3.6.7Spacecraft Structures & Systems Engineering

Shell buckling — thin-walled cylinder under axial load

1,816 words8 min readdifficulty · medium

WHAT is buckling?


HOW we build the classical formula (derivation from scratch)

We want the critical axial stress σcr\sigma_{cr} that a perfect cylinder can carry before buckling. We derive it by analogy and energy balance rather than memorising.

Step 1 — A plate/column analogy

A flat plate strip of thickness tt resists bending with flexural rigidity D=Et312(1ν2).D = \frac{E t^3}{12(1-\nu^2)}.

Step 2 — Two ways the shell resists a dimple

When the wall pushes inward/outward by a small radial displacement ww, two restoring effects fight it:

  1. Bending stiffness DD — resists the curvature of the dimple. Dominant for short-wavelength (many waves).
  2. Membrane (hoop) stiffness — pushing the wall radially stretches/compresses the circumference, storing energy EtR2\propto \dfrac{E t}{R^2}. This is the term a flat plate does not have; it comes purely from curvature 1/R1/R.

The competition between these two sets a natural buckle wavelength, and minimising the buckling load over that wavelength gives a beautifully simple result.

Step 3 — The energy minimisation (sketch)

Assume a buckling pattern w=w0sin ⁣mπxLcosnyRw = w_0 \sin\!\frac{m\pi x}{L}\cos\frac{n y}{R}. The energy per unit area has the form UDk4  +  EtR2kx4k4,U \sim D\,k^4 \;+\; \frac{E t}{R^2}\,\frac{k_x^4}{k^4}, where kk is the total wavenumber. The applied axial stress does work σtkx2\sigma t\, k_x^2. Setting the destabilising work equal to the restoring energy and minimising over the wave numbers cancels the length LL and radius geometry into a clean combination: σcrt  =  2DEtR2.\sigma_{cr}\,t \;=\; 2\sqrt{D \cdot \frac{Et}{R^2}}.

