Electronic Thesis and Dissertation Repository

Thesis Format

Integrated Article

Degree

Master of Engineering Science

Program

Civil and Environmental Engineering

Supervisor

Zhou, Wenxing

Abstract

One of the main causes of pipeline failures is corrosion, which leads to a localized loss of the pipe wall thickness and hence compromises the capacity of the pipeline. Composite repair is a method to rehabilitate corroded pipelines. Design codes such as ASME PCC-2 are commonly used to design the repair thickness. As the predictive accuracy to determine the burst capacity of a composite repaired pipeline with existing models is generally poor, the main objective of the present thesis is to provide insights on the parameters that affect the burst capacity and propose improvements to the prediction of burst capacity of composite repaired corroded pipelines.

The first study investigates the influence of defect width on the burst capacity of corroded pipelines repaired with composite materials. Parametric finite element analysis is conducted to assess the burst capacities of composite-repaired corroded pipelines containing localized and full-circumferential corrosion defects. The analysis reveals that composite-repaired pipes with localized defects exhibit considerably lower burst capacities compared to those with full-circumferential defects. Furthermore, the burst capacity model derived from the ASME PCC-2 code´s design equation is deemed non-conservative for composite-repaired pipes with localized defects based on the parametric finite element analyses. To address this issue, an empirical equation for the defect width correction factor is developed, demonstrating its high effectiveness in enhancing the predictive accuracy of the PCC-2 burst capacity model.

The second study addresses the limitations of existing prediction models that fail to account for the complexities of composite materials and the impact of defect dimensions on the burst capacity of composite-repaired corroded pipelines. An improved equation is proposed to enhance the ASME PCC-2 design code's ability to predict the burst capacity of such pipelines by incorporating a correction term into the model. To determine the correction term, a machine learning model called Gaussian process regression (GPR) is employed, utilizing seven input variables. The finite element parametric analysis data is divided into a training set (70%) and a test set (30%) using a stratified random approach to ensure equal representation of failure modes in both sets. The GPR model is constructed using a squared exponential kernel and a zero-mean function based on the training set data. The performance of the model is then evaluated using the test set. Results indicate that the proposed model accurately predicts the burst pressure of the validation set with a mean absolute error of 4.0%. In comparison, the mean absolute error for the ASME PCC-2 model was 48%, demonstrating a significant improvement in accuracy.

Summary for Lay Audience

Pipeline failures often occur due to corrosion, which weakens the pipe and reduces its capacity. Composite repair is a method used to fix corroded pipelines. However, current models for predicting the strength of repaired pipelines are not very accurate. This thesis aims to improve the prediction of burst capacity (maximum pressure it can withstand) for composite-repaired corroded pipelines.

The first study focuses on the width of the corrosion defect. It is found that pipes with localized defects repaired using composites have lower burst capacities compared to those with full-circumferential defects. The existing design equation from the ASME PCC-2 code is found to be unreliable for localized defects. To address this, a new equation is developed, correcting for the defect width and significantly improving the accuracy of burst capacity prediction.

The second study addresses the limitations of existing prediction models that don't consider the complexities of composite materials and the impact of defect dimensions. An improved equation is proposed by incorporating a correction term into the ASME PCC-2 model. To determine this correction term, a machine learning model called Gaussian process regression (GPR) is used. The GPR model is trained using data from finite element analysis and seven input variables. The model's performance is evaluated, and the results show that it accurately predicts the burst pressure of the pipeline with a small error of 4.0%. In contrast, the ASME PCC-2 model has a much larger error of 48%, indicating a significant improvement in accuracy with the proposed model.

Creative Commons License

Creative Commons Attribution 4.0 License
This work is licensed under a Creative Commons Attribution 4.0 License.

Download

Share

COinS