PdM PlatformThe ToolboxServicesAcademyLibraryAboutContactOpen PdM →Open the Toolbox →
Reliability Engineering

Availability, reliability & RAM: how uptime is engineered

A machine can be highly reliable and still have poor availability โ€” if every rare failure takes a week to fix. Availability is the number the business actually feels, and it has just two levers: fail less often, or recover faster. This guide separates reliability from availability, builds the A = MTBF/(MTBF+MTTR) formula, shows how systems combine in series and through redundancy, and puts the “nines” in perspective โ€” with an interactive calculator.

A = MTBF/(MTBF+MTTR)RedundancyRBDThe nines
In short

Reliability = probability of running without failure over a period. Availability = the fraction of time the asset is actually up: A = MTBF/(MTBF+MTTR) = uptime/(uptime+downtime). The two levers are MTBF (fail less) and MTTR (repair faster).

Components in series multiply โ€” A = Aโ‚ยทAโ‚‚ยทโ€ฆ โ€” so more parts in a chain means lower availability (the weakest link rules). Redundancy (parallel) does the opposite: A = 1 โˆ’ (1โˆ’Aโ‚)โฟ drives availability up fast.

Availability is counted in “nines”: 99% (two nines) โ‰ˆ 88 h/yr down; 99.9% โ‰ˆ 8.8 h/yr; 99.99% โ‰ˆ 53 min/yr. RAM modelling rolls reliability + maintainability across a whole plant to predict production availability.

Reliability is not availability

The two words are used interchangeably in conversation and mean very different things in engineering:

That difference is why a very reliable item can still have mediocre availability: if it only fails once a year but each failure means a two-week wait for a spare and a specialist, the downtime dominates. Conversely, something that fails often but is back in minutes can post excellent availability. The business feels availability; reliability is one of the two ingredients that produce it.

The availability formula

Availability comes straight from the mean times between and to repair:

A = MTBF / (MTBF + MTTR) = uptime / (uptime + downtime) MTBF = mean operating time between failures. MTTR = mean time to repair (often expanded to mean time to restore, including detection and logistics). Raise A by growing MTBF or shrinking MTTR โ€” the two independent levers.

The two levers map onto the two halves of the Academy. Growing MTBF is the reliability agenda โ€” better design, RCM, precision maintenance, defect elimination via RCA. Shrinking MTTR is the maintainability agenda โ€” fast detection (condition monitoring), planned, kitted jobs, spares on the shelf, accessible design. RAM analysis is just doing this arithmetic across an entire plant.

Three kinds of availability

“Availability” gets quoted at three levels, and the gaps between them are where money hides:

The lesson: chasing inherent availability through better hardware while ignoring the logistics-and-planning gap leaves most of the prize on the table.

Combining systems: series vs redundancy

Real plant is many components together, and how they’re arranged in the reliability block diagram (RBD) changes everything:

Series:  A = Aโ‚ ยท Aโ‚‚ ยท โ€ฆ ยท Aโ‚™     Parallel (1-of-n):  A = 1 โˆ’ (1โˆ’Aโ‚)โฟ In series, all must work, so availabilities multiply โ€” adding components only lowers the total. In active parallel, any one working keeps the system up, so the unavailabilities multiply โ€” redundancy crushes downtime.

Series is sobering: ten components each at 99% give a system at 0.99ยนโฐ โ‰ˆ 90% โ€” the chain is much weaker than any link. Redundancy is the cure: two parallel units each at 99% give 1 โˆ’ 0.01ยฒ = 99.99%, a hundred-fold cut in downtime. That is why critical duties run installed spares (2ร—100%) or voting arrangements (2-out-of-3). The calculator shows both effects โ€” set the single-unit MTBF and MTTR, then add redundant units:

Interactive — Availability & redundancy

Live model
Mean operating time between failures — uptime only, repair time excluded (fail less)
Mean time to restore (repair faster)
1 = single ยท 2/3 = installed spares
Single-unit availability
โ€”%
MTBF/(MTBF+MTTR)
System availability
โ€”%
single unit
Downtime / year
โ€”
at 8760 h/yr
The nines
โ€”
โˆ’logโ‚โ‚€(1โˆ’A)
Annual downtime vs repair time
How MTTR and redundancy drive downtime โ€” note the log scale
single2 parallel3 parallelyou are here
Model: Aโ‚ = MTBF/(MTBF+MTTR); redundant A = 1โˆ’(1โˆ’Aโ‚)โฟ (active 1-of-n parallel); annual downtime = 8760ยท(1โˆ’A) h; nines = โˆ’logโ‚โ‚€(1โˆ’A). Idealised: assumes independent failures, perfect switchover and unlimited repair crews. Real RAM models add common-cause failure, repair-resource limits, logistics delay and partial-capacity states via simulation.

The nines

Availability is so often near 100% that it’s spoken of in nines โ€” and each extra nine is a ten-fold cut in downtime, and usually a large step in cost:

Availability“Nines”Downtime per year
90%one nine~36.5 days
99%two nines~3.65 days (88 h)
99.9%three nines~8.8 hours
99.99%four nines~53 minutes
99.999%five nines~5.3 minutes

Two practical truths fall out. First, the cost of each extra nine climbs steeply โ€” pushing from two to three nines is usually achievable through better maintenance management; three to four often needs redundancy and design changes; five-nines is a deliberate, expensive architecture. Second, you should only buy the nines a duty actually needs: spend them where downtime is genuinely costly or unsafe, decided by the same criticality ranking that drives the rest of the strategy.

This is where reliability becomes a number the business buys. RAM modelling lets you compare options โ€” an extra spare pump vs faster spares logistics vs a more reliable seal โ€” on one currency: production availability, and therefore revenue. It draws its inputs from Weibull life data and OREDA/ISO 14224 failure rates, and its MTTR side from the planning and spares systems Bluestream implements in the CMMS.

Key takeaways

Related guides