Mean time between failures, plus MTTR, availability, and mission reliability
| MTBF (hrs) | MTTR (hrs) | Ratio | Availability |
|---|---|---|---|
| 100 | 5 | 20:1 | 95.2% |
| 120 | 3 | 40:1 | 97.6% |
| 200 | 10 | 20:1 | 95.2% |
| 250 | 5 | 50:1 | 98.0% |
| 500 | 8 | 63:1 | 98.4% |
| 1,000 | 10 | 100:1 | 99.0% |
| 2,000 | 20 | 100:1 | 99.0% |
Notice the two 95.2% rows and the two 99.0% rows: availability depends on the ratio, not the raw MTBF. A plant that doubles MTBF but lets repairs stretch twice as long gains nothing.
MTBF answers one question: how long does this asset run, on average, between breakdowns? It's the core metric of reliability engineering for repairable systems, and it feeds directly into availability targets, spare-parts decisions, and maintenance planning.
MTBF = operating hours ÷ failures. MTTR = total repair time ÷ repairs. Availability = MTBF ÷ (MTBF + MTTR). Failure rate is the reciprocal of MTBF, shown here per 1,000 hours so the numbers stay readable. Mission reliability uses the exponential model: R(t) = e^(−t/MTBF), the probability the asset survives t hours of operation without a failure.
Pull a period from your maintenance log or CMMS: the hours the asset was scheduled and running, how many unplanned failures stopped it, and the wrench time spent getting it back. Keep one definition of "failure" and stick with it. Most teams count any unplanned stop needing a work order and exclude scheduled maintenance.
A packaging line logs 720 operating hours in a quarter and 6 unplanned failures. MTBF = 720 ÷ 6 = 120 hours. The 6 repairs consumed 18 hours total, so MTTR = 3 hours, and availability = 120 ÷ 123 = 97.6%. The failure rate works out to 8.33 failures per 1,000 operating hours, which at around-the-clock operation is roughly 73 per year (8,760 ÷ 120).
For the mission estimate: the chance of getting through a 24-hour production run without a stop is e^(−24/120) = 81.9%. Stretch the run to a full week, 168 hours, and it drops to 24.7%, which matches what the crew sees: week-long clean runs are rare, not normal. A second asset, a process pump with 3 failures across 8,760 operating hours, works out to MTBF = 2,920 hours and 98.7% availability at 38 total repair hours; the odds of a 30-day (720-hour) clean run are 78.1%.
There is no universal good number, because MTBF scales with how often an asset runs and what you count as a failure. A packaging line at 120 hours and a pump at 2,900 hours can both be healthy. The honest target is the MTBF-to-MTTR ratio, which sets availability: 120 hours between failures with 3-hour repairs gives 97.6% availability, while 200 hours with 10-hour repairs gives 95.2%. Track your own baseline, then drive the ratio up.
No, and this is the most common misreading. MTBF is the average interval between failures of a repairable asset, not how long it lasts. An asset with an MTBF of 120 hours can run for years if each failure is fixed in an hour. Lifetime is a retirement question; MTBF is a breakdown-frequency question.
MTBF measures how often something fails: operating time divided by the number of failures. MTTR measures how long each failure lasts: total repair time divided by the number of repairs. Availability combines the two: MTBF ÷ (MTBF + MTTR). Cutting MTTR is often cheaper than stretching MTBF, which is why spares and repair procedures matter as much as the hardware.
The standard formulas assume a constant failure rate, which only holds during an asset's useful-life period, after infant mortality but before wear-out. In that window, reliability over a mission of t hours follows R(t) = e^(−t/MTBF). If your failures cluster early (bad installs) or late (wear-out), the exponential model understates or overstates the risk, so treat the percentages as estimates, not guarantees.
Add up the hours the asset was actually running, count the failures that stopped it, and divide. Decide up front what counts as a failure; most teams exclude planned maintenance and include any unplanned stop that required a work order. Keep the definition consistent, because a metric that changes definitions every quarter can't show a trend.