Session: 03-03-01 Assessment Model Enhancements - Cracking

Paper Number: 87207

87207 - Development of a Near-Neutral Ph Stress Corrosion Cracking Growth Model for Pipelines Using Machine Learning Algorithms 

Near-neutral pH stress corrosion cracking (NNpHSCC) is one of the leading causes of failure for buried pipelines.  NNpHSCC defects grow over time and compromise the pipeline’s pressure containment capacity, i.e. burst capacity.  Characterizing NNpHSCC growth in the through wall thickness direction is significant to the safety of operating pipeline systems but still remains a challenging task in the industry.  A recent study on existing NNpHSCC growth models in the literature suggests that improvements on these models are required or new growth models need to be developed such that adequately accurate predictions of the NNpHSCC growth rates can be achieved.  In this study, an NNpHSCC growth model for buried pipelines is developed based on experimental data obtained from full-scale tests conducted at the CanmetMATERIALS of Natural Resources Canada of pipe specimens that are in contact with near-neutral pH environment and subjected to cyclic internal pressures, which well simulate actual field conditions.  The crack growth per unit time, da/dt, rather than the crack growth per loading cycle, da/dN, is predicted for better practicality.  Due to the unavailability of consistently recorded chemical data at the testing environment, the maximum stress intensity factor K_max and stress intensity factor range ΔK are used for the prediction.  Four machine learning algorithms, namely the random forest (RF), extreme gradient boosting (XGB), Gaussian process regression (GPR) and artificial neural network (ANN), are employed to establish the connections between the input features K_max and ΔK, and the output da/dt, utilizing the open-source platform PYTHON.  The machine learning models are well trained through hyperparameter tuning and k-fold cross validation to improve model robustness, prevent overfitting and generate the best outputs.  Model performances are validated and compared on the same independent test dataset.  An illustration of the growth model application to internal pressure records obtained from a real pipeline is also presented.  The new crack growth model will facilitate the pipeline integrity management practice with respect to NNpHSCC.

Presenting Author: Haotian Sun The University of Western Ontario

Presenting Author Biography: Haotian Sun is a Ph.D. Candidate in the Department of Civil and Environmental Engineering at The University of Western Ontario.