LAUSR

Abstract Pipelines are used to carry fluids across long distances. Given the costly and hazardous nature of some of these fluids, timely and accurate inspection for damages is imperative to prevent harmful financial and environmental consequences. In the previous research, a fiber optic based hoop strain sensor is developed and reported to accomplish the goal of pipeline corrosion and leakage monitoring. The sensors form a foundation upon which advanced damaged detection algorithms can be carried out. In this paper, the application of distributed hoop strain sensing information combined with support vector regression (SVR) is demonstrated to pinpoint leakage position on a long-distance pressurized pipeline. The SVR parameters were further optimized by a genetic algorithm (GA) in order to improve accuracy. The resulting leakage detection system had a mean squared error as low as 0.076 when no noise was present. The effect of noise (approximated by Gaussian white noise) was studied in a simulation, showing an effective 5% error for a 55 km simulated pipeline. Results also showed that the use of more sensors corresponded to heightened robustness towards different noise levels.