To ensure that your results are valid, consider the following guidelines when you collect data, perform the analysis, and interpret your results.
- The data that you collect are usually failure times
- For example, you might collect failure times for units that are tested at a specific temperature. You might also collect samples of failure times under different temperatures, or under different combinations of stress variables. Alternatively, the data might measure product usage in units other than time, such as the amount of mileage for a tire until it fails.
- Use this analysis for data with any censoring scheme.
- Censoring schemes include right-censored, interval-censored, left-censored, and
combinations of different types.
- Exact failure data
- Record the failure time at a fixed point in time before the end
of the study or reliability test
- Right-censored data
- Record a time when the unit still survived at the end of the
study or reliability test. For example, a patient may quit a
study before it ends, or a unit is still operating after the
allocated testing time. The exact failure time is unknown but
will certainly occur beyond the allocated study time.
- Interval-censored data
- The exact failure time is unknown, but it occurred in a known
interval. For example, a unit was still operating at 12 when the
reliability engineer left for a mid-day break. Upon the
engineer’s return at 12:30 they noticed that the unit has
failed. Thus, the unit failed in the interval starting at 12 and
ending at 12:30.
- Left-censored data
- Left-censored data is a special case of interval-censored data
point where the beginning point of the interval may be time 0 or
may be unknown. The engineer only knows that the unit failed by
some specific time since the beginning of the reliability
test.
For more information on censored data, go to Data censoring.
- If your data are multiply censored, you must have a column of censoring indicators
- If your test units are censored at different times, your data are multiply censored. With
multiply-censored data, failure times are intermixed with censoring times. In
that case, you must have a column with text or numeric values that indicate
whether each observation was an actual failure or a unit that was removed from
the test before failure (censored). For more information on censored data, go to
Data censoring.
- To evaluate each cause of failure separately, record failure modes
- A failure mode indicates the cause of a failure. Because different failure modes often have different failure distributions, consider grouping the data by failure mode, when possible. To obtain results for each failure mode, as well as for the overall system, record the cause of each system failure in a worksheet column. For more information, go to What is a failure mode?.
