When a manufacturer rates how long a bearing, motor, or wear-prone component lasts, they rarely quote the average life, because the average hides how many units fail well before it. Instead they quote a low percentile, the age by which a small, defined fraction of the population has failed, and the most common of these is the B10 life. It answers a question a designer actually cares about: how long can I run before an unacceptable share of units has worn out? This page defines B10 as the age at which ten percent of a population has failed, shows how to read it off a Weibull fit, and explains why a low percentile is chosen over the mean for rating rotating and wear-prone field equipment.
B10 life in one line: The B10 life is the age at which ten percent of a population is expected to have failed, meaning ninety percent are still running, and it is a specific percentile of the failure-time distribution rather than an average. You calculate it by fitting a life distribution, most often a Weibull, to failure data including any survivors, then reading off the time at which the cumulative failure probability reaches ten percent. Designers of bearings and other wear-prone equipment prefer this low percentile to the mean because it describes early failures, which is what a conservative design life needs to avoid.
B10 is a percentile of a life distribution, not a typical or average value. Its definition is direct: it is the age at which ten percent of the population has failed and, equivalently, ninety percent has survived. The B stands for the family of Bx or percentile lives, where the number after it is the failure percentage of interest, so B5 is the age at five percent failed and B50 is the age at which half the population has failed. In rolling-element bearing work the same quantity is traditionally called the L10 life, and B10 and L10 refer to the same ten-percent-failed age; the different letters are just conventions from different fields.
Because it is defined as a fraction of the population failed, B10 is a property of the whole distribution of lives, not of any single unit. It does not say that a particular bearing will last exactly the B10 time; it says that if you ran many identical bearings, you would expect around ten percent of them to have failed by that age. This population framing is what makes it useful for design and warranty decisions, where you care about the behavior of a fleet of parts rather than the fate of one. It also means B10 is meaningful only alongside the distribution it comes from, since the same ten-percent point sits at very different ages for a tightly clustered wear-out and a broadly spread random failure pattern.
The choice of ten percent as the reference fraction is a convention balancing conservatism and practicality. A lower percentile such as B1 would describe an even earlier, rarer failure and give a very short, very cautious life, while a higher one such as B50 approaches the median and tolerates far more failures. Ten percent has become a widely used standard for rating rotating and wear-prone equipment because it captures early failures without being so extreme as to be impractical, and because it aligns with long-established bearing-rating practice, which makes B10 and L10 figures broadly comparable across the industry.
Calculating B10 starts with fitting a life distribution to failure data, and for wear-prone parts the Weibull distribution is the usual choice because its shape can represent the increasing failure rate that wear-out produces. The fit uses the recorded failure ages together with the ages of any units still running, the censored survivors, which must be included so the estimate is not biased toward short lives. Fitting yields the Weibull parameters that describe the whole life distribution, from which any percentile, including the ten-percent point, can be computed.
With the fitted distribution in hand, B10 is read as the age where the cumulative probability of failure first reaches ten percent. On a Weibull probability plot, where the fitted line represents cumulative failure probability against age, you find the ten-percent level on the probability axis and read across to the fitted line and down to the corresponding age; that age is the B10 life. Reliability software does the same thing numerically, returning the time at which the fitted distribution predicts ten percent failed. Either way the B10 is a direct consequence of the fit, so it is only as trustworthy as the data and the distribution behind it.
Two things determine where B10 lands relative to other summary numbers. The scale of the distribution sets the overall timeframe, while the shape sets how spread out the failures are, and a more spread-out distribution pushes B10 much earlier than the mean, because the early tail of failures arrives well before the average. This gap between B10 and the mean is exactly why the two are not interchangeable, and it is largest for populations with wide variability. Getting a defensible B10 therefore depends on a good fit to enough data, including survivors, rather than on plugging a single number into a formula.
The reason designers rate wear-prone equipment on B10 rather than the mean life comes down to what each number protects against. The mean life is pulled upward by the long-lasting units and says nothing about how many parts fail early, so designing to the average would leave a substantial fraction of a fleet failing before its rated life, which for rotating equipment in the field means unplanned downtime and callouts. B10 instead pins the rating to the early-failure end of the distribution, guaranteeing that by the rated life only a defined small fraction has failed, which is the conservative, fleet-protecting basis a design life needs.
Producing a credible B10 depends on real operating data, and this is where monitoring earns its place. A trustworthy Weibull fit needs the failure ages of the units that failed and, just as importantly, the accumulated ages of the survivors that have not, so that censoring is handled and the estimate is not biased low. A control or SCADA system supplies both by tracking runtime hours for every unit and logging when failures occur. A cloud SCADA platform such as Merobix can gather this runtime and failure history across a whole fleet of similar rotating assets and dispersed sites, giving the population of failures and survivors that a percentile estimate like B10 requires to be meaningful.
Beyond computing a static B10, ongoing monitoring lets the estimate stay honest and turns it into operational action. As more units accumulate hours and more failures occur, the fit and its B10 can be updated to reflect how the equipment is actually wearing in your duty and environment, which may differ from the manufacturer's rating. And because B10 marks the age by which a meaningful fraction is expected to have failed, tracking each unit's runtime against that figure in the monitoring platform lets maintenance flag assets approaching their B10 for inspection or replacement before they contribute to the failing ten percent, converting a population statistic into a per-asset maintenance trigger.
Yes, B10 and L10 refer to the same quantity, the age at which ten percent of a population is expected to have failed and ninety percent survives. The L10 term comes from rolling-element bearing practice while B10 is the more general reliability term, but both name the ten-percent-failed percentile of the life distribution. The different letters are simply conventions from different fields describing the identical concept.
The average life is pulled up by long-lasting units and says nothing about how many parts fail early, so designing to it would leave a large fraction of a fleet failing before the rated life. B10 pins the rating to the early-failure end of the distribution, guaranteeing only a small defined fraction has failed by the rated age, which is the conservative basis a design life needs. For rotating and wear-prone field equipment, avoiding early failures matters more than describing the average, which is why the low percentile is preferred.
Yes, a sound B10 requires including the ages of units still running, the censored survivors, alongside the failure ages, because leaving survivors out biases the fit toward shorter lives and understates B10. The Weibull fit should be built from both the failures and the suspensions so it reflects the whole population. A SCADA or maintenance system that records runtime for every unit and flags which have failed provides exactly the failure-and-survivor data the calculation needs.
Merobix reads your field devices into a cloud SCADA - the real thing behind these terms, live in days from any browser.