Availability: Inherent, Achieved and Operational, and Why the Distinction Changes Your Number

Report an availability figure for the same crusher to two different people at the same site and both can walk away with a correct number that disagrees with the other's by several points. That is not a rounding error and it is not a reporting mistake. Inherent, achieved and operational availability are three distinct, legitimate calculations, and the gap between them is exactly where a system's calculated reliability meets its measured performance in practice, the same gap this series left open at the close of reliability block diagrams.

Availability looks like the simplest number in reliability engineering: one percentage, one line on a report, seemingly beyond argument. It is also the number most often quoted without saying which of three definitions produced it, which is how the same asset ends up reported at 99.5% in one meeting and 94.3% in another with nobody actually wrong. This page builds inherent, achieved and operational availability from one worked asset, a primary gyratory crusher, carries it through all three formulas in turn, and itemises exactly what each added layer of definition costs.

An inherent availability figure describes the design in isolation. An operational availability figure describes the organisation running it, parts store, permitting process and production schedule included. Neither is wrong on its own terms, and treating one as a stand-in for the other is exactly where a defensible number turns into a disputed one, and where an available asset can quietly stop being a reliable one.

Inherent Availability: What the Design Alone Promises

Inherent availability is the first of three availability figures for the same asset (MIL-HDBK-338B 1998), and it is the number an equipment designer can actually be held to, and nothing else.

It counts only corrective maintenance, the repair that follows an unplanned failure, and excludes everything an organisation does around the asset: no waiting for parts, no permit, no preventive task. That is a deliberate, narrow scope, not an oversight. It is the figure that isolates design and build quality from everything operations later adds or subtracts, and it is the version of "availability" most vendor specifications are actually quoting (MIL-HDBK-338B 1998).

The formula is a plain ratio of up-time to total time inside that narrow, corrective-only world: MTBF hours running for every MTTR hours down, repeated indefinitely. Treat the MTTR in this ratio as a mean, and treat that mean with some caution. Repair times are typically right-skewed and lognormal rather than normal (MIL-HDBK-338B 1998), so a handful of awkward repairs sit well out past the average, and a shutdown window planned against the mean alone is exactly the plan a long-tail repair breaks. MTBF belongs here because it is a repaired item's event stream; MTTF, the mean life of a non-repairable item and a distinct quantity covered earlier in this series, has no place in this formula, because a crusher is restored to service, not discarded, on failure.

A primary gyratory crusher, typically the highest-consequence, lowest-redundancy asset in a hard-rock processing circuit and the kind of equipment an asset-criticality assessment ranks at the very top, sets the numbers for every calculation on this page. Its corrective failure history gives MTBF = 2,000 operating hours and mean active corrective repair time MTTR = 20 hours, so Ai = 2,000 / (2,000 + 20) = 0.9901, or 99.01%.

When the repair-to-interval ratio is small, MTTR/MTBF < 0.10, a shortcut holds: Ai ≈ 1 − MTTR/MTBF. Here MTTR/MTBF = 0.01, and the shortcut gives 99.00% against the exact 99.01%, close enough for a quick estimate. The shortcut stops being trustworthy exactly where it would matter most: an item with MTBF = 80 hours and MTTR = 20 hours has the same arithmetic form but a ratio of 0.25, and the shortcut returns 75% against an exact 80%, a five-point error in the wrong direction for anyone deciding whether the number is acceptable. Reach for the exact ratio whenever MTTR climbs past about a tenth of MTBF.

Achieved Availability: Bringing the Maintenance Plan Into the Number

Achieved availability asks a slightly bigger question: what happens to the figure once the maintenance plan, not only the failure history, is switched on.

MTBM folds two clocks into one. λ is the corrective failure rate, 1/MTBF; f is the preventive maintenance frequency, however often a scheduled task is due. Adding a preventive task to the calendar always shortens the mean interval between actions of any kind, because it is one more event competing for the same timeline, and that is the whole point of achieved availability: it prices in the maintenance strategy, not only the failure rate.

The crusher's liner and lubrication programme calls for a preventive task every 400 hours, mean active time 4 hours. That gives λ = 1/2,000 = 0.0005 per hour and f = 1/400 = 0.0025 per hour, so MTBM = 1/(0.0005+0.0025) = 333.3 hours. M̄, the mean active time weighted by how often each type of work actually happens, works out to 6.7 hours. Aa = 333.3/(333.3+6.7) = 0.9804, or 98.04%, already almost a full point below Ai, and every one of those points is a preventive task that was scheduled on purpose.

Operational Availability: What the Organisation Around the Asset Actually Delivers

Operational availability is the number the plant actually lives inside, because it is the only one of the three that counts what happens before a technician can put a spanner on the machine at all.

MDT is M̄ with two more clocks added on top (MIL-HDBK-338B 1998): mean logistic delay time, the wait for parts or an available crew, and mean administrative delay time, the wait for a permit, an isolation or a confirmed production window. Neither delay involves a tool touching the asset, and neither shows up in Ai or Aa at all, which is exactly why Ao routinely disappoints anyone who expected it to sit close to the design figure.

The crusher's mean logistic delay time runs 10 hours, parts and crew mobilisation combined, and mean administrative delay time 4 hours, permit and production-window coordination combined, giving MDT = 6.7 + 10 + 4 = 20.7 hours. Ready time, scheduled hours the crusher is staffed and available but idle for reasons upstream of it, such as a blast-clearance interruption to the ore feed, runs 10 hours. Ao = (333.3+10)/(333.3+10+20.7) = 0.9432, or 94.32%.

