Interpreting Test Results
Abstract
Carl and Fred discussing a listener question about reliability demonstration testing; specifically how to deal with unrelated failures when using “success testing.”
Key Points
Join Carl and Fred as they discuss “success testing,” and the assumptions and principles that need to be considered.
Topics include:
- Planning and interpreting binomial testing
- Pass-fail testing versus test-to-failure
- What assumptions go into “success testing”?
- Example of cracked glass in a device
- The importance of understanding the underlying failure mechanisms
- What happens when different mechanisms of failure are lumped together?
- The role of confidence interval in success testing
- What happens when unrelated failures are not independent?
- Can system testing be designed, without using Weibull parameters?
- When designing a test, how do you include the user conditions?
- When can you assume independence of failure mechanisms in a complex structure?
- Ask questions about how the planned tests represents the real world
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.
- Social:
- Link:
- Embed:
Show Notes
Information about the relationship between sample size, testing time, confidence, and reliability can be found in a book titled “Statistical Design and Analysis of Engineering Experiments,” by Charles Lipson and Narenda Sheth, McGraw Hill, 1973. Reference section 5.4, page 178.
I really like the oboe opening the last movement of Danzi wind quintet Opus 67! I got it for Vineyard Wind Quintet.
Good question Fred, and good answers Carl! Reminds me of load sharing and dependence of glass cracking on other failure modes. I.e., P[Glass crack| other] depends on other failure modes. Imagine the case failing softer and cushioning the glass. Censoring distribution may not be independent of time to failure distribution.
Stay tuned for article on “Credible Reliability Test Planning,” based on Bayes and “Credible Reliability Prediction.” It’s supposed to be ASQ Silicon Valley statistics discussion group Nov. 17. I’ll turn it into article after that.