Step 4 — Plug in DD and simplify

= \frac{2E}{R}\sqrt{\frac{t^2}{12(1-\nu^2)}}. $$ Simplifying $\sqrt{1/12}=1/(2\sqrt3)$: > [!formula] Classical (theoretical) axial buckling stress > $$ \boxed{\;\sigma_{cr} = \frac{E}{\sqrt{3(1-\nu^2)}}\,\frac{t}{R}\;}\qquad P_{cr}=\sigma_{cr}\,(2\pi R t). $$ > For metals with $\nu = 0.3$, $\sqrt{3(1-0.3^2)} = \sqrt{2.73}\approx 1.65$, so > $$ \sigma_{cr} \approx 0.605\,E\,\frac{t}{R}. $$ The number **0.605** is the famous **classical buckling coefficient** for an axially loaded cylinder. --- ## The reality gap: the knockdown factor > [!mistake] Steel-man: "Use $0.605\,E t/R$ and I'm safe." > **Why it feels right:** it's the rigorously derived answer for a *perfect* cylinder, and the derivation is clean. > **Why it's dangerous:** real cylinders buckle at only **20–70%** of this value. Tiny imperfections (dents, weld distortions, thickness variation) of size ~$t$ drastically cut the load because the axial cylinder buckling is *imperfection-sensitive* (many buckle modes have almost the same energy, so the system is on a knife-edge). > **The fix — knockdown factor $\gamma$ (NASA SP-8007):** > $$ \sigma_{allow} = \gamma\,\sigma_{cr},\qquad \gamma = 1 - 0.901\left(1-e^{-\phi}\right),\quad \phi=\tfrac{1}{16}\sqrt{R/t}. $$ > Always design with $\gamma$, not the bare classical stress. --- ## Worked examples > [!example] Example 1 — Aluminium tank wall stress > Given: aluminium $E=70$ GPa, $\nu=0.3$, $R=1.8$ m, $t=3$ mm. Find classical $\sigma_{cr}$ and $P_{cr}$. > > **Step 1** — ratio $t/R = 0.003/1.8 = 1.667\times10^{-3}$. > *Why:* the formula depends only on this geometry ratio. > > **Step 2** — $\sigma_{cr}=0.605\,E\,t/R = 0.605\times70\times10^9\times1.667\times10^{-3}$. > *Why:* using $\nu=0.3$ coefficient. $=70.6$ MPa. > > **Step 3** — cross-section area $A=2\pi R t = 2\pi(1.8)(0.003)=0.0339$ m². > *Why:* axial load is stress × load-bearing wall area. > > **Step 4** — $P_{cr}=\sigma_{cr}A = 70.6\times10^6\times0.0339 = 2.39$ MN (classical). > > **Step 5 (reality)** — $\phi=\frac1{16}\sqrt{R/t}=\frac1{16}\sqrt{600}=1.531$, $\gamma=1-0.901(1-e^{-1.531})=0.297$. > So realistic allowable $\approx 0.297\times70.6 = 21$ MPa, $P\approx0.71$ MN. *Massive* difference — that's why the knockdown matters. > [!example] Example 2 — Forecast-then-verify: double the thickness > **Forecast:** If I double $t$, how much does classical $\sigma_{cr}$ change? Guess before reading. > **Verify:** $\sigma_{cr}\propto t/R$, so doubling $t$ **doubles** the critical stress, and $P_{cr}\propto \sigma_{cr}\cdot t \propto t^2$ **quadruples**. *Why:* both the stress capacity *and* the area rise linearly with $t$. > [!example] Example 3 — Compare with yield > Steel $E=200$ GPa, yield $\sigma_Y=250$ MPa, $R/t=400$. Classical $\sigma_{cr}=0.605\times200000/400=302.5$ MPa **theoretically**. Knockdown: $\phi=\frac1{16}\sqrt{400}=\frac1{16}\times20=1.25$, so $\gamma=1-0.901(1-e^{-1.25})=1-0.901(1-0.2865)=0.358$. Thus $\sigma_{allow}=0.358\times302.5\approx108$ MPa. So this shell fails by **buckling at ~108 MPa**, well below the 250 MPa yield. *Why this step:* it proves buckling, not yield, governs — you can't just use $\sigma_Y$. --- > [!recall]- Feynman: explain to a 12-year-old (click to reveal) > Take an empty soda can and press straight down on it with your hands. For a while nothing happens — then *bang*, the sides suddenly crumple into little diamonds and it collapses. The metal wasn't "too weak"; it was **too thin to stay round** under the squeeze, so it folded. Rockets are giant soda cans, so engineers calculate exactly how hard they can push before the folding starts — and because real cans always have tiny dents, they only trust them to a fraction (roughly a third to a half) of the perfect-can number. > [!mnemonic] Remember the formula > **"0.6 E times t-over-R"** → *"Six Elephants Trample Rockets."