1. 1

   Record MTBF and mean corrective repair time from the failure history alone, excluding every planned task, and compute Ai.

   This isolates the design-level figure the equipment and its build quality can actually be held to, before any operating decision is added.

2. 2

   Add the preventive maintenance interval and its mean active time, then compute MTBM = 1/(λ+f) and Aa.

   This is the first figure a maintenance-strategy decision moves, and it belongs to whoever sets the preventive programme.

3. 3

   Add mean logistic delay, mean administrative delay and ready time, then compute MDT and Ao.

   This is the number operations actually lives inside, and the only one of the three a parts, permitting or scheduling fix can shift.

4. 4

   Report all three figures together, and name which function owns the gap between each pair.

   A single reported number lets the reader assume whichever definition suits them, and the gap between the three is exactly the information one number throws away.

The Same Crusher, Three Numbers

Line the three figures up and the arithmetic does something a single number never could: it shows exactly where each percentage point went.

Two things stand out. First, preventive maintenance costs less than a point here, 0.97, because a well-run programme is a small, predictable tax on the calendar rather than a drain on it. Second, logistics and administration cost nearly four times as much, 3.72 points, and that is the largest single number on the page. Nothing about the crusher's mechanical condition changed between Aa and Ao, only how long the organisation around it took to get a technician in front of it with the right part and the right paperwork. Design engineering owns the first gap. Maintenance planning owns the second. Operations, procurement and the permitting process own the third, and it is usually the one a site can move fastest once it is named.

Availability Is Not Reliability

Two assets can report the identical availability figure and describe two completely different risk profiles, because availability and reliability answer different questions.

Availability asks what fraction of time an item is up. Reliability asks the probability an item survives a stated mission, under stated conditions, without failing at all, covered at the start of this series, and an asset can score well on one while scoring poorly on the other.

Consider two items, each returning Ai = 99.5% and each accumulating exactly 44 hours of corrective downtime a year. Item A fails often and is fixed almost immediately: MTBF = 199 hours, MTTR = 1 hour, roughly 44 events a year, each one a minor nuisance. Item B fails rarely and takes a while to put right: MTBF = 3,980 hours, MTTR = 20 hours, roughly two events a year, each one a genuine outage. Identical availability, identical annual downtime, and yet across a single 12-hour production shift Item A's reliability is 94.15%, a meaningful chance of tripping at least once that shift, while Item B's is 99.70%, almost certain to run the whole shift without incident, but with a consequence twenty times longer on the rare occasion it does not.

The two items demand different responses, and availability alone cannot tell you which one you are looking at. Item A needs root-cause attention to nuisance-trip frequency, because each event disrupts a process even when it costs almost no downtime. Item B needs consequence management: spares on the shelf, a standby unit, or a faster response plan, because its rare event is the one that actually stops the plant. An item that fails weekly and is restored in minutes can report excellent availability and mediocre reliability, and the fix for each failure mode is not the same fix.

Where the Gap Actually Comes From

Every point of the operational gap traces back to a named, fixable cause, which is what makes Ao worth calculating in the first place.

Waiting for parts and waiting for a crew both sit inside mean logistic delay time. Waiting for a permit and waiting for a confirmed production window both sit inside mean administrative delay time. None of the four is a design lever. All four are operational levers, owned respectively by procurement, workforce planning, the permit-to-work process and production scheduling, and moving any one of them moves Ao without touching the crusher itself.

One more limitation is worth naming honestly: the Ai/Aa/Ao model assumes a binary up-or-down state, and a crusher running at reduced throughput on a partially worn liner or a partially blocked chute is neither. It is available by every definition above and still not delivering full rate. That partial-capacity condition is a production-availability question, not an Ai/Aa/Ao question, and a site that only tracks the binary model can report excellent availability on an asset that is quietly under-delivering tonnes.

The same caution applies to redundant systems. A calculated system reliability built from independent component figures, covered earlier in this series, assumes the components fail independently of one another. A shared power supply, a shared control signal or a shared installation error breaks that assumption, and it is exactly why a redundant system's measured availability so often sits below its calculated figure: the independence the formula relies on was assumed rather than confirmed (Applied R&M Manual, ch. 30). Redundancy does not multiply availability quite as cleanly as the arithmetic suggests, for the same reason it does not multiply reliability quite that cleanly either.

The Distinction Is a Management Report, Not a Rounding Error

A 4.7-point gap between two availability figures sounds like a footnote until it is converted into the units a capital committee actually reads.

Take the 4.69-point gap between this crusher's Ai and Ao and apply it across a year of operation: 8,760 hours a year multiplied by 4.69% is close to 411 equivalent hours a year that inherent availability alone never revealed. On a crusher constraining a circuit rated at 2,000 tonnes an hour, at an illustrative margin of $18 a tonne, that gap is worth on the order of $14.8 million a year, not because the crusher is unreliable but because parts, permits and production windows are taking longer to align than the design ever assumed. That is the translation an availability figure needs before a capital committee can act on it: not a percentage, but a throughput and margin number attached to a named, ownable cause.

The question "what is this asset's availability" only has a defensible answer once it is three answers: inherent for the design, achieved for the maintenance strategy, and operational for everything else the organisation adds on top. Report all three, name what separates them, and the gap stops being an argument and starts being a work plan.