Weibull statistics
/ VY-bull /
A chain is only as strong as its weakest link. Pull on it and it breaks not at its average link but at its single worst one, so a chain with more links is more likely to hide a bad one and snap sooner. A brittle ceramic behaves exactly the same way: it fails at its worst flaw, and because the population of flaws differs from part to part, the strength scatters. Weibull statistics is the mathematics that turns this weakest-link picture into predictions, describing not a single strength but the whole spread of strengths you would measure across many nominally identical parts.
The Swedish engineer Waloddi Weibull captured it in 1939 with a survival law. The probability that a part survives a stress sigma is P_s = exp(-(V/V_0) times (sigma/sigma_0)^m), so its probability of failure is P_f = 1 - P_s. Here sigma_0 is a characteristic strength (the stress at which about 63 percent have failed), V is the stressed volume, and m is the Weibull modulus, which measures the scatter. The form comes straight from the weakest-link idea: the whole part survives only if every one of its small volume elements survives, and multiplying all those independent survival probabilities together produces the exponential law. In practice you break twenty or thirty specimens, rank them, and plot the double logarithm ln(ln(1/(1-P_f))) against ln(sigma); the data fall on a straight line whose slope is m.
Weibull statistics is what lets an engineer design safely with a material that has no single strength. Instead of a fixed value you choose a design stress that keeps the failure probability below some target, say one in a million, and the same law predicts how strength scales with size. Two honest caveats keep it from being magic. It is an empirical weakest-link fit, not a law of nature, and it assumes one independent population of flaws; a real ceramic often has two populations, say surface scratches and internal pores, which produce a kinked, bimodal Weibull plot that a single line misrepresents. And a trustworthy modulus needs many specimens; estimating m from only a handful gives a wide, unreliable answer.
Thirty alumina bars are broken and ranked by strength. Their double-log Weibull plot falls on a straight line, giving a Weibull modulus of about 10 and a characteristic strength of 350 MPa. From this the engineer reads off that for a one-in-a-million failure probability the part must be kept below roughly 120 MPa, far under the average.
A straight double-log plot yields the modulus (slope) and characteristic strength, turning scatter into a design rule.
Weibull is an empirical weakest-link description, not a physical law, and it assumes a single flaw population. Real ceramics often have two (surface and internal), producing a bent, bimodal plot; forcing one straight line through it hides the truth. It also needs many specimens to estimate reliably.