Aiding Leak Detection in Pipelines Using System Identification - Year two
The ultimate goal of this project is to detect and localize leaks in pipelines in real time. Hifi Engineering has developed
distributed fiber optic sensors. Measurements are obtained at evenly spaced intervals along the pipeline (called
channels). This project aims to develop data processing tools to improve leak detection and localization. Many events
occur along a pipeline whose effects are registered by sensors (trucks driving by, compressors turning on, leaks).
Events are registered in many channels as sounds propagate down the pipeline. We aim to remove the component of
a channel signal caused by events occurring in neighboring channels. This results in a de-cluttered channel signal. We
do this by constructing a data-driven dynamic model of the pipeline acoustics, and use it to filter out events that
occurred in neighboring channels. Subsequently, Hifi Engineering will process the de-cluttered signal to classify each
event (i.e. was the event a truck, leak,...).