Iron Lake Reliability

Weibull is pronounced WYE-bull (hard I).

MTTF calculator

MTTF is the mean of the life distribution — the average time to failure if every unit is run to death. For two-parameter Weibull it is not characteristic life η and it is not B10. Use it for expected energy, spares at the mean, or comparison to an exponential baseline; use B-lives when you care about a failed fraction.

Open the Iron Lake Reliability workbook Open Weibull Calculator Compute mean life from a fit

Formula

MTTF = E[T] = η Γ(1 + 1/β)

Γ(z) is the gamma function. Γ(1) = 1, Γ(1.5) = √π / 2 ≈ 0.886227.

The exponential special case is β = 1: Γ(2) = 1, so MTTF = η. That is the only shape where “characteristic life” and “mean life” coincide. Iron Lake Reliability keeps both on the calculator so they cannot be confused.

Worked example (β = 2, η = 1 000 h)

Shape 2 is a clean wear-out Weibull (Rayleigh). Then 1 + 1/β = 1.5 and

MTTF = 1 000 × Γ(1.5) = 1 000 × √π / 2 ≈ 886 h

Do not swap mean, median, and scale
NameWhat it isβ = 2, η = 1 000 h
B1010% failed quantile325 h
Median B5050% failed quantile833 h
MTTFmean E[T]886 h
η63.2% failed (scale)1 000 h

The gamma value is the whole ratio

Γ(n) = (n − 1)! for positive integers, so Γ(1) = 1 and Γ(2) = 1. Between 1 and 2 it dips below 1. For a Weibull the argument is 1 + 1/β: it equals 2 at β = 1, equals 1.5 at β = 2, and slides back toward 1 as β grows. MTTF / η is exactly that gamma value. The table holds η at 1 000 h.

Mean life as a fraction of characteristic life
β1 + 1/βΓ(1 + 1/β)MTTF
0.532.0002 000 h
0.82.251.1331 133 h
1211 000 h
1.51.66670.9027902.7 h
21.50.8862886.2 h
31.33330.8930893.0 h
41.250.9064906.4 h

For every wear-out row (β > 1) the mean is shorter than η, and the shortfall is not a straight line in β. The ratio is 0.886 at β = 2, 0.893 at β = 3, and 0.906 at β = 4. A very large shape concentrates the failures near η, so the mean climbs back toward the scale. For infant mortality the mean is longer than η, because of a long right tail. β = 0.5 gives MTTF = 2 000 h from Γ(3) = 2, and the B10 on that same scale is only 11.1 h. The average of the tail does not describe the early failures.

Hand check for β = 2

Shape 2 is the Rayleigh case, and Γ(1.5) has a closed form. Γ(1/2) = √π and Γ(z + 1) = z Γ(z), so Γ(1.5) = √π / 2 = 0.886226925. Times 1 000 h that is 886.227 h, the 886 h in the summary table. The median on the same model is 832.6 h, so the mean sits a little to the right of the median. Neither is B10 (324.6 h) and neither is η.

1 / MTTF = 1 / 886.227 h = 0.001128 / h. People sometimes call that “the failure rate.” It equals h(t) at every age only when β = 1. On this model the hazard is 0.0005 / h at 250 h, 0.001 / h at 500 h, and 0.002 / h at 1 000 h. One of those is near 1/MTTF. The others are not.

When to use the mean

Use MTTF for an average over the whole population: expected hours if every unit runs to failure, or a comparison with an exponential baseline that was quoted as a mean. Do not use it as a warranty life, a maintenance interval, or a demonstration target. Those are quantiles or test designs. B10 answers “when have 10% failed?” The Weibull analysis page is where β and η come from.

MTTF is for items that fail once. MTBF is the mean time between failures of a repairable item, a different process. A constant-rate parts-count model is the exponential case, β = 1, so there the mean, the scale, and 1/λ do coincide. As soon as the life data say β is not 1, that coincidence is over.

Common mistakes

Open Weibull Calculator → Mean life after a workbook fit. The bearing-sample MLE (β ≈ 2.17, η ≈ 716 h) has MTTF ≈ 634 h — again below η, as wear-out requires. Related: B10 life and hazard rate.

FAQ

What is MTTF for a Weibull distribution?

Mean time to failure is the expected life: MTTF = η Γ(1 + 1/β), where Γ is the gamma function. It is the mean of the distribution, not a reliability quantile such as B10.

When does MTTF equal η?

Only when β = 1 (exponential). Then MTTF = η = 1/λ and the hazard is constant. For wear-out (β > 1) MTTF is less than η; for infant mortality (β < 1) MTTF is greater than η.

Is MTTF the same as MTBF?

MTTF is for non-repairable items (time to first failure). MTBF is for repairable systems (mean time between failures). Iron Lake Reliability is a life-data workbook for units that fail and are suspended — it reports mean life MTTF from the fitted LifeFit, not a repairable-process MTBF.

Where is mean life in Iron Lake Reliability?

Fit the grid on the Weibull Workbook, then Weibull Calculator → Life → Mean life. Mean remaining life at age T is also there; at T = 0 it matches MTTF.

Open the Iron Lake Reliability workbook Open Weibull Calculator