MTBF, MTTR and Availability Explained

Reliability engineering asks how often equipment fails, how long it takes to fix, and how much of the time it is available. These questions affect safety, production output, maintenance planning and spare parts. MTBF, MTTR and availability are the most widely used metrics to answer them.

Definitions and formulas

MTBF for a repairable item over an observation period:

$$MTBF = \frac{\text{total operating time}}{\text{number of failures}}$$

MTTR (often called mean time to restore):

$$MTTR = \frac{\text{total downtime for repairs}}{\text{number of repairs}}$$

MTTR may include diagnosis, waiting for parts and staff, repair, testing and restart; definitions should be stated clearly.

Failure rate for a constant rate:

$$\lambda = \frac{1}{MTBF}$$

often expressed in failures per hour, per million hours, or FIT (failures per 10⁹ hours) for electronics.

Inherent availability:

$$A = \frac{MTBF}{MTBF + MTTR}$$

Operational availability also includes logistics delays, preventive maintenance downtime and administrative delays, so it is lower.

Reliability over time

For items with a constant failure rate (random failures), the probability of surviving a mission of length t without failure follows the exponential distribution:

$$R(t) = e^{-\lambda t} = e^{-t/MTBF}$$

At t = MTBF, R = e⁻¹ ≈ 0.37: only 37% of units are still working. This is why MTBF must not be confused with expected life.

The bathtub curve

Download diagram (WebP)

• Early life: failures from manufacturing defects and installation errors, reduced by burn-in, quality control and careful commissioning.
• Useful life: random failures at a roughly constant rate, where MTBF is meaningful.
• Wear-out: failures rise from fatigue, corrosion and wear; preventive replacement before this stage is often economic.

The Weibull distribution models all three regions with a shape parameter β (β below 1 for early failures, β = 1 for constant rate, β above 1 for wear-out) and is widely used to analyse real failure data.

System reliability

Series systems

If any one of n components failing causes system failure, failure rates add and reliabilities multiply:

$$\lambda_s = \sum \lambda_i \qquad R_s = \prod R_i$$

Parallel (redundant) systems

If the system works as long as at least one of two identical items works:

$$R_s = 1 - (1 - R)^2$$

System availability for two independent units in parallel is similarly A_s = 1 − (1 − A)². Redundancy is widely used in pumps, power supplies, controllers and safety systems, as discussed in safety integrity level.

How to collect MTBF and MTTR data

Good metrics need good data. Every failure should be recorded in the maintenance management system with the asset, date and time of failure, time of restoration, failure mode, cause and work done. Operating hours come from run hour meters, control system logs or production records. Teams must agree on what counts as a failure (for example, any unplanned loss of function, not minor adjustments) and whether MTTR includes waiting for parts and staff. Grouping similar assets, such as all pumps of one type, gives more failures and therefore more reliable estimates than a single machine.

MTBF, MTTR and maintainability

MTTR is a measure of maintainability: how quickly and easily equipment can be restored. It is influenced in design by accessibility, modularity, standard parts, built in diagnostics and clear labelling, and in operation by spare parts stocking, tools, skills and procedures. Reducing MTTR is often the fastest way to raise availability, because it does not require changing how often equipment fails.

Using MTBF and MTTR in practice

• Benchmarking and trend tracking: rising MTBF shows maintenance and reliability improvements; rising MTTR may reveal spare parts or skills problems.
• Maintenance strategy: choosing between run to failure, preventive and predictive types of maintenance.
• Spare parts planning: expected failures per year = operating hours ÷ MTBF.
• Design and procurement: comparing suppliers’ reliability data, while checking how it was derived.
• Safety analysis: failure rates feed FMEA, fault trees and PFD calculations.

Mean time between failures for software and systems

The same ideas apply to control systems, networks and IT services, where uptime targets such as 99.9% (about 8.8 hours of downtime per year) or 99.99% (about 53 minutes) are common. For such systems, MTTR often depends more on detection and diagnosis time than on repair work itself.

Common mistakes

The following errors appear often in reliability reports and lead to wrong decisions about maintenance and spares.

• Treating MTBF as product life.
• Mixing calendar time and operating time.
• Counting different failure definitions inconsistently.
• Using manufacturer MTBF predictions (often based on handbook models such as MIL-HDBK-217 or field data from benign conditions) without considering the actual operating environment.
• Ignoring wear-out, where the constant failure rate assumption no longer holds.

Used carefully, MTBF, MTTR and availability turn maintenance records into insight, helping engineers decide where reliability improvements will pay off most, often guided by root cause analysis of repeat failures.