* > - **Six** = 0.605 coefficient > - **Elephants** = $E$ > - **Trample** = $t$ (thin wall) > - **Rockets** = $R$ (in the denominator: bigger radius ⇒ weaker). --- ## Active recall #flashcards/physics What kind of failure is shell buckling — strength or stiffness? ::: A stiffness/stability failure (geometry-driven), not material yield. Classical axial buckling stress of a thin cylinder? ::: $\sigma_{cr}=\dfrac{E}{\sqrt{3(1-\nu^2)}}\dfrac{t}{R}\approx0.605\,E\,t/R$ for $\nu=0.3$. Why does a $(1-\nu^2)$ appear in the plate rigidity $D$? ::: Poisson coupling stiffens a wide plate (anticlastic curvature is suppressed), unlike a slender beam. Why is $\sigma_{cr}$ the geometric mean $2\sqrt{DK}$? ::: Two competing restoring effects (bending vs membrane hoop) balance at the optimum buckle wavelength; the optimum of a trade-off is a geometric mean. What is the physical origin of the membrane (hoop) restoring term $Et/R^2$? ::: Radial dimpling stretches/compresses the circumference; this exists only because of curvature $1/R$ (a flat plate lacks it). Why do real cylinders buckle far below the classical value? ::: Imperfection sensitivity — near-degenerate buckle modes make the cylinder a knife-edge; dents ~$t$ slash the load. Typical knockdown factor range for axial cylinders? ::: $\gamma\approx0.2$–0.7; NASA SP-8007 gives $\gamma=1-0.901(1-e^{-\phi})$, $\phi=\tfrac1{16}\sqrt{R/t}$. If wall thickness doubles, how does classical $P_{cr}$ change? ::: It quadruples, since $P_{cr}\propto\sigma_{cr}\,t\propto t^2$. Flexural (plate) rigidity formula? ::: $D=\dfrac{Et^3}{12(1-\nu^2)}$. Does $\sigma_{cr}$ depend on cylinder length $L$? ::: For moderate/long cylinders, no — $L$ cancels in the minimisation (classical result is length-independent). --- ## Connections - [[Euler column buckling]] — the 1-D analogue; here membrane action replaces the single spring. - [[Plate bending and flexural rigidity D]] - [[Imperfection sensitivity and knockdown factors]] - [[NASA SP-8007 buckling of thin-walled cylinders]] - [[Rocket tank and interstage structural design]] - [[Yield vs stability failure modes]] - [[Hoop stress in pressurised cylinders]] — pressurisation *raises* buckling load (stabilising). ## 🖼️ Concept Map ```mermaid flowchart TD A[Thin-walled cylinder] -->|loaded by| B[Axial compression] B -->|triggers at Pcr| C[Buckling instability] C -->|is a| D[Stiffness failure] C -->|NOT a| E[Yield strength failure] C -->|occurs below| E A -->|geometry R t L| F[Flexural rigidity D] F -->|scales as t^3 and 1-nu^2| G[Bending stiffness] A -->|curvature 1/R gives| H[Membrane hoop stiffness] G -->|competes with| H H -->|energy prop Et/R^2| I[Buckle wavelength] G -->|resists dimple curvature| I I -->|minimise over wavenumbers| J[Critical stress sigma_cr] J -->|guides| K[Structural design margin] ``` ## 🔊 Hinglish (regional understanding) > [!intuition]- Hinglish mein samjho > Dekho, socho ek khaali cold-drink can ko tum seedha upar se dabate ho. Metal itni jaldi "crush" nahi hota — pehle to kuch nahi hota, phir achanak diwaar diamond-shape mein fold hokar collapse ho jaati hai. Yehi hai **shell buckling**. Rocket ke fuel tank aur interstage bhi bade thin-walled cylinder hi hain, jinke upar thrust aur weight ka load axial (lambai ke direction mein) aata hai. Yaad rakho: yeh **strength** failure nahi hai (material yield nahi ho raha), yeh **stiffness/stability** failure hai — geometry ki wajah se. > > Classical formula hai $\sigma_{cr} \approx 0.605\,E\,t/R$. Iska matlab: patli deewar ($t$ chhota) ya bada radius ($R$ bada) matlab jaldi buckle. Formula ke andar do cheezein compete karti hain — **bending stiffness** (deewaar ko mudne se rokti hai) aur **membrane/hoop stiffness** (radial dimple banane se circumference stretch hoti hai, jo sirf curvature $1/R$ ki wajah se aata hai). In dono ka optimum balance geometric mean $2\sqrt{DK}$ deta hai, isliye woh clean formula banta hai. > > Sabse important practical baat: **real cylinder classical value ka sirf 20–70% hi jhelta hai**. Kyun? Kyunki chhoti-chhoti dents aur weld imperfections ($\sim t$ size ki) load bahut girा deti hain — isko **imperfection sensitivity** kehte hain. Isliye engineer hamesha ek **knockdown factor** $\gamma$ (NASA SP-8007) se multiply karke design karte hain, warna rocket udne se pehle hi fold ho jaayega. Exam aur real design dono mein: pehle classical nikaalo, phir $\gamma$ se ghata do. ![[audio/3.6.07-Shell-buckling-—-thin-walled-cylinder-under-axial-load.mp3]]

Go deeper — visual, from zero

Test yourself — Spacecraft Structures & Systems Engineering