instead of just a flat safety factor?

Imperfection sensitivity grows with $R / t$ , so the required reduction is not constant. The exponential fit (Figure 3) captures how the empirical scatter worsens for thinner shells, giving a curve-fit lower bound rather than one blunt number — see NASA SP-8007 buckling of thin-walled cylinders.

Why can't we treat the tank like a solid column and use Euler's buckling formula?

Euler column buckling is global bending of a slender bar governed by

E I

and length

L

; shell buckling is local wall wrinkling into rings or diamonds governed by

D

and the hoop term

K

. They are different instabilities — see Euler column buckling.

Why is axial cylinder buckling especially imperfection-sensitive compared with, say, a plate?

Many distinct buckle patterns (rings and diamonds among them) have almost the same critical energy (near-degenerate modes — the red bowl in Figure 1 is very flat near its bottom), so the perfect-cylinder equilibrium is a knife-edge; any small dent nudges it toward the lower-load mode, whereas a plate's modes are better separated.

Edge cases

What happens to the classical formula as

t / R \to 0

(extremely thin wall)?

σ_{cr} \propto t / R \to 0

: an infinitely thin wall carries essentially no axial buckling stress, and

γ \to 0.099

makes the realistic allowable even smaller — such shells are governed entirely by stability, never yield.

What happens as the wall becomes thick enough that

γ σ_{cr}

exceeds the yield stress

σ_{Y}

The failure mode switches: yield now governs and the shell fails as a material-strength problem, not a stability one. The correct design stress becomes

min (γ σ_{cr}, σ_{Y})

— see Yield vs stability failure modes.

Can the empirical formula ever give

γ = 1

for a truly perfect shell?

No — the SP-8007

γ

depends only on

ϕ (R / t)

, so it never returns exactly

1

; it always lies below the perfect ceiling. Recovering

γ \to 1

requires a different description: an imperfection-amplitude model in which the dent size is sent to zero, which reproduces the classical value only in the (unreachable) zero-imperfection limit. The empirical curve deliberately does not "know" about perfection — it fits real, always-imperfect hardware.

Does the buckling analysis change if the tank is also pressurised internally?

Yes. Internal pressure adds tension that stabilises the wall against dimpling, effectively raising the buckling load; the hoop tension partly cancels the destabilising axial compression — connect to Hoop stress in pressurised cylinders and Rocket tank and interstage structural design.

What if the applied stress is tension rather than compression?

No buckling occurs. Buckling requires compressive membrane stress to do destabilising work; tension straightens imperfections and only yield or fracture can then govern.

For a very short, stubby cylinder (

L / R

small), is the local geometric-mean formula still valid?

Not necessarily. The derivation assumed enough length for the optimal wavelength to fit; a stubby shell is boundary-condition dominated, the axisymmetric ring mode tends to win, and it may instead behave like a short column or a ring — so

L

re-enters and the clean

0.605 E t / R

no longer holds.

For a very long cylinder (

L / R

large), which buckle mode dominates?

The asymmetric diamond mode with many circumferential waves; the shell has ample length to fit the optimal wavelength, so

L

drops out and the classical

0.605 E t / R

applies — until the length becomes so great that global Euler column buckling of the whole tube takes over instead.

Recall One-line self-test (click to reveal)

Cover every answer above and re-derive the reasoning for the three you found hardest. If you can state the trap AND why it's a trap, you own the concept.