Definition
MTBF stands for mean time between failures. It applies to repairable systems: it is the total operating time divided by the number of failures that occurred. For non-repairable products the term is MTTF, mean time to failure.
Relationship to failure rate
If the failure rate is constant, which corresponds to the flat section of the bathtub curve, the MTBF is the inverse of that rate. A higher MTBF means a lower frequency of failures per unit of operating time. The relationship holds only at a constant rate; if the rate changes with age, the average hides that variation.
The most common misunderstanding
A long MTBF does not mean that each unit lasts that long. At a constant rate, the probability that a unit survives to a time exactly equal to the MTBF is only a fraction, about one third. Many units fail earlier. MTBF describes a population in operation, not the life of one unit. Nor does it include wear-out: a product can have a very high MTBF during its useful life and degrade at the end through wear.
How it is obtained
- Prediction: from models based on component failure rates. It is an estimate that depends on the assumptions.
- Testing: operating many units for a long time and counting failures. It is reported with a confidence level.
- Field data: actual failures of installed units, which are the most representative source.
The methods yield different values, and comparing products requires the method and conditions to be the same.
How to interpret it
It is useful for comparing designs, planning spare parts and calculating availability, together with the mean time to repair. It is a statistical indicator of a population with a defined use profile. A value without conditions, such as temperature and stress, or without the method used to obtain it, has no meaning.
When it does not apply
It does not describe useful life or individual safety, and it is not suitable when wear-out mechanisms dominate. For those cases it is better to use life distributions such as Weibull.