What reliability is
Reliability is the probability that a piece of equipment performs its function without failing for a given time and under defined operating conditions. It is a probability, so it always comes with a time frame: it makes no sense to say a pump is reliable without saying for how many hours.
It helps to separate it from two related concepts. Availability is the fraction of time in which the equipment is able to operate. Maintainability describes how quickly it is returned to service after a failure.
The three starting indicators
A worked example
A feed pump ran for 8,000 hours in the year, had 4 failures and accumulated 40 hours of repair.
- MTBF
- 8,000 ÷ 4 = 2,000 hours
- MTTR
- 40 ÷ 4 = 10 hours
- Availability
- 2,000 ÷ 2,010 = 99.5%
If the failure rate is constant, reliability for a mission time t is calculated as follows:
For one month of continuous operation (720 hours), R = exp(−0.36) = 69.8%. In other words, the same pump that shows 99.5% availability has roughly a 30% chance of failing within the month. Both numbers are correct and they answer different questions.
When the failure rate is not constant: Weibull
In practice, many machines fail more often early in life or as they wear out. The Weibull distribution describes both cases with two parameters, the characteristic life η and the shape parameter β.
- β below 1: early failures, linked to installation, commissioning or spare part quality.
- β close to 1: random failures, with a constant rate.
- β above 1: wear-out; the probability of failure grows with operating hours.
The value of β changes the strategy. Replacing components on a time basis only makes sense when there is wear-out. With early or random failures, preventive replacement does not reduce failures and can increase them.
Common mistakes
- Calculating MTBF from very few failures and treating it as an exact value.
- Leaving out equipment that has not failed yet. That censored data also carries information and must be included in the fit.
- Mixing different failure modes in the same analysis.
- Using availability as if it were reliability.
At Delta Mechanics we run this analysis on your plant's failure history. Our Farellón platform fits the distributions with censored data and compares the models, and an engineer reviews the result against the physical failure mechanism.
