In short. Reliability is the chance a system works for a period; availability is the share of time it is ready. Redundancy raises both, but only when failures are independent. The study calculates the numbers, finds common-cause weak points and shows how maintenance and repair time change the result.
Core measures
Availability = MTBF ÷ (MTBF + MTTR)
| Term | Meaning |
|---|---|
| MTBF | Mean time between failures |
| MTTR | Mean time to repair, including spares and travel |
| Failure to start | A distinct, often dominant mode for standby generators |
Series and parallel
In series, every element must work, so availabilities multiply and the total is lower than the weakest element. In parallel, the system fails only if all paths fail, so unavailabilities multiply.
Aparallel = 1 − (1 − A₁)(1 − A₂)
Why real systems fall short
- Common-cause failures: shared fuel, shared control, shared switchgear, one maintenance error.
- Standby failure on demand: an idle set may fail to start even if reliable while running.
- Repair delays: missing spares extend MTTR.
- Maintenance outages: during maintenance, redundancy is reduced.
- Human error: a leading cause in critical facilities.
Improving the result
- Remove single points of failure. See FMEA.
- Add redundancy where it matters. See N, N+1, 2N architectures.
- Reduce MTTR with spares, training and remote diagnostics.
- Test regularly on load. See load-bank testing.
- Track failures and trends. See AMC.
Illustrative example
A single generator has an availability of 0.98 on demand. Two independent units in parallel give 1 − 0.02 × 0.02 = 0.9996. But if both depend on the same fuel line with availability 0.995, the fuel line caps the system near 0.995 in practice. Adding a second fuel route or day tanks per set restores the benefit.
Three measures, not one
| Term | Meaning |
|---|---|
| Reliability | Probability of operating without failure over a period |
| Availability | Fraction of time the system is able to supply |
| Redundancy | Spare capacity or paths that keep supply when a part fails |
Availability = MTBF ÷ (MTBF + MTTR)
Worked example (illustrative)
A generator with MTBF of 2,000 hours and repair time of 24 hours has availability 2,000 ÷ 2,024 = 98.8 percent. Two independent sets, either of which can carry the load: unavailability of each is 1.2 percent, so both down together is 0.012² ≈ 0.014 percent, giving availability of about 99.986 percent. Shared fuel supply or shared controls reduce this, because the failures are no longer independent.
Where studies find weakness
- Common-mode failures: shared fuel, shared battery, shared control.
- Repair time, which is often the largest lever; spares on site shorten it.
- Start failure rate, usually the dominant failure of standby sets, which testing reduces.
Data sources
Use maker data, industry databases and the site’s own failure history. Label values as estimates and show sensitivity.
Decision use
The result supports a choice between N+1, 2N or other architecture, balanced against cost. See N to 2N architecture and FMEA.
Common mistakes
- Counting redundancy without independence.
- Using MTBF without standby failure on demand.
- Ignoring repair time.
What to send us
Send failure history and the architecture. Send your application details.
Questions engineers ask
What does 99.99 percent mean?
About 53 minutes of unavailability a year. Define what counts as downtime.
Is redundancy enough?
No. It needs independence, testing and good maintenance.
Why do standby sets fail to start?
Common causes are batteries, fuel and control faults, which are found by regular testing.
How to get reliability data?
From your own records first, then manufacturer and industry sources.
Values are illustrative. Use actual data for real calculations.