MTTF and MTTR appear to be simple reliability measures. The formulas are straightforward. The difficult part is ensuring the data being used actually represents the operating state of the asset being measured.
That is where many plant performance and reliability systems get it wrong.
For an automated system to calculate accurate asset-level reliability metrics, the measurement point needs to represent the actual operating state of that asset. In most cases, this means using a tag directly associated with whether the equipment is running or stopped — not simply an instrument or process tag located somewhere around the equipment.
The distinction is critical.
Consider a crusher. If the system determines crusher operation from a downstream belt scale, tonnes-per-hour measurement or another process instrument, a zero value does not necessarily mean the crusher has failed. The crusher may have stopped because there is no upstream feed, because downstream equipment is unavailable, because of an operational decision, or because the instrument itself is producing an incorrect signal.
If the software interprets every one of those conditions as a crusher failure, the reliability calculation is immediately compromised.
The same problem occurs with MTTR. The system needs to know when the actual asset stopped and when that same asset genuinely returned to service. Using an indirect process signal can move either of those timestamps, making the calculated repair duration longer or shorter than the real event.
This matters because MTBF — formally used for repairable assets — is calculated from operating time and the number of failures, while MTTR is calculated from the duration of failure or repair events. If the underlying start and stop times are wrong, the mathematics may be perfect while the result is still wrong. IBM's reliability guidance describes these metrics in exactly these terms, while maintenance software provider Fiix specifically notes that accurate MTBF depends on correctly tracking when an asset is actually online and offline.
Measurement Point Selection Matters
A good reliability model therefore starts before any calculation takes place.
For each critical asset, the system should identify a measurement point that directly and consistently represents the asset's operating condition. Process instrumentation can then provide additional context, but it should not automatically be treated as proof that the asset itself has failed.
This approach also allows upstream, downstream and unrelated process stoppages to be classified correctly rather than incorrectly appearing as failures against every affected asset.
This principle is central to how FixedPlantIQ approaches fixed plant measurement. The platform uses configured operational measurement points to automatically detect downtime and slow-running events, record event start and finish times, and calculate reliability measures including MTTF and MTTR.
The objective is not to collect every available tag from the historian. Fixed plants already have enormous amounts of data.
The objective is to select the right measurement point for the right asset.
Because when the measurement point is wrong, the failure history is wrong. And when the failure history is wrong, MTTF, MTBF and MTTR can become precise-looking numbers that do not accurately represent the reliability of the equipment.
Good reliability analysis starts with good measurement design.