- 21 views
- 45 downloads
Novel Artificial Neural Network Model for Predicting Failure Pressure of Thin-walled Pipes Containing Axially-oriented Surface Cracking Subjected to Internal Pressure
-
- Author / Creator
- Zhang, Xinfang
-
Axial cracking is a major integrity threat for oil and natural gas transmission pipelines in Canada since its presence can lead to detrimental oil leaks or ruptures, resulting in pipeline incidents, and eventually causing severe damage to the property and the environment. Therefore, periodically monitoring the conditions of transmission pipelines during their service life becomes paramount, and predicting their failures is also essential. Different analytical models are available for predicting the burst capacity of pipelines containing axial external cracks, with the CorLASTM model being the most widely used. The accuracy of a predictive model can be objectively measured by comparing the test reported failure pressure with the one predicted by the model. However, recent studies have shown that the CorLASTM model has a slightly conservative mean of test-to-predicted failure pressure ratio and a considerably high coefficient of variation (CoV; standard deviation divided by the mean). The primary aim of this research is to develop a predictive model capable of accurately predicting the failure pressure of pipelines with axial cracks using artificial neural network (ANN) based on datasets generated from the extended finite element method (XFEM) simulations.
The first part of the study examines the accuracy of the latest version of the CorLASTM model using experimental data collected from the literature. A comprehensive reliability-based assessment of cracked pipelines is performed based on the CorLASTM model. The effects of several factors such as pipe grade, pipe dimension, crack size, and CorLASTM model error on the probability of failure (PoFs) are also investigated. The results indicate a significant influence of model error poses on the PoFs.
The second part of the study assesses the effectiveness of the XFEM coupled with the cohesive segment modelling approach implemented in the finite element software ABAQUS for evaluating cracked pipelines. Within this modelling approach, failure is governed by two damage properties. XFEM models are calibrated and validated based on more than 100 full-scale burst test data, from which, a correlation between the fracture toughness and XFEM damage properties (i.e., maximum principal strain and fracture energy) is established. In addition, the XFEM predictions are compared with the ones computed by CorLASTM. The comparison shows that XFEM results in more accurate predictions than CorLASTM.
The third part of the study presents the development of an ANN model for pipelines with axial cracks subjected to internal pressure only. Given that the performance of an ANN model is highly rely on the accuracy of the input data, XFEM models are validated against more than 100 full-scale burst tests before using their outputs to train the ANN model. Parametric studies are conducted in ABAQUS to examine the effects of pipe and crack sizes on the failure pressure. Based on XFEM parametric analyses, an ANN model is developed using the open-source libraries Scikit-learn and TensorFlow in PYTHON. The trained ANN model is further validated using 25 full-scale burst tests data reported from the open literature. Results of ANN predictions are in good agreement with the experimental data with the mean of 1.01 and a low CoV of the test-to-predicted failure pressure ratio at 4.76%, implying that the ANN model is unbiased within the trained range and can serve as a reliable alternative for evaluating cracked pipelines. -
- Graduation date
- Spring 2024
-
- Type of Item
- Thesis
-
- Degree
- Doctor of Philosophy
-
- License
- This thesis is made available by the University of Alberta Libraries with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.