Session: 03-03-05 Feature Assessment Case Studies Cracking Session II

Paper Number: 133509

133509 - Utilizing Partial Safety Factors to Better Manage Stress Corrosion Cracking on Pipelines 

Abstract: 

Fitness for service assessments of stress corrosion cracking (SCC) crack fields (CF) must account for in-line inspection (ILI) tool measurement error. In deterministic assessments, the vendor-specified tool tolerance is typically added to both the CF depth and overall length. This methodology is very conservative and leads to a large number of false positives (resource outliers) when evaluated in the field. The resource outliers are defined as cases where the safety factor on the failure pressure (Pf) assessed using the ILI measurements is below 1.25 (Pf / MOP), but the field safety factor did not indicate a need for repair at this level.

In the present study, a semi-probabilistic model was calibrated to match lower percentiles (0.1st and 0.5th percentiles) drawn from distributions of predicted burst pressures generated using a full probabilistic model that is solved using Monte Carlo simulation. The Monte Carlo simulations account for measurement error in the ILI called wall thickness (WT), crack field depth, and crack field length, using probability distributions. The distributions are calibrated to reflect the difference between in-line inspection (ILI) measurements and field non-destructive examination (NDE) measurements. The calibration for length incorporates the ratio between the total ILI-called field length and the NDE measurement of longest-interacting length.

The semi-probabilistic model approximates the lower percentiles (0.1st and 0.5th percentiles) of the probabilistic burst pressure distributions using a deterministic evaluation of CorLAS, incorporating partial safety factor (PSF) adjustments on the feature’s length and depth.

The PSFs are solved through an optimization which is run against probabilistic results generated for a comprehensive set of feature sizes and pipe attributes. The numerical optimization is done in two steps to ensure that a global optimum solution has been found, first constrained global search is performed using simulated annealing and then steepest descent is used to fine tune to the best solution local to the outcome from the first step.

The performance of the model was evaluated to ensure a proper balance is achieved between the level of conservatism in the predicted failure pressure and the dig efficiency. When evaluated against historical digs, the results showed a 75% reduction in the number of recommended digs that were ultimately found to be false positives (resource outliers). Based on these historical field measurements, the expected probability of a false negative call (near miss) is less than 1 x 10^-6.

Presenting Author: Ryan Stewart Integral Engineering

Presenting Author Biography: Ryan is a UBC Integrated Engineering graduate and E.I.T. at Integral Engineering. He has 5 years of experience in the pipeline industry across design, construction, and operations. In Ryan's current role he focuses on the application of statistical and machine learning methods to risk and reliability assessments of pipelines and related facilities.

Authors:

Ryan Stewart Integral Engineering
Thomas Dessein Integral Engineering
Lyndon Lamborn Lamborn Engineering Inc
Vitaly Vorontsov Enbridge
Evelyn Rawlick Enbridge