|
Activity Number:
|
516
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 5, 2009 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Physical and Engineering Sciences
|
| Abstract - #304741 |
|
Title:
|
Enhanced Monte Carlo Estimation of Extremely Small Probabilities of Failure
|
|
Author(s):
|
Peter W. Hovey*+ and Brian Krilov
|
|
Companies:
|
University of Dayton and University of Dayton
|
|
Address:
|
300 College Park Dr, Dayton, OH, 45469-2316,
|
|
Keywords:
|
Monte Carlo ; Extreme Value ; Reliability
|
|
Abstract:
|
Safely critical systems generally require high levels of reliability. For example, aircraft turbine engines must be designed to achieve an extremely high reliability. Current design strategies are focused on achieving a specific probability of failure for the engine. Traditional Monte Carlo techniques require excessive computing time because of the complexity of the finite element calculations that determine when a failure occurs and the large number of trials required to estimate a probability that is close to 0. A new method for analyzing Monte Carlo results based on extreme value theory is discussed that significantly decreases the number of simulations that are required, thus increasing computation speed. A comparison is made between maximum likelihood estimation, which is biased, and a simple linear estimate that is unbiased.
|
