Erosion Detection in Potash Pipelines Using Dynamic Pressure Response and Machine Learning
Date
2023-01-26
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
ORCID
Type
Thesis
Degree Level
Masters
Abstract
Erosion in pipelines can result in leakage if not monitored and maintained properly. Hence, it is very important for pipelines transporting slurries to conduct pipeline health condition monitoring continuously. In theory, erosion can be detected by inserting a pressure impulse and measuring the delay time of the reflected wave as the level of erosion and its distance to the pressure source can affect the reflected wave speed. This research is focused on investigating the possibility of this innovative continues pipeline condition monitoring method, which could change the future of pipeline health monitoring techniques, in different flow regimes and in the presence of non-idealities.
The Method of Characteristics is used to simulate pressure responses in the inlet of an arbitrary eroded pipeline. Equations of continuity and momentum are solved for both laminar and turbulent flow using frequency-dependent friction terms from the literature. The impact of non-idealities such as varying temperature, noise and limited bandwidth of pressure transducers and flow sources are studied for the laminar flow. Turbulent flow of potash brine in an eroded pipeline is then used to generate the turbulent dataset. Transient pressure response in pipelines with turbulent flows including all three regions of smooth, transition and fully rough is then compared to the pressure response of the laminar flow.
Machine Learning is used to extract important features in the transient pressure signal of an arbitrary eroded pipeline and learn the relationships between erosion parameters (severity, length, and location) and the reflected pressure wave. In real-world applications such as erosion detection in potash pipelines, inserting and receiving pressure signals can be performed by using pressure transducers at the two ends of each test segment.
Results from this study showed that with the represented continuous condition monitoring technique, high-cost smart pigging inspection can be decreased significantly. This method is able to detect the severity, length and location of an eroded section in pipelines even under non-ideal conditions such as varying temperature, presence of noise and limited bandwidth of the transducers and flow sources for both low and high Reynolds numbers. However, the accuracy of each parameter varies according to the studied non-ideality. Overall, thickness detection has the highest accuracy among all three erosion parameters. This is suitable because we can classify erosion with the represented continuous method and only use smart pigging inspection or other costly techniques when needed.
Description
Keywords
Erosion, Potash, Potash Pipelines, Pipeline Integrity, Pipeline Condition Monitoring, Pipeline defects, Pressure Response, Machine Learning
Citation
Degree
Master of Science (M.Sc.)
Department
Mechanical Engineering
Program
Mechanical Engineering
Advisor
Wiens, Travis