Real-time Pipeline Burst Detection and Estimation using the Damping of Fluid Transients

Abstract

Water is a vital human resource, and the corresponding water distribution pipelines are critical to modern cities' infrastructure. Any burst in the water distribution pipelines can cause significant economic losses, interrupt traffic, and result in negative publicity for water utility companies. As a result, pipeline burst detection technologies have been widely explored; however, the applications of some of the previously-explored techniques, in both the time and the frequency domains, require multiple sensors, which leads to a high cost for the sensing system. Additionally, some of the techniques cannot be applied to real-time data monitoring, indicating that the burst cannot be captured in real-time and, thus, the negative effects from the burst cannot be minimized. Consequently, this thesis aims to investigate an approach for real-time pipe burst detection and estimation (i.e., localization and size quantification) in a pipeline using the measured data in the whole frequency range of the transient pressure signals that are measured by one sensor alone. Earlier research has shown that a pipeline leak can cause additional damping of the transient pressure signals. The additional damping exists in burst problems. Additionally, according to the burst damping equation that is developed in this thesis, the harmonic components of the transient signal are damped differently in all the resonance responses in a newly-developed discrete harmonic spectrogram in the frequency domain, due to the burst. Accordingly, the damped resonance responses can be utilized to detect and estimate the burst. This specific analysis makes pipe burst detection and estimation utilizing transient pressure signals possible in the frequency domain. The research proposes a novel technique for pipe burst detection, localization, and size quantification using transient pressure signals in the frequency domain. The damping caused by both the friction and the burst exists in the transients of the pressure head. The technique can analyze the resonance responses of the transient pressure signals caused by the burst measured by one sensor alone, during every fundamental period of the signal, to determine the total damping. The steady friction damping can be calculated, and a developed unsteady friction water hammer model for burst initiation enables the unsteady friction damping caused by the burst to be estimated. Therefore, the burst damping can be calculated by subtracting the steady and unsteady friction damping from the total damping, so that the burst can then be identified, localized, and quantified utilizing the burst damping ratios. Only the damping of the first three resonance harmonics is utilized to minimize the number of possible solutions to the burst location. This technique has been further developed from the perspective of real-time data monitoring. The corresponding algorithm enables the transient pressure signals to be analyzed window by window with a user-defined window length and window gap instead of a fixed value, fundamental period of the signal. The userdefined window gap makes real-time data monitoring possible by letting the window gap equal the reciprocal of the data sampling rate. Additionally, the required data transmission and sampling rates are reduced and only require the Nyquist frequency of the third resonance harmonic of the analyzed signal, smce only the damping of the first three resonance harmonics is required. Further research has been conducted to ease the restriction that only the first three resonance responses of the signal can be utilized. The research has developed an algorithm that enables any sequence of harmonics to be utilized for real-time pipe burst detection and location estimation. The burst location estimation can be fulfilled with acceptable accuracy, and the majority of the incorrect, possible solutions to the burst location can be excluded. In order to discuss and compare the performance of those techniques that use the damping of fluid transients and those techniques that use the frequency response diagram, the specific relationship between these two techniques has been explored. The detailed and exact relationship between these two techniques has been revealed from both a mathematical perspective and via practical utilization. The two techniques have been discussed and compared from the perspectives of input bandwidth, problem type, low sampling rate capability, robustness, and real-time data monitoring capability, thereby providing an insight into the detailed impact of utilizing different techniques for pipeline inspection. The overall contribution of this research is the development of novel techniques for pipe burst detection and estimation in real-time, utilizing any sequence of harmonics in the whole frequency range of the transient pressure signals. To date, the developed techniques have been verified both numerically and experimentally. The techniques have the advantages of being applied to real-time data monitoring with low sampling rates and utilizing one sensor alone.