A Study on Optimal Sensor Placement for Detecting Rupture in Pipeline System

Abstract

Pipeline systems of infrastructures like power plants and petrochemical facilities are critical to maintain human daily life. Rupture or damage of pipeline systems can interrupt the reliable operation of the infrastructures and lead human loss. In the recent years, there is increasing interests in the research area of structural health monitoring (SHM) to prevent such undesired accidents. This study presents optimal sensor placement for the detection of pipeline rupture. First, a probabilistic simulation model of pipeline systems is described. Physical uncertainties and measurement error are employed in the simulation model to include the inherent randomness in real world. Then, damage scenarios are determined based on expert knowledge and failure history analysis. For individual damage scenarios, pressures at the nodes of pipeline systems are calculated from the probabilistic simulation model. Second, the performance of sensor networks is calculated with a detectability metric. The detectability metric quantifies the capability of classifying individual damage scenarios with the pressures at nodes where sensors are placed. Finally, optimal sensor nodes are computed with a mixed-Integer nonlinear programming method (genetic algorithm). To demonstrate the validity of optimal sensor placement, a small-scale testbed and a real-scale warship application are presented. The accuracy and speed of the proposed sensor placement method is compared those of exhaustive search and intuitive sensor placement.