MTBF

What is MTBF?

MTBF (Mean Time Between Failures) is the average operating time between two consecutive failures of a repairable system, device or component, calculated as the total operating time divided by the number of failures in that period. MTBF is one of the most important KPIs for reliability in maintenance, asset management and the design of industrial installations. The higher the MTBF, the less often an installation fails on average. Together with MTTR, MTBF determines the availability of a machine or system.

🧮 How is MTBF calculated?

The basic formula is simple:

MTBF = total operating time / number of failures

A machine that runs for 1,000 hours and fails 4 times in that period has an MTBF of 1,000 / 4 = 250 hours.

In practice you usually calculate over a population of identical assets. Suppose 20 identical pumps each run for 8,000 operating hours and 8 failures occur in total. The MTBF is then (20 × 8,000) / 8 = 20,000 hours. Pumps that have not yet failed also count towards the operating time; if you only include the units that failed, you will badly underestimate the MTBF.

A few ground rules apply:

• Operating time, not calendar time - count only the hours the system is actually running; downtime for planned maintenance or because there is no production does not count
• A fixed failure definition - decide what counts as a failure (loss of function, unplanned stop, alarm requiring intervention) and keep that definition constant, otherwise figures cannot be compared across periods and sites
• Repairable systems only - for parts that are replaced when they fail, such as a sensor or a fuse, use MTTF (Mean Time To Failure)
• Failure rate - with a constant failure rate λ, MTBF = 1/λ; an MTBF of 20,000 hours corresponds to λ = 5 × 10⁻⁵ failures per hour

📉 What does MTBF tell you about service life, and what not?

The biggest misconception is that MTBF is the expected service life of a single device. It is not. MTBF is an average across a population during the period in which the failure rate is roughly constant: the flat middle part of the so-called bathtub curve, between infant mortality (early failures) and wear-out at the end of life.

With a constant failure rate, the probability that a device survives time t without failure is R(t) = e^(−t/MTBF). It follows that at t = MTBF only about 37% of devices are still running without a failure. More than 63% have already failed at least once.

A useful rule of thumb is the annualised failure rate (AFR): with continuous operation roughly 8,760 / MTBF. A hard disk or power supply with a stated MTBF of 1,000,000 hours will therefore fail in about 0.9% of units per year. That does not mean a single unit will last 114 years.

🔄 What is the difference between MTBF, MTTF, MTTR and availability?

Inherent availability is MTBF / (MTBF + MTTR). A pump with an MTBF of 20,000 hours and an MTTR of 10 hours has an availability of 20,000 / 20,010 = 99.95%. Note that this formula only uses the technical repair time. Waiting for a technician or a spare part is excluded, so availability in practice turns out lower.

Some sources define MTBF as MTTF + MTTR, i.e. including the repair time. Because repair time is usually small compared with operating time, the difference is often negligible, but state which definition was used when you compare figures.

🧱 How do you determine the MTBF of a system made of components?

• Series configuration - when every component is needed, the failure rates add up: 1/MTBF(system) = Σ 1/MTBF(component). A PLC CPU of 250,000 hours, an I/O module of 500,000 hours and a power supply of 200,000 hours together give an MTBF of about 90,900 hours, lower than that of the weakest link.
• Parallel redundancy - two identical units of which one is sufficient (1oo2) greatly increase reliability. Without repair, the mean time to system failure rises to 1.5 × the MTBF of a single unit; with prompt repair of the failed unit the gain is far larger. That is the principle behind redundant DCS controllers, UPS systems and high availability clusters.
• Prediction standards - for electronics without field data, manufacturers calculate MTBF from part failure rates using MIL-HDBK-217F (last revised by Notice 2 in 1995), Telcordia SR-332, the Siemens standard SN 29500 or the French FIDES method. IEC 61709 (edition 3, 2017) provides reference conditions and stress models to convert failure rates to other temperatures and loads. The same circuit can yield very different MTBF values depending on the method, so only compare datasheets that use the same method and conditions.

🏭 How do you use MTBF in maintenance and safety?

