When I studied Reliability Engineering in graduate school, I was taught that failure rate data came from field failure studies. I assumed that operating companies always kept accurate records when a piece of equipment failed. I also assumed that someone would investigate the failures and assign a root cause. Later, I found that is not quite reality in many industries. Even so, field failure data is a valuable source of information even when the reports are incomplete and information is missing.
This past summer, exida completed a large field failure study on pressure transmitters and remote seals. This study was based on manufacturer warranty return data. Experts know that type of data should not be used by itself to calculate failure rates, primarily because not all failures are reported to the manufacturer. However, the data is useful. It can be analyzed to reveal existing failure modes, establish ratios of failure modes, and provide bounds on failure rates.
At exida, we use a model-based failure rate / failure mode prediction technique. Models for particular products, even specific manufacturers and model numbers, are established using detailed production design information via an FMEDA. Given that model, the results from field failure studies can be compared in a feedback loop method to reveal any weakness in the model. Model parameters will be adjusted to improve accuracy.
The field failure studies we have done hit the two billion unit operational hour mark. This quantity of data has been used to tune our component failure database. When that database is used in the FMEDA model technique, the failure rate / failure mode predictions seem to match the field data quite reasonably. One, of course, can never stop gathering field failure data and checking the models in the quest for better accuracy.
Tagged as: SIL safety lifecycle reliability engineering pfh calculations pfh PFDavg fmeda field failure studies failure rate data Dr. William Goble