Weibull Distribution
The Weibull Distribution: Modelling Time-to-Failure
The Weibull distribution is a flexible continuous distribution widely used in reliability engineering and business analytics to model waiting times, product lifetimes, and time-to-failure data. It is governed by two positive parameters: a shape parameter k and a scale parameter λ.
Weibull probability density functions with scale λ=1 and three different shape parameters, showing decreasing, constant, and increasing hazard behaviour.
Key formulas
For x≥0:
Probability density function:
Cumulative distribution function (probability of failure by time x):
Cumulative Failure Probability of a Bearing Component (Weibull Model)
Survival (reliability) function (probability of surviving past x):
Role of the shape parameter
k value | Hazard rate behaviour |
|---|---|
k<1 | Decreasing over time (early/"infant mortality" failures) |
k=1 | Constant (reduces exactly to the exponential distribution) |
k>1 | Increasing over time (wear-out failures) |
The scale parameter λ sets the "typical" size of x: plugging x=λ into the CDF always gives F(λ)=1−e−1≈0.632, regardless of k.
Worked example
A batch of light bulbs has a lifetime modelled by a Weibull distribution with k=1.5 and λ=800 hours. Find the probability a bulb lasts more than 1000 hours.
R(1000)=e−(1000/800)1.5=e−(1.25)1.5=e−1.397≈0.247. About 24.7% of bulbs survive past 1000 hours.
The figure above shows three Weibull probability density functions, all with scale parameter λ=1 but different shape parameters k. Which curve corresponds to a Weibull distribution that reduces exactly to the exponential distribution?
Weibull probability density functions with scale λ=1 and three different shape parameters, showing decreasing, constant, and increasing hazard behaviour.
Cumulative Failure Probability of a Bearing Component (Weibull Model)
The lifetime (in hours) of a certain type of electronic sensor follows a Weibull distribution with shape parameter k=2 and scale parameter λ=5000 hours. What is the probability that a randomly selected sensor operates for more than 6000 hours before failing?