Reliability Calculator
The chance that a unit runs a set time without a failure, from an exponential model (MTBF) or a Weibull model (shape β, scale η). Gives the probability of failure, B10 life, hazard rate and mean life. Updates as you type.
This works from a life model for one item. To get MTBF from a failure log, use the MTBF and MTTR calculator; for several units working together, the system availability calculator.
Your unit
Pick a life model, then enter its numbers and the mission time. The defaults are the worked example below.
Reliability curve R(t)
If it has already run
Conditional reliability = R(T + t) ÷ R(T).
Want a second look at this result?
Leave your work email if you would like the OEE Lab team to get in touch about your numbers. To keep the result, use Copy link or Print above.
No newsletter, no spam.
See every stop as minutes of OEE lost.
A life model is only as good as the failure times behind it. Fabrico reads every stop from the PLC or IoT sensors and shows it as minutes of OEE lost per line and shift, so MTBF and MTTR build up per machine from real stops, and where cameras are fitted the video of any stop can be streamed on request next to its machine data.
Request a demo Contact usWhat the reliability calculator does
It gives the chance that one unit runs a set time, the mission time, without a failure. A unit can be a bearing, a motor, a pump or a machine treated as one item.
You pick a life model. Exponential assumes a constant failure rate and needs only MTBF or the failure rate λ; Weibull adds a shape β, so the risk can rise or fall with age.
The tool returns reliability R(t), the probability of failure F(t), the hazard at that age, B1, B10 and B50 life, mean life, and the chance that a unit which has already run T hours survives the next mission.
The formulas
These are the standard life distribution formulas, as set out in the NIST/SEMATECH e-Handbook of Statistical Methods, section 8.1.6.2 (Weibull), and in C. E. Ebeling, An Introduction to Reliability and Maintainability Engineering (McGraw-Hill, 1997). The handbook uses other letters for the same parameters.
R(t) = e^-((t - γ) ÷ η)^β
- Weibull: R(t) = e^(-((t - γ) ÷ η)^β) for t above γ, and 100% before it. F(t) = 1 - R(t).
- Exponential: R(t) = e^(-λt), with λ = 1 ÷ MTBF. It is the Weibull with β = 1 and γ = 0, so the hazard is λ at every age.
- Hazard: h(t) = (β ÷ η) × ((t - γ) ÷ η)^(β - 1), the failure rate at age t among units still running.
- B life: Bp = γ + η × (-ln(1 - p))^(1/β). B10 uses p = 0.10, B1 uses 0.01 and B50 uses 0.50.
- Mean life: γ + η × Γ(1 + 1/β), where Γ is the gamma function. The tool computes Γ with the Lanczos approximation; for the exponential, mean life is the MTBF.
- Already ran T hours: the chance of surviving t more is R(T + t) ÷ R(T).
Worked example: the default numbers
Example: a gearbox bearing follows a Weibull life with β = 2 and η = 1,000 running hours, and γ = 0. You want the chance it survives a 500 hour run.
| Step | Working | Result |
|---|---|---|
| (t - γ) ÷ η | 500 ÷ 1,000 | 0.5 |
| Raised to β | 0.5^2 | 0.25 |
| Reliability R(500) | e^-0.25 | 77.88% |
| Probability of failure | 1 - 0.7788 | 22.12% |
| Hazard at 500 h | (2 ÷ 1,000) × 0.5 | 0.001 /h |
| B10 life | 1,000 × 0.10536^(1/2) | 324.6 h |
| B50 life | 1,000 × 0.69315^(1/2) | 832.6 h |
| Mean life | 1,000 × Γ(1.5) | 886.2 h |
So the run has a 77.88% chance of finishing without a failure. B1 life, the age by which 1 in 100 fails, is 100.3 hours.
Now say the bearing has already run 500 hours. The chance it survives the next 500 is R(1,000) ÷ R(500) = e^-1 ÷ e^-0.25 = e^-0.75 = 47.24%, down from 77.88% when new.
Compare an exponential model with the same mean life, an MTBF of 886.2 hours. It gives R(500) = e^(-500 ÷ 886.2) = 56.88%, at any age. Same average life, a different answer, so the choice of model matters.
What β tells you
- β below 1: early-life failures. The risk per hour falls with age, as weak units fail first. Look at build quality, installation and start-up.
- β near 1: random failures. The risk per hour barely changes with age, so replacing a part on a timer does not lower it.
- β above 1: wear-out. The risk rises with age, so a planned replacement before the steep part of the curve can pay off. The preventive maintenance calculator puts a number on that.
Over a whole life, these three phases are often drawn as the bathtub curve. This tool models one phase at a time, with one β.
How to read the result
- R(t) and F(t) are probabilities for one unit, or shares of a large group of identical units. They always add up to 100%.
- Hazard at t is the rate at that age for units still running, not the chance over the whole mission.
- B10 is a common way to state a planned life for parts such as bearings. Plan spares from the failures you expect; the spare parts inventory calculator covers the stock side.
- The chart runs from 0 to about three times your mission. The steeper the drop after the dashed line, the faster risk builds with each extra hour. To push the curve to the right on the floor, see how to improve MTBF.
Common mistakes
- Using MTBF as the time before the first failure. With a constant failure rate, only e^-1 = 36.8% of units reach the MTBF without a failure, so 63.2% fail sooner. What is a good MTBF explains what the figure can and cannot tell you.
- Mixing calendar hours and running hours. If η or MTBF comes from running hours, the mission time must be running hours too. A machine on one 8 hour shift runs a third of the calendar day.
- Fitting β from too few failures. A handful of failure times can give a β far from the true value, and units that have not failed yet must count as survivors in the fit. Treat a β from a small sample as a rough guide.
- Giving a whole machine one Weibull. A machine with many parts that fail in different ways rarely follows one clean curve. Model the critical parts one by one; reliability centered maintenance is a method for picking them.
How do you calculate reliability from MTBF?
With a constant failure rate, R(t) = e^(-t ÷ MTBF). An MTBF of 1,000 hours and a 100 hour mission give e^(-0.1) = 90.48%, so 9.52% of units are expected to fail in that time.
What is B10 life?
It is the age by which 10% of a population is expected to have failed. For a Weibull with β = 2 and η = 1,000 hours, B10 = 1,000 × (-ln 0.9)^(1/2) = 324.6 hours.
What does the Weibull shape parameter β tell you?
Below 1, the risk per hour falls with age, which points to early-life failures. Near 1, failures are random in time. Above 1, the risk rises with age, which points to wear-out.
What is the Weibull scale parameter η?
It is the characteristic life: the age by which 63.2% of units have failed, whatever the value of β, when the location γ is 0. It is not the mean life unless β is 1.
Does the chance of a failure depend on how long a unit has run?
Only if the failure rate changes with age. With a constant rate it does not, which is called the memoryless property. With β above 1, the chance of surviving the next mission falls as the unit gets older.
Is reliability the same as availability?
No. Reliability is the chance of running a set time with no failure. Availability is the share of time a unit is able to run once repairs are counted, which the system availability calculator works out.