TY - JOUR
T1 - Uncertainty propagation and sensitivity analysis in system reliability assessment via unscented transformation
AU - Rocco Sanseverino., Claudio M.
AU - Ramirez-Marquez, José Emmanuel
PY - 2014/12
Y1 - 2014/12
N2 - The reliability of a system, notwithstanding it intended function, can be significantly affected by the uncertainty in the reliability estimate of the components that define the system. This paper implements the Unscented Transformation to quantify the effects of the uncertainty of component reliability through two approaches. The first approach is based on the concept of uncertainty propagation, which is the assessment of the effect that the variability of the component reliabilities produces on the variance of the system reliability. This assessment based on UT has been previously considered in the literature but only for system represented through series/parallel configuration. In this paper the assessment is extended to systems whose reliability cannot be represented through analytical expressions and require, for example, Monte Carlo Simulation. The second approach consists on the evaluation of the importance of components, i.e., the evaluation of the components that most contribute to the variance of the system reliability. An extension of the UT is proposed to evaluate the so called "main effects" of each component, as well to assess high order component interaction. Several examples with excellent results illustrate the proposed approach.
AB - The reliability of a system, notwithstanding it intended function, can be significantly affected by the uncertainty in the reliability estimate of the components that define the system. This paper implements the Unscented Transformation to quantify the effects of the uncertainty of component reliability through two approaches. The first approach is based on the concept of uncertainty propagation, which is the assessment of the effect that the variability of the component reliabilities produces on the variance of the system reliability. This assessment based on UT has been previously considered in the literature but only for system represented through series/parallel configuration. In this paper the assessment is extended to systems whose reliability cannot be represented through analytical expressions and require, for example, Monte Carlo Simulation. The second approach consists on the evaluation of the importance of components, i.e., the evaluation of the components that most contribute to the variance of the system reliability. An extension of the UT is proposed to evaluate the so called "main effects" of each component, as well to assess high order component interaction. Several examples with excellent results illustrate the proposed approach.
KW - Importance index
KW - Monte Carlo simulation
KW - Sensitivity analysis
KW - Uncertainty propagation
KW - Unscented transformation
UR - http://www.scopus.com/inward/record.url?scp=84906698841&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84906698841&partnerID=8YFLogxK
U2 - 10.1016/j.ress.2014.07.024
DO - 10.1016/j.ress.2014.07.024
M3 - Article
AN - SCOPUS:84906698841
SN - 0951-8320
VL - 132
SP - 176
EP - 185
JO - Reliability Engineering and System Safety
JF - Reliability Engineering and System Safety
ER -