In practice the MTBF comes from your own failure history in a CMMS or from the historian of the SCADA system. A five-step approach:

1. Define assets and failures - choose a fixed asset hierarchy and failure codes. ISO 14224, originally written for the petroleum, petrochemical and natural gas industries, offers a practical taxonomy for collecting reliability and maintenance data.
2. Record operating time - pull running hours automatically from the PLC or DCS instead of estimating them.
3. Calculate MTBF per asset type - for example quarterly over a rolling year, so that a single incident does not dominate the trend.
4. Analyse the worst performers - use root cause analysis or FMEA to eliminate recurring failures.
5. Adjust your strategy - choose between preventive maintenance, condition monitoring or predictive maintenance and track the effect in the MTBF trend.

MTBF is also a building block for OEE, TPM and RAMSHEEP analyses, which are widely used in Dutch infrastructure projects, for example by Rijkswaterstaat. Within ISO 55000, the MTBF trend supports decisions on replacement versus overhaul.

Functional safety works differently. Calculations under IEC 61508 and IEC 61511 for a SIL level do not use MTBF itself, but failure rates per failure mode, mainly the dangerous undetected failures (λDU), from which the PFDavg or PFH follows. In machinery safety, ISO 13849-1 uses MTTFd: low (3 to 10 years), medium (10 to 30 years) or high (30 to 100 years). For pneumatic and electromechanical components it is derived from the B10d value: MTTFd = B10d / (0.1 × number of operations per year).

🧮 How do you calculate availability from MTBF and MTTR?

Enter MTBF and MTTR and choose whether components are in series or redundant. The calculator shows the availability and the expected downtime per year.

❓ Frequently asked questions

What is a good MTBF value?

There is no universally good MTBF; the value depends heavily on the type of asset and the operating conditions. An industrial power supply may have a stated MTBF of hundreds of thousands of hours, while a mechanical pump often reaches thousands to tens of thousands of hours. Always compare MTBF within the same asset class, against your own history or against industry data.

Does an MTBF of 100,000 hours mean a device will last 11 years?

No, MTBF is not a service life but an average across a population during the period with a constant failure rate. With an MTBF of 100,000 hours and continuous operation, about 8.4% of devices fail each year, and the actual service life is determined by wear and ageing. MTBF mainly tells you how many failures to expect in a large population.

When do you use MTTF instead of MTBF?

You use MTTF for components that are replaced rather than repaired after a defect, such as lamps, fuses and many sensors. MTBF belongs to repairable systems that return to service after a failure, such as pumps, machines and control cabinets. The calculation is similar, but the distinction makes clear whether a failure is followed by a repair or a replacement.

How do you increase the MTBF of an installation?

MTBF rises when you eliminate the causes of recurring failures with root cause analysis and FMEA, keep components loaded within their specifications and use condition monitoring to detect wear in time. Better lubrication, alignment, cooling and clean power also raise MTBF noticeably. Redundancy mainly increases MTBF at system level, not that of the individual parts.

Can you trust the MTBF on a datasheet?

A datasheet MTBF is usually a calculated prediction according to a standard such as MIL-HDBK-217F, Telcordia SR-332 or SN 29500, at a fixed reference temperature. In a hot control cabinet or under heavy load, the actual MTBF is often lower. Ask the manufacturer which method and which ambient conditions were used.

How are MTBF and availability related?

Inherent availability is MTBF divided by the sum of MTBF and MTTR. A higher MTBF means failures occur less often, and a lower MTTR means a shorter repair per failure. Both increase the availability of an installation.

📌 In summary

MTBF measures the average operating time between failures of repairable systems and, together with MTTR, forms the basis of availability. It is a population average, not a guaranteed service life: at t = MTBF only about 37% of devices are still running without a failure. Reliable MTBF figures require a fixed failure definition, real operating hours and an understanding of the method behind datasheet values.

MTBF indicates how often a failure occurs on average; if you treat it as a population average rather than a service life, you plan maintenance, spare parts and redundancy on facts.