The three numbers and what they tell you
MTBF, mean time between failures, is how long the equipment runs, on average, from one breakdown to the next. It measures reliability: how often things go wrong.
MTTR, mean time to repair, is how long it takes, on average, to get it running again once it has failed. It measures maintainability: how quickly you recover. In practice it includes everything from the moment the machine stops to the moment it runs again: noticing the failure, finding someone, finding the part, the repair itself and the restart.
Availability combines the two into the share of time the equipment is ready to run:
A worked example
Three identical pumps run around the clock for a year. Between them they fail nine times, and the repairs take 54 hours in total. That is 26,280 unit-hours in service, of which 54 were spent under repair. MTBF comes out at about 2,900 hours, MTTR at 6 hours, and availability at 99.8%.
That sounds excellent, until you look at it the other way: the chance of any one pump running a full month without a failure is only about 78%. Availability hides how often you get disturbed. For a plant with no standby, the failure rate is often the more useful number.
Which one should you work on?
Availability can be improved in two ways, and they cost very different things. Raising MTBF means fewer failures: better lubrication, alignment, operating closer to the design point, replacing chronic bad actors. Lowering MTTR means faster recovery: spares on the shelf, clear repair procedures, tools and people ready.
A useful habit is to split MTTR into its parts in your failure log: time to respond, time waiting for parts, actual repair time. In many plants the repair itself is the smallest of the three. If waiting for parts dominates, the critical spare parts calculator will tell you which spares pay for themselves. If you are not recording repairs yet, the equipment log template has a failure log ready to use.
Assumptions behind the reliability figure
The probability of running a mission time without failure is calculated with the exponential model, R(t) = e−t / MTBF, which assumes failures happen at a constant random rate. That is a fair approximation for a mixed population of equipment in its useful life. It is not accurate for parts that wear out, such as bearings, seals and belts, whose failure rate rises with age, or for new equipment still in its early failure period. For those, a Weibull analysis on individual failure times gives a truer picture.
Frequently asked questions
Should MTBF use calendar time or operating time?
Operating time, if you can. A pump that runs only 8 hours a day accumulates a third of the running hours of one in continuous service, and its MTBF should reflect that. Enter the operating hours per day and the calculator converts the observation period into running hours.
What counts as a failure?
Any unplanned stop that needed a repair or intervention to restore the equipment. Planned maintenance does not count. Be consistent: if you count short resets in one period and not in another, the MTBF trend means nothing.
What is a good availability?
It depends on the equipment and the consequences of a stop. Critical continuous equipment is often expected to exceed 98 to 99 percent. But compare like with like, and look at the trend over time rather than a single figure.
Why is my MTBF longer than the time I observed?
Because it is an average across units. Three pumps observed for one year give three unit-years of operating time. With only one failure, the MTBF would be about three years even though nobody watched a single pump for that long. With few failures, treat the figure as a rough estimate.
More free engineering tools
Free, no sign-up, built by an engineer who runs heating and water supply networks.