Electronic Thesis and Dissertation Repository


Doctor of Philosophy


Civil and Environmental Engineering


Dr. Wenxing Zhou


One of the major threats to the oil and gas transmission pipeline integrity is metal-loss corrosion. Pipeline operators periodically inspect the size of the metal loss corrosion in a pipeline using in-line inspection (ILI) tools to avoid pipe failure which may lead to severe consequences. To predict pipe failure efficiently, reliability-based corrosion management program is gaining popularity as it effectively incorporates all the uncertainties involved in the pipe failure prediction. The focus of the research reported in this thesis is to investigate the unaddressed issues in the reliability-based corrosion assessment to assist in better predicting pipe failure.

First, a methodology is proposed to facilitate the use of RSTRENG (Remaining Strength of Corroded Pipe) and CSA (Canadian standards association) burst pressure capacity models in reliability-based failure prediction of pipelines. Use of RSTRENG and CSA models require the detail geometric information of a corrosion defect, which may not be available in the ILI reports. To facilitate the use of CSA and RSTRENG models in the reliability analysis, probabilistic characteristics of parameters that relate the detailed defect geometry to its simplified characterizing parameters was derived by using the high-resolution geometric data for a large set of external metal-loss corrosion defects identified on an in-service pipeline in Alberta, Canada.

Next, a complete framework is proposed to quantify the measurement error associated with the ILI measured corrosion defect length, effective length, and effective depth of oil and gas pipelines. A relatively large set of ILI-reported and field-measured defect data is collected from different in-service pipelines in Canada and used to develop the measurement error models. The proposed measurement error models associated with the ILI reported corrosion defect length, effective length, and effective depth is the weighted average of the measurement errors of the corresponding Type I and Type II defects and the weighted factor is the likelihood of ILI reported corrosion defect being a Type I defect (without cluster error) or a Type II defect (with clustering error). A log-logistic model is proposed to quantify the weighted factor. The application of the proposed measurement error models is demonstrated by evaluating probability of failure of a real corroded pipe joint through system reliability analysis.