Weibull analysis is a statistical method that fits your equipment failure data to a Weibull distribution to reveal the pattern behind those failures: whether parts are dying young, failing at random, or wearing out with age. From that fit you read two numbers, the shape parameter (beta) and the scale parameter (eta), and those two numbers tell you which maintenance strategy actually fits the machine in front of you. It is the quantitative backbone of reliability engineering, and it turns a pile of breakdown dates into a defensible plan.
Every failure record has a life value: hours run, cycles completed, or days in service before the asset broke. Weibull analysis takes a set of those life values and finds the distribution curve that best describes them. Because the Weibull distribution is so flexible, one equation can model failures that speed up over time, stay constant, or slow down. That flexibility is why reliability engineers reach for it before almost any other tool. The output is not a vague opinion about reliability; it is a fitted curve you can use to estimate the probability of failure at any future point in a component's life.
The shape parameter, beta, is the headline result. It tells you the direction of the failure rate over time, and it maps directly onto the three zones of the bathtub curve:
Knowing which zone you are in changes everything. You do not schedule time-based replacements on a part with beta below 1, because a fresh part is statistically more likely to fail than the aging one you would remove.
The scale parameter, eta, is the characteristic life. It is the point by which 63.2 percent of the population will have failed, regardless of the beta value. Think of it as the natural time scale of the failure process. A pump family with eta of 8,000 hours has a longer characteristic life than one with eta of 3,000 hours. Together, beta tells you the shape of failure and eta tells you the timing, and you need both to plan.
Suppose you log bearing failures on a conveyor and a Weibull fit returns beta = 2.5 and eta = 6,000 hours. Beta of 2.5 is comfortably above 1, so these bearings are in a wear-out pattern. Eta of 6,000 hours means 63.2 percent will have failed by 6,000 hours of run time.
The Weibull reliability formula is R(t) = e^(-(t/eta)^beta). Ask: what fraction survive to 3,000 hours, half of the characteristic life?
So about 83.8 percent of bearings survive to 3,000 hours, meaning roughly 16 percent have already failed by the halfway mark. Now check 4,500 hours: (4,500/6,000) = 0.75, and 0.75^2.5 = 0.487, so R = e^(-0.487) = 0.614. Reliability drops from 83.8 percent to 61.4 percent as you move from 3,000 to 4,500 hours. That steepening decline is the wear-out signature in action.
The maintenance decision: because beta is well above 1, a time-based replacement is justified. A sensible planned-replacement interval sits below eta where reliability is still high, perhaps around 3,000 to 3,500 hours, rather than running bearings to failure. If beta had come back near 1, scheduled replacement would waste good parts and you would lean on condition-based maintenance instead, watching vibration and temperature to catch the random failure before it cascades.
Weibull results translate cleanly into strategy. Beta below 1 points to fixing your installation and commissioning process, not to a replacement schedule; the problem is quality and break-in, so a burn-in test protects you. Beta near 1 favors condition monitoring and inspection because age gives you no warning. Beta above 1 rewards planned, time-based interventions and is the classic case for proactive maintenance over waiting for the break. Feeding these findings into a structured FMEA sharpens which failure modes deserve the most attention and cost.
Weibull analysis complements the summary numbers you already track. A single MTBF figure averages away the failure pattern; two machines can share an identical MTBF while one suffers infant mortality and the other pure wear-out, demanding opposite responses. Weibull recovers that lost detail. When beta equals 1, mean time between failures and characteristic life align neatly, but for any other beta they diverge, and the divergence is exactly the insight a raw average hides.
Weibull analysis is only as good as the failure data you feed it, and this is where most programs quietly fail. Guessed dates, missing run hours, and uncounted micro-stops produce a garbage fit no matter how good the statistics. Fabrico does not run Weibull software or predictive models for you; instead it supplies the clean, timestamped foundation those methods depend on. Its real-time OEE and production monitoring captures precise run time and stoppage events, including on machines without a PLC through camera and computer-vision monitoring, so every failure has a true life value attached. Its CMMS logs work orders, asset histories, and spare-part usage, giving you the failure timeline and repair records a Weibull fit requires. Feed that verified CMMS history into your analysis and your beta and eta reflect reality, not a spreadsheet of estimates. Cleaner input also improves your broader OEE picture.
More is better, but a rough rule is that even 5 to 10 failure points can suggest the beta direction, while 15 to 20 or more give confidence in both beta and eta. Fewer points widen the uncertainty, so treat a small-sample beta as a hint rather than a hard number, and keep collecting accurate life data as failures accumulate.
No. Fabrico is a real-time OEE monitoring and CMMS platform, not statistical software, so it does not fit Weibull curves or run reliability simulations for you. What it does is capture the accurate run-time, stoppage, and repair data that a Weibull analysis in your chosen tool relies on, so your fitted parameters describe what actually happened on the floor.
Eta is the characteristic life, the point by which 63.2 percent of units have failed, while MTBF is the average time between failures. They are equal only when beta equals 1. For wear-out patterns (beta above 1), MTBF is typically a bit below eta, so mixing the two up will mis-time your replacements.
Want your Weibull inputs built on real data instead of guesswork? Book a Fabrico demo to see how real-time monitoring and CMMS give reliability engineering an accurate foundation.
Login
Blog
