Reliability analysis of corroded pipes using conjugate HL–RF algorithm based on average shear stress yield criterion

• We present a reliability scheme to compute failure probability of corroded pipeline.
• We establish a new sensitivity vector to determine conjugate vector of HL–RF method (CHL–RF).
• CHL–RF scheme is more robust than HL–RF method for nonlinear performance of pipeline.
• Probabilistic burst model is defined for corroded pipe using analytical and empirical models.
• Failure probability to versus elapsed time is more sensitive to defect depth at 0.2t < d(T)<0.65t.

In this paper, an algorithm with a new sensitivity vector for reliability analysis is presented in reliability analysis. The sensitivity vector is computed using the conjugate gradient vector of limit state function based on Hasofer-Lind and Rackwitz-Fiessler (HL–RF) method (CHL–RF algorithm). An appropriate probabilistic model which is included model error and burst model of corroded pipe that is defined based on burst pressure of intact pipes and remaining strength due to corrosion defect, is developed using the average shear stress criterion. The best burst pressure model of the corroded pipeline was chosen based on the confidence index which was a statistical comparison for efficiency goodness-of-fit measure between experimental burst of corroded pipes and predicted pressure by burst model of pipe. Finally, effects of the mean and standard deviation of random variables have been investigated using the CHL–RF algorithm for pipeline with time-depended corrosion defect. Results indicate that the CHL–RF algorithm is more robust than the HL–RF scheme and small error is achieved using the probabilistic model for predicting burst of corroded pipe. Failure probability of corroded pipe is more sensitive to defect depth.