MTBF and MTTR: How to Calculate Them from Machine Data

The monthly maintenance report says "filling machine failed four times". Production says the machine was down for almost a shift; maintenance says repairs took two hours. Both are right by their own records, and the meeting ends without a decision.

MTBF and MTTR are the two numbers that settle arguments like this: how often a machine fails, and how quickly it runs again. This guide covers the definitions, a worked example, the data you need, and how to get downtime records straight from machine data instead of handwritten logbooks.

What MTBF and MTTR mean

• MTBF (mean time between failures) is the average operating time between two failures. Higher means the machine fails less often.
• MTTR (mean time to repair) is the average time from a breakdown stop until the machine runs normally again. Lower means faster recovery.

The IEC 60050-192 dependability vocabulary defines MTBF as the mean operating time between failures, so time spent down for repair is not counted. MTBF also applies only to repairable items. For parts that are simply thrown away when they fail, the term is MTTF (mean time to failure).

The R in MTTR is read in different ways: repair, restore, or recovery. The label matters less than when the clock starts and stops, and that rule must be the same for every machine.

How to calculate MTBF and MTTR, with an example

• MTBF = total operating time ÷ number of failures
• MTTR = total repair time ÷ number of failures
• Availability = MTBF ÷ (MTBF + MTTR)

The example below is an illustration, not real plant data: one filling machine running three shifts for 30 days, with one planned maintenance stop.

Item	Value
Calendar time (30 × 24 hours)	720 hours
Planned stop (preventive maintenance)	8 hours
Planned production time	712 hours
Failures	4: 1.5 h, 3 h, 0.5 h, 2 h
Total repair time	7 hours
Operating time (712 - 7)	705 hours
MTBF (705 ÷ 4)	About 176 hours
MTTR (7 ÷ 4)	1.75 hours (1 h 45 min)
Availability (176.25 ÷ 178)	About 99.0%

Planned stops are left out. Preventive maintenance is not a failure, and mixing the two makes the numbers look worse than reality.

What counts as a failure

This is where figures from different lines stop being comparable. Agree the rules before anyone starts logging, for example:

• A failure is an unplanned stop because the machine cannot perform its function, lasting longer than a threshold you set, such as 5 minutes.
• Short stops below that threshold stay out of MTBF. They belong in the performance part of OEE monitoring.
• External stops, such as waiting for material, a power cut, or no orders, are not machine failures. Log them separately.
• The MTTR clock runs from the stop until normal production resumes, including waiting for a technician, diagnosis, waiting for parts, the repair itself, and the test run.

Splitting MTTR into those stages pays off. If most of the time goes on waiting for parts, the fix is in the store room, not in technician skills.

The data you need

For every stop, record:

• Machine ID and location.
• When the stop began and when the machine ran again, ideally to the minute.
• Category: failure, planned, or external.
• The component or failure mode, such as bearing, motor, sensor, or pneumatics.
• What was done and which spare parts were used.

Add running hours per machine. The trouble is that many plants still write this by hand: times get rounded, short stops are forgotten, and forms are filled in after the shift. Our guide to digitising production logsheets covers how to move them to a digital system.

Capturing downtime automatically from the machine

Stop and restart times do not need to be typed in. They can be read from signals the machine already has:

• Run status from the PLC. Many PLCs hold a running or stopped status in a register that can be read over Modbus RTU.
• Motor current. An energy meter on the machine panel shows when the main motor draws current. The rule can be simple, such as treating the machine as running when current stays above a set level for more than a minute.
• Vibration. A vibration sensor on a motor or pump also shows whether it is turning, and gives early warning of wear.

From these signals, stop and restart times come straight from the data, along with running hours, without relying on anyone's memory. What a sensor cannot know is why the machine stopped. So the most reliable pattern is a mix: times are captured automatically, and the operator or technician only picks the category and cause.

Reading and using MTBF and MTTR

• Watch the trend per machine, not one monthly figure. MTBF falling several months in a row points to wear or missed maintenance.
• Compare similar machines. Twin pumps with very different MTBF deserve a look at their load, installation, or repair quality.
• Build a Pareto of failure modes. A handful of causes usually account for most failures.
• Raise MTBF through root cause analysis, well-chosen preventive maintenance, and predictive maintenance that turns sudden breakdowns into planned repairs.
• Cut MTTR with critical spares on hand, repair SOPs, and alarms that reach technicians directly, so the initial wait shrinks.

Limits to keep in mind

• Few failures make the figure unstable. With two failures a year, one extra event swings MTBF sharply. Use a longer period or pool similar machines.
• MTBF is not machine life. An MTBF of 176 hours does not mean the machine will fail at exactly 176 hours.
• Brochure MTBF figures are measured under specific conditions and cannot be compared directly with numbers from your shop floor.
• The numbers are easy to dress up by moving failures into the planned category. Written category rules and automatically captured times reduce the temptation.

Frequently asked questions

What is a good MTBF?

There is no universal figure, because it depends on machine type, load, and how failure is defined. The most useful benchmark is the history of that same machine and of similar machines in the same plant.

How do MTBF and MTTR relate to OEE?

Both explain the availability part of OEE. Availability can drop because a machine fails too often (low MTBF) or takes too long to fix (high MTTR), and the remedies differ.

Does waiting for spare parts count toward MTTR?

In the definition used here, yes, because the machine is not producing. Some plants split it out. Pick one approach, write it down, and use it consistently.

MTBF and MTTR with INCLUDE

InAsset monitors vibration, temperature, and running hours for each machine in real time. From the last 30 days of vibration trend, it estimates how many days remain before a machine reaches the ISO 10816 damage limit, with service schedules and alarms that can be forwarded to Telegram. On the same platform, PLC run status can be read through IncludeGateways (Modbus RTU) and motor current through an energy meter. The data history can be exported to calculate MTBF and MTTR with your own category rules.

Once there is enough failure history and sensor data, the INCLUDE predictive analytics and machine learning team can build models that warn before a failure, so more repairs can be planned rather than forced by a sudden breakdown.