Effects of correlation between the adjacent components on time dependent failure probability of corroded pipelines

This study aims to investigate the effects of correlation of random variables between the adjacent components on time variant system reliability of corroded pipelines. The system reliability analysis is conducted using a combination of Latin Hypercube Sampling and Cholesky decomposition methods for four different failure modes (Composite (C), Corrosion Perforation (CP), Local Burst (LB) and Rupture (R)). Since the multivariate target correlation matrix might be non-positive definite, a combination of two efficient optimization techniques is implemented to find the minimum perturbation for transforming the non-positive definite correlation matrix into a positive definite one. The findings of the paper indicate that there is no monotonic order in either lower or upper bound of the system failure probability of corroded pipelines with uniform variation in target correlation matrix. Furthermore, some correlations can change the upper bound of the system failure probability of corroded pipelines about 10%. In addition, the cumulative upper bounds of the system failure probability of the investigated case study for C, CP, LB and R failure modes reach unity after 9, 10, 11 and 13 years, respectively.

Automated summary

This study investigates the effects of random variables correlation on the system reliability of corroded pipelines. The results show that there is no monotonic order in the system failure probability with uniform variation in correlation matrix. Some correlations can change the upper bound of the system failure probability by about 10%. Additionally, the cumulative upper bounds of system failure probability for different failure modes reach unity after 9 to 13 years.