Unpiggable Oil and Gas Pipeline Condition Forecasting Models

Abstract

Although they are the safest method of transporting oil and gas, pipelines are still subject to different degrees of failure and degradation. It is therefore important to efficiently monitor oil and gas pipelines to optimize their operations and to reduce their failures to an acceptable safety limit. Several models have recently been developed to predict oil and gas pipeline failures and conditions. However, most of these models were limited to the use of corrosion features as the sole factor in assessing pipeline condition. In addition, the use of internal corrosion features in the condition assessment requires the pipe to be piggable, which is not always the case. Modifying pipelines with pigging facilities is not always an easy option and can be very costly and time consuming. This paper presents the development of condition forecasting models for unpiggable oil and gas pipelines based on factors other than those related to internal corrosion. In addition, the paper examines the degree of confidence of the unpiggable model by comparing its results to those obtained using piggable models. Unpiggable models can save both time and cost of usual scheduled in-line inspections. Regression analysis, artificial neural network (ANN), and decision tree techniques were used to develop the models based on historical inspection data of existing pipelines in Qatar. All necessary statistical diagnoses have shown sound results for the developed models. When they were validated, the models showed robustness with a satisfactory average validity percentage.