A single sensor measurement is always a superposition of waves traveling up and down the tube. The applicability of fiber optic pressure sensors for the transient detection of pipeline leaks is confirmed by laboratory experiments. To the authors' knowledge, this is the first time that in-pipe fiber optic sensor arrays have been used for hydraulic transition-based leak detection.
Details of the in-tube optical sensor array design and laboratory verification of hydraulic transition measurement and leak detection are discussed in the following sections. As mentioned above, the in-tube fiber optic pressure sensor array is a modification of the FBG-based manometric catheter developed by Arkwright et al. This change in strain causes a shift in the reflected wavelength of the FBG that is proportional to the change in ambient pressure [31].
Schematic of the fiber optic pressure sensor showing the arcuate preload applied to the fiber.
Hydraulic transient measurement
Presentation of the experimental system: (a) pipeline configuration; and (b) enlarged view of the placement of the fiber optic pressure sensors (drawings are not to scale and are representative only). The cumulative effective opening [36] of the leak is calculated as 2.9 mm2 from the opening equation. To investigate the frequency response, the Fourier transforms of the hydraulic flow noise signals were calculated and shown in Figure 7 .
Only the transient data measured 20 s after the start of the leak were used in the spectral analysis to avoid the impact of the leak opening operation. The total duration of the data used in the frequency analysis was 26 s for each sensor. Single-sided amplitude spectrum of the leak-induced signal: (a) measured by FBG3; and (b) measured by the piezoresistive pressure transducer (PT).
In water distribution systems, the maximum observable frequency of the transient pressure waves is usually less than 100 Hz. A stepwise pressure wave was generated by abruptly closing the valve, and the pressure responses of the pipeline were measured by the fiber optic sensor array and the piezoresistive pressure transducer. The calibration of the fiber optic sensors was checked by matching the peak and valley measurements of the FBG and PT sensors.
Using this calibration method, the sensitivities of the FBG sensors were slightly less than those indicated by the steady-state measurements (Table 2). Transient response of the experimental pipeline system to a stepwise incident pressure wave: (a) when there is no leak and (b) when a leak is present. This small increase is believed to be the effect of the in-pipe fiber optic cable.
Because the inserted portion of the cable was only about 3.1 m [Figure 4(b)], the impact is localized and has a very limited impact on leak detection as discussed later. Note that the measurements shown in Figure 8(b) appear to be noisier due to hydraulic noise caused by leakage.
Transient-based leak detection and localisation
Consequently, the measured wave reflection can be caused by a defect located either on the top or bottom side of the transient generator, or it can even be the combined result of defects on both sides. The data shown in Figure 10 are difficult to interpret directly due to interference from hydraulic and measurement noise. The time shift of the data is performed according to the data recorded by FBG2.
FBG2 is the leftmost sensor among FBG2 to FBG5 [Figure 4(b)], negatively shifting the time bases of FBG3, FBG4 and FBG5 (i.e. subtracting the corresponding wave travel time from the original time base) will adjust waves coming from the left page. The principle behind this is that waves coming from the left side of the sensor array first arrive at FBG2, then FBG3, FBG4 and FBG5 with increasing time delays (wave travel times). On the other hand, positively shifting the time bases (ie adding the corresponding wave travel time on top of the original time base) will set up waves coming from the right side.
In Figure 11(a), where a negative time shift was applied, the rising edge due to the closure of the transient generator and the reflection from the closed ball valve are aligned, indicating the left-right direction of the waves. Conversely, in Figure 11(b) where a positive time shift was applied, the pressure drops due to the reflection from the leak are aligned, indicating a right-left traveling wave. Compared to the original pressure traces in Figure 10, the filtering of the moving average significantly improves the clarity of the leakage-induced reflections.
It is expected that the same filtering approach will also be effective in field pipelines, as the main purpose of the filtering is to suppress the intrinsic measurement noise (which has a characteristic similar to white noise). The impact of this local wave speed change on the leak localization is negligible in field pipelines where the length of the normal pipe section is much longer than that in the laboratory pipeline. Overall, the occurrence of the pressure drop when the leak is opened and the agreement between the calculated and measured distance between FBG2 and