Abstract

ABSTRACT


Vibration-based Damage Detection and Localization in Pipelines Using Data Analysis
Mohammadsajad Salimi
Pipelines are important indicators of modern infrastructure, which allow for easy transportation of significant resources like gas, water, and petroleum over long distances. They often cross very sensitive environments, and any damage to them poses severe and irreversible consequences for marine ecosystems. Also, global warming brings deterioration in infrastructure, which calls for urgent needs for advanced monitoring and maintenance solutions. Their challenges bring about the development of efficient and practical systems to meet the current and future demands of industry. This research focuses on the prediction of pipeline behavior through the vibration signals as a primary method for detection, location, and showing the difference in the extent of damage. Methods like this provide critical details about pipe structural integrity and hence notify early stages of a possible problem well before failure occurs. In this work, pipeline conditions were simulated with numerical modeling using the ANSYS software and then validated with experimental data. Features extracted from sensors were specifically velocity and acceleration. Analysis was done by Principal Component Analysis (PCA), which tries to diffuse data complexity and emphasizes only the most significant variations, where the first principal component carries the most critical information about pipeline conditions and is used as our desirable feature. Independent Component Analysis (ICA) as a method for finding statistically independent components for refining the data is used for detection phase. Application of ICA in this area helps maximize the detection rate for anomalies such as corrosion or structural damage. Then, it combines Mahalanobis distance and K-nearest neighbor methods to accurately localize damage in the pipeline. Initial results using data reveal that the algorithm works well, hence may be applied to real life.