Page 152 - Reliability and Maintainability of In service Pipelines

138 Reliability and Maintainability of In-Service Pipelines

While evaluating the probability of failure for a pipeline, it is important that
the various probabilities associated with the violation of the individual limit states
are combined into a system analysis. In Fig. 5.8, the benefit of this approach of
reliability analysis is apparent as the probability of pipeline system failure
exceeds each of the individual limit state probabilities at every point. This is espe-
cially important during the initial 50 years of the graph, which is indicative of a
typical pipeline’s service life. It is evident from Fig. 5.8 that considering each
failure mode individually in the reliability analysis of pipelines can mislead the
reliability analysis, since it ignores the impact of the other limit states involved.
For instance, considering only the leakage limit state, the service life of the pipe-
line is estimated more than 200 years, while it takes only 59 years for the system
failure to happen (assuming acceptable failure probability of 0.05). Additionally,
it can be seen that, the system failure is highly affected by wall thrust and flexural
failure.
Using the results of Fig. 5.8, one can determine the time of failure (t c ) based
on accepted probability of failure (P a ). Ascertainment of t c is critical for design
engineers and asset managers, as they must decide whether to repair or replace
the structure in accordance with their budget allocation and risk analysis
(Mahmoodian and Li, 2016a,b). For example, using the criteria of system failure,
it can be obtained from Fig. 5.8 that, for an acceptable estimated probability of
failure of 5%, the time of pipeline system failure is 59 years. If there is no repair
or maintenance action during the service period of 59 years for the pipe, the pipe
will not be serviceable after 59 years.

5.2.2.1 Sensitivity Analysis

The effect of the random variables on the failure of the pipeline can be analyzed
using a parametric sensitivity analysis. As previously mentioned, these variables
are involved in the corrosion process and thus, the parameters associated with
determining the probability regarding the violation of the limit state functions.
Consequently, determining the variables that influence this process the most, is of
high importance. In order to study the effect of a variable using sensitivity analy-
sis, first a deterministic value is given to the selected variable (i.e., its mean value)
while the rest of the variables remain random. Then an upper and a lower value
are given to the variable to observe the changes in failure probability curves.
The results of this parametric sensitivity analysis for four of the random vari-
ables are presented in Figs. 5.9 5.12. It should be noted that, these
figures represent the estimated probability of failure of the pipeline system and
not individual limit states. For instance, considering an acceptable probability of
failure of 5%, it can be seen that having a wall thickness 1 mm less than that of