Pipelines are a safe and economic means for transporting oil and natural gas with the number of failures, defined as a loss of product, being relatively low compared to other means of transportation. However, it should be pointed out that even a single major failure, such as a rupture, will have a significant financial and environmental impact.

The focus of this thesis is the assessment of corrosion defects in pipelines with the goal of providing a more complete understanding of the failure of these defects and addressing the conservatism in the currently accepted assessment procedures. At present, these procedures do not take advantage of the detailed corrosion measurements available from high-resolution inspection devices and the plastic material behaviour, which has a significant effect on the failure of corrosion defects, is characterized with a simple flow stress. The currently accepted procedures were developed from experimental burst tests on machined defects with simple geometries but have not been investigated with complex natural corrosion defects.

Two forms of simply shaped defects, long grooves and single pits, were investigated in detail to develop a fundamental understanding of corrosion defect failure. A solution to predict the failure pressure of a long groove was developed and agrees with Finite Element analyses and experimental test results. In addition, a solution to predict the failure pressure of uncorroded pipe was developed since this is an upper bound for the failure pressure of a corrosion defect. Single pit defects were investigated in detail using the Finite Element method. The effect of element mesh density, defect circumferential dimension, varying material properties and defect interaction on the defect failure pressure were considered. In addition, the validity of the Folias factor used in RSTRENG was investigated and found to be inconsistent with the Finite Element results.

Burst test results of 40 pipe sections removed from operating pipelines due to the presence of corrosion defects are presented in an experimental database. This database was used to investigate the accuracy of the currently accepted assessment procedures. The RSTRENG procedure was found to be the most accurate due to the consideration of the actual corrosion geometry. Other proposed methods based on the total defect length and maximum depth were also considered but were not as accurate as RSTRENG. Twenty-five of the defects in the experimental database were analyzed using three-dimensional elastic-plastic Finite Element analysis. When accurate material properties and defect measurements are available this is the most accurate method of assessment. However, use of the Finite Element method requires expertise in interpretting the results and inaccurate corrosion measurements can significantly reduce the accuracy of the failure pressure predictions.

A new model for predicting the failure pressure of corrosion defects is proposed. This model uses the elastic-plastic material properties, as determined from uniaxial tensile tests, and corrosion measurements in the same form as the currently accepted RSTRENG procedure to predict the defect failure pressure. This procedure is based on a Weighted Depth Difference model and uses the long groove and plain pipe failure pressures as lower and upper bounds respectively for the defect failure pressure. The actual defect failure pressure is then determined by considering each point within the corrosion defect and evaluating the effect of the adjacent material loss through the proposed weighting scheme. This method is iterative in nature and applied with a computer program called Corroded Pipe Strength (CPS). When used to evaluate the corrosion defects in the experimental database, this method provided the most accurate failure pressure predictions. This solution was developed from fundamentals and utilizes the actual material properties so that it is applicable to other materials not considered in the experimental database. In addition the Weighted Depth Difference model identifies the predicted failure location.

A statistical model, based on the experimental database, is presented for the various assessment procedures considered and a three-level assessment procedure is proposed when detailed material properties are available. This procedure consists of the solution for a long groove, the Weighted Depth Difference model and three-dimensional Finite Element analysis. When detailed material properties are not available, an alternate two-level assessment procedure, which makes use of the currently accepted assessment procedures, is proposed.