Research on key technologies for early warning-position-angle monitoring and diagnosis of nuclear-grade pipeline leaks

Abstract

Nuclear power plants (NPPs) are complex systems where most accidents stem from equipment failures. When such failures occur, personnel often struggle to identify the root cause accurately, promptly, and effectively, which can lead to worsening conditions and potential radioactive leaks. To reduce the risk of coolant loss accidents, it is crucial for NPPs to develop robust leakage monitoring systems that can control the spread of harmful consequences and improve overall safety. Although researchers have made significant strides in developing online pipeline leak detection systems, challenges remain, particularly in providing sufficient early warning information and accurately diagnosing the exact location of pipeline ruptures. To address these issues, this paper introduces an advanced monitoring and diagnostic method that combines threshold techniques with qualitative trend analysis to offer early warnings of nuclear-grade pipeline leaks. This method establishes a comprehensive "early warning - location - angle" monitoring and diagnostic framework. Specifically, the combined use of threshold methods and qualitative trend analysis enhances monitoring sensitivity, while a novel sensor positioning technique improves the accuracy of detecting abnormal locations, achieving precise localization within 0.275 m. Additionally, a machine learning-based technique is proposed for diagnosing breach angles, achieving a 95 % accuracy rate in monitoring breach angles.