Should reliability prediction be part of your reliability plan?
Join Carl and Fred as they discuss the controversial subject of reliability prediction. The discussion centers around the value of reliability prediction and whether or not it should be part of a reliability plan.
Some of the discussion points include:
- reliability prediction often depends on outdated manuals and guides.
- reliability prediction can obscure needed reliability tasks when the predictions are without sufficient merit.
- initial reliability predictions can be integrated with system reliability models to provide the reliability-wise mathematics.
- initial reliability predictions should be updated with more objective data when available.
- don’t use reliability predictions unless required or unless you have substantial and credible information supporting the analysis.
- If reliability predictions are required, be sure to advise the customer of the limitations and assumptions.
Enjoy an episode of Speaking of Reliability, where you can join friends as they discuss reliability topics.
Join us as we discuss topics ranging from design for reliability techniques to field data analysis approaches.
SOR 069 Should reliability prediction be part of your reliability plan?Carl Carlson
Tim Gaens says
So you guys take the statement not to do “reliability prediction”, because the numbers have no use.
But on the other hand you have an example where you discover a problem in an early phase by doing a “reliability prediction”.
What is the lesson we need to take here?
Fred Schenkelberg says
Reliability predictions are a tool and when used for the purpose where they add value, it’s find.
The problem is using a parts count prediction as the end all estimate of reliability performance. Or, as the only activity in DFR or the entire program. Placing prediction results on a data sheet is just plain wrong.
Predictions while not useful to estimate future reliability performance, they are useful to spot trouble areas (too hot, not enough margin, too many components, etc.) They are also useful when comparing different architectures for a design – for example how much parallel structure do we need to get the desired performance.
Also, they can be useful to encourage the design team to ask better questions about the reliability impact of design decisions. The relative change of a component failure rate on the overall impact while not absolute, does provide a relative magnitude impact.
Like all modeling work – the model is wrong and sometimes useful. Unfortunately we use the term “prediction” when in fact the common prediction tools are patently not useful for predictions.