Approximate Evaluation of Reliability and Availability Via Perturbation Analysis

The progress on a three-year effort to examine approximate reliability evaluation techniques for fault tolerant control and sensor system is described. The motivation for the work is provided by the fact that the reliability models for these system tend to be finite state semi-Markov models with lar...

Full description

Saved in:
Bibliographic Details
Main Authors: Walker, Bruce K, Chu, Sin-Kwong, Wereley, Norman M
Format: Report
Language:English
Subjects:
BEHAVIOR
COMPUTER ARCHITECTURE
Computer Systems
CONTINUITY
CONTROL
DETECTORS
EXPANSION
FAILURE
FAULT TOLERANT COMPUTING
MARKOV PROCESSES
MOTIVATION
PARAMETERS
PE61102F
PERTURBATIONS
RATES
RELIABILITY
SIZES(DIMENSIONS)
STEADY STATE
TEST AND EVALUATION
TIME
TRANSIENTS
WUAFOSR2304A5
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:The progress on a three-year effort to examine approximate reliability evaluation techniques for fault tolerant control and sensor system is described. The motivation for the work is provided by the fact that the reliability models for these system tend to be finite state semi-Markov models with large dimension that evolve relatively slowly in time due to the rare occurrence rate of component failures. The transient behavior of these models is of interest because the steady state behavior is trivial and not of practical importance. The evaluation of the transient behavior of such models, however, is intractable even for relatively simple system architectures because of the widely varying rates at which events occur in the model. The research effort concentrates on generating useful limit theorems that approximate the behavior of these models asymptotically well as the small component failure rates become vanishingly small. Using the work of Korolyuk as a starting point, such limit theorems are generated for both continuous and discrete time models that are representative of fault tolerant system behavior. In particular, the limit theorems of Korolyuk are expanded to cover models where the classes of the decomposed models include trapping states when the small parameter vanishes and to cover models where the holding times are not necessarily scaled by the small parameter.