Time Between Failures, Repair Time and Availability

Why it matters

One piece of equipment may give few but long failures. Another may have many short failures. Even if the two cases produce the same total stoppage, the maintenance need is not the same. Mean time between failures, abbreviated MTBF, divides run exposure by the number of failures. Mean time to repair, abbreviated MTTR, divides the defined repair scope by the number of completed repairs. The first indicator approaches the frequency dimension, the second the duration dimension. These averages do not say when individual events will occur. They are not an appointment that the next failure will occur exactly at the average value. They must be interpreted together with the raw event record and information on the distribution.

Before the calculation, the meaning of the times is determined. If run time describes the time the equipment actually ran, stoppages are not added to this total. Active repair, in the example of this lesson, is the elapsed time of the intervention jobs actually performed to remove the failure. Waiting for parts and waiting for the arrival of the authorized person are kept separately. In another report MTTR may cover the whole time to restore. Comparing the two approaches directly, because they carry the same name, is wrong. The definition used must be visible in the report and must not be silently changed within the period. As much as a single number, what times went into that number is also a learning outcome.

Simple availability can be calculated from MTBF and MTTR under conditions that cover only run and defined repair times. It does not automatically represent a whole shift that contains planned maintenance, waiting for parts or externally caused stoppages. Also, availability alone does not show the soundness of the parts produced or the suitability of the run rate. Equating this ratio with the overall equipment effectiveness in M9 makes different losses invisible. While equipment is ready to run there may be no demand; while running it may produce slowly or defectively. The purpose of the indicator is not to melt all these results into a single number. It is to answer the relevant question with the right time scope.

Event boundaries directly affect the average. Counting the interrupted interventions of the same failure as separate failures lowers MTBF; hiding failed attempts in a single record can also conceal repeat behavior. The start of the failure, the moment the function returns and the restart verification must be defined. The end of an event still open at the end of the period is not invented. The observed accumulated stoppage can be reported; the final repair time is not yet known. Similarly, infinite reliability cannot be deduced from an observation period with zero failures. Indicators provide decision support only together with the number of events, the scope and information on missing records.

Records with the same average may have different distributions. When some durations are very short and some very long, a single average is not enough to describe the reliability of the maintenance window. No new model is built for the distribution in this lesson; keeping the events individually is what is asked. In a later review, the conditions in which long repairs occur can be investigated from these records. When comparing equipment, the length of the observation should be explained along with the product load and the record threshold. A change seen over a short period may be coincidental. For this reason, when an improvement is seen in the indicator, a lasting cause is not declared immediately. The technical change, a comparable period and the event evidence are assessed together. Numerical accuracy does not mean uncertainty has been removed; even when the calculation is right, the interpretation may still be too definite.

Formula