Date of Award
Thesis - Restricted
Master of Science (MS)
Electrical and Computer Engineering
Situations in which historical data is not available are simulated using availability estimation and importance sampling. Determining functional time periods through though simulation of highly dependable systems can be a lengthy process. The actual computer run time can be hours, days or possibly even months. This is due to the fact that failure of these highly dependable systems is rare. In addition to this rarity of failure, numerous failures of these systems are required to produce an accurate estimation of their availability and the variance of this availability. Unlike availability estimation importance sampling has been used to reduce simulation time of highly dependable, unrepairable systems. Methods have been devised to utilize uniformization and exponential transformation techniques specifically for this purpose. Under some loosely defined conditions, exponential transformation has been shown to be a successful technique. Availability estimations which rely on numerous independent and computational variables present in the system. The functionality of these availability estimators produced by exponential transformation has been unexplored. To speed up the failure process, in this investigation importance sampling will be used. Importance sampling speeds up the failure process by accelerating the failure rate, allowing more failures to occur in a reasonable amount of time. A scaling technique is then used to transform these accelerated failures back into the original time frame. A reasonable estimation of system parameters may be acquired once a sufficient number of failures (samples) have been obtained. In the investigation of importance sampling and exponential transformation, a number of non-linear phenomenon occur due to the interaction of variables when obtaining an estimation of availability. During this investigation, relevant simulation parameters and their influence on the availability estimation will be obtained through importance sampling and explored based upon accuracy and stability considerations.
Nowicki, Erica, "Availability Estimation via Importance Sampling: A Parameter Investigation" (1996). Master's Theses (1922-2009) Access restricted to Marquette Campus. 4603.