Accurate Leak Detection of Upstream Pipelines using Passive and Active Methods utilizing Artificial Intelligence

管道在碳基能源产品的生产和运输中至关重要。然而，管道系统会受到内部和外部腐蚀、裂缝、土壤移动造成的应力以及第三方损坏等威胁。失液事件可能危及人类生命，对环境有害，并影响公众对管道系统的信任。为了解决这些威胁并降低失液事件的可能性，需要进一步发展与管道泄漏检测和结构健康监测相关的技术。多相流是管道泄漏检测的难点之一。然而，多相流中存在的不同流型引入了瞬态扰动，使传感器数据的解释复杂化。此外，管道网络可以跨越数公里的长度。信号衰减可能导致泄漏信号低于传感器噪声底，导致信号变得无法检测。多相流管道需要快速可靠的泄漏监测系统，以避免管道故障造成的破坏事件。卡尔加里大学的团队提出了一种新的泄漏检测方法，通过结合被动和主动泄漏检测技术来表征多相流管道的泄漏。该团队将研究来自不同传感器系统的数据，以了解这个问题。利用这些结果，将使用实验室和现场数据训练机器学习/人工智能分类方案。研究小组将使用现有的多相流设备，该设备包括一个可互换的土壤箱，以便研究外部土壤对信号衰减的影响。一旦对多相流和土壤衰减的影响有了重要的了解，该团队将继续研究泄漏检测系统的现场数据。该项目对环境保护以及对工业赞助商的HQP培训至关重要。

Pipelines are critical in the production and transport of carbon-based energy products. Pipeline systems are, however, subject to threats including internal and external corrosion, cracks, stress due to soil movement, and third-party damage. Loss-of-fluid events can endanger human life and be harmful to the environment, and this impact public trust in pipeline systems. In order to address these threats and reduce the probability of loss-of-fluid events, technology related to pipeline leak detection and structural health monitoring needs to be further advanced. One of the challenges in pipeline leak detection is dealing with multiphase flow. The different flow regimes that exist in multiphase flow, however, introduce transient disturbances that complicate the interpretation of sensor data. Moreover, pipeline networks can span kilometers in length. Signal attenuation may cause the leak signature to fall below the sensor noise floor, resulting in the signature becoming undetectable. There is a need for fast and reliable leak monitoring systems for multiphase flow pipelines to avoid the damaging events that result due to pipeline failure.The University of Calgary team proposes a novel approach in leak detection by combining passive and active leak detection techniques to characterize leaks from a pipeline with multiphase flow. The team will study data from different sensor systems to build an understanding of the problem. Using the results, a Machine Learning/Artificial Intelligence classification scheme will be trained using both laboratory and field data. The research team will use an existing multiphase flow facility that includes an interchangeable soil box, enabling the effect of external soil on signal attenuation to be investigated. Once a significant understanding of the effects of multiphase flow and soil attenuation has been achieved, the team will move on to field data of the leak detection system. The project is critically important for protecting environment as well as training of HQP with the industrial sponsors.