Mean time to failure (MTTF): the lifespan of the parts you replace instead of repair

The instrumentation technician is replacing a batch of identical proximity sensors across the plant, and this time he logs each one: the day it went in, the day it stopped working. A year later he has ten little lifespans written down. Some sensors lasted barely three months, one soldiered past twelve, the rest landed somewhere in between.

On their own, each failure looked like bad luck. Averaged together, those ten lifespans become something far more useful: the number he needs to set a replacement interval, stock the right quantity of spares, and stop being ambushed at midnight.

What it actually measures

Mean time to failure is the average operating life a non-repairable component manages before it fails. It is the metric for the things you replace rather than repair: bearings, sensors, fuses, light fittings, electronic modules. For equipment you fix and return to service, the sibling metric is mean time between failures, and the two are sometimes both called mean life. The clock is usually counted in operating hours, but it can just as easily be cycles, distance or volume, whatever best matches the way the part actually wears.

How to work it out

MTTF = Total operating time to failure of the sample (hours) / Number of items run to failure

Ten identical sensors are run until they fail, at 100, 152, 192, 297, 433, 485, 689, 757, 823 and 951 hours.

MTTF = 4,879 / 10 = 487.9 hours

The average sensor lasts roughly 488 hours, and that single number is the seed of a plan.

Set the interval from the distribution, not the average

Here is the trap, and almost everyone falls into it once: you cannot safely set a replacement interval at the average. Look back at those ten sensors. They did not cluster politely around 488 hours; they sprawled from under 100 to past 950. Replace them all at 488 and you will be throwing away a good many that had hundreds of hours left, while others will already have failed and taken the line down with them. The mean describes a life that few of the actual parts lived.

The honest method is to set the interval from the spread, not the centre. You look at how the failures are distributed and pick a point early enough that very few units have failed by the time you replace, accepting that you sacrifice some good life to avoid the failures that matter. How early depends on the stakes: a safety-critical part is replaced while the probability of failure is still tiny, while for a cheap, harmless component you let it run longer and trade a few failures against the cost of replacing too often. Duty bends the curve, too. The very same part lives a comfortable life handling clean water and a brutal one handling abrasive slurry, so an MTTF borrowed from a gentler application will quietly betray you on a harsh one.

What good looks like

There is no universal benchmark; the right life depends on the component, the duty and the conditions it works in. What matters is the comparison: against the manufacturer's stated life, against the same part on another line, and against your own history. A rising MTTF means your components are lasting longer, whether through better installation, kinder operating conditions, or a better choice of part in the first place.

Where it can mislead you

• Use it only for non-repairable items that are replaced on failure. For anything you repair and reuse, mean time between failures is the right tool.
• It needs a population, not an anecdote. A single failure tells you almost nothing; a handful of like-for-like components gives you a number worth planning around, and the more you have, the more you can trust the shape behind it.
• The average hides the spread, and here the spread is the whole story. A 488-hour mean that contains a unit dead at 100 and another running past 950 is telling you to study the distribution, not to circle the average.
• A short MTTF against the design life is a flashing light. Treat it as the trigger for a closer look at the application, the installation or the specification, not as a fact of life to be quietly accepted.

A single part failing is a shrug. The same part failing on a schedule you can actually predict is an opportunity. MTTF, read as a distribution rather than a lone number, is what turns a drawer full of identical replacements into rational replacement intervals, the right spares on the shelf, and an end to the midnight callouts that come from being surprised by something you could have planned for.