What is Weibull analysis?

Weibull analysis is the tool of choice for most reliability engineers when they consider what to do with age-to-failure data. It uses the Weibull distribution which says mathematically that reliability, R(t) = e-(t/h)^b where t is time, h is a scale factor know as the characteristic life (most of the Weibull distributions have tailed data and lack an easy way to describe central tendency as the mode≠median≠mean, however, regardless of the b-values, which is a shape factor, and all of the cumulative distribution function values pass through the h value at 63.2% which thus entitles it to be know as the single point characteristic life).

Why use Weibull analysis?

The Weibull distribution is so frequently used for reliability analysis because one set of math (based on the weakest link in the chain will cause failure) described infant mortality, chance failures, and wear-out failures.

When to use Weibull analysis?

Use Weibull analysis when you have age-to-failure data. When you have age-to-failure data by component, the analysis is very helpful because the b-values will tell you the modes of failure which no other distribution will do this! When you have age-to-failure by system, the b-values have NO physical significance and the b-, h-values only explain how the system is functioning-this means you loose significant information for problem solving.

Where to use Weibull analysis?

When in doubt, use the Weibull distribution to analyze age-to-failure data. It works with test data. It works with field data. It works with warranty data. It works with accelerated testing data. The Weibull distribution is valid for ~85-95% of all life data, so play the odds and start with Weibull analysis. The major competing distribution for Weibull analysis is the lognormal distribution. For additional information read The New Weibull Handbook, 5th edition by Dr. Robert B. Abernethy and use the WinSMITH Weibull and WinSMITH Visual software for analyzing the data (both software are bundled for a reduce price as SuperSMITH).

These definitions are written by H. Paul Barringer

Return to Reliability Tools