Session: 07-04-01 Crack Management

Paper Number: 86906

86906 - Probabilistic Analysis Applied to the Risk of Scc Failure 

This paper will discuss a model developed and applied to evaluate the probability of Stress Corrosion Cracking (SCC) failure in a large gas pipeline system spanning approximately 5,600 miles.  A machine learning algorithm (neural network) was applied to the system, which has experienced nearly 500 prior instances of SCC.  Subject matter experts were interviewed to help identify key system factors that contributed to the prevalence of SCC and these factors were incorporated in the neural network algorithm. Key factors such as coating type, vintage, operating stress as a percentage of SMYS, distance to compressor station, and seam type were evaluated in the model for correlation with SCC occurrence.  A Bayesian analysis was applied to ensure the model aligned with the prevalence of SCC.  A Probabilistic Fracture Mechanics (PFM) model was then applied to relate the probability of SCC existing to the probability of rupture. 

Statistical distributions of SCC features were developed specific to each pipeline in the system based on the pipeline operator’s crack registry (database of pipeline segments with SCC along with characterization of the SCC features).  These were combined with distributions for key material properties (such as strength and toughness).  A series of Monte Carlo simulations using these distributions and applying a Bathtub Curve crack growth model to evaluate whether a segment will rupture with each simulation as a function of time.  The model was used to evaluate susceptibility to failure at different pressure levels and flow directions for a subset of pipe segments. 

The neural network and PFM results were then combined to produce a joint probability (likelihood of SCC existing times the probability of rupture given that SCC exists).  These results were overlaid with the pipeline operator’s decision-making risk matrix used to determine whether a designated pipeline segment was safe to operate at a designated pressure level.  Different assessment methods and schedules were then assigned to various pipeline segments based on these results.

Presenting Author: Owen Malinowski Structural Integrity Associates

Presenting Author Biography: Mr. Malinowski is an engineer with ten years of combined experience in the power generation and natural gas pipeline industry. Owen has significant expertise in developing software for data acquisition and control systems, as well as web-based data aggregation and analysis software. Owen also has significant expertise in nondestructive evaluation and has led/supported several research and development projects, as well as field inspection projects, in this capacity. Owen has his Bachelor of Science degree in Engineering Science and Master of Science degree in Engineering Mechanics from The Pennsylvania State University.