Continuous Reliability Models for Systems of Non-identically Distributed Correlated Components  
Author Bentolhoda Jafary


Co-Author(s) Lance Fiondella


Abstract Most methods to model system reliability assume component failures are statistically independent. However, in many cases, components of a system can experience correlated failures. Existing continuous multivariate distributions to model correlation between the failures of components that deteriorate over time impose additional restrictions such as the life distribution each component follows and the correlation structure between these components. Moreover, these methods to model a system with correlated components commonly introduce an exponential number of parameters. This paper presents a method to model the reliability of a continuous system composed of non-identically distributed correlated components. The method allows each component to follow a unique distribution and introduces a quadratic number of correlation parameters to characterize correlations between the failures of each pair of components. We demonstrate our approach through examples, which illustrate the impact of correlated component failures on system reliability. Several important metrics for continuous systems are also considered, including mean time to failure, availability, and mean residual life. These examples illustrate correlated failures can negatively influence system reliability and the associated metrics. Thus, the approach can quantify the improvements to system reliability and metrics that could be achieved by lowering the correlation between the failures of a system’s components.


Keywords Continuous reliability model, correlated failure, mean time to failure
    Article #:  2084
Proceedings of the 20th ISSAT International Conference on Reliability and Quality in Design
August 7-9, 2014 - Seattle, Washington, U.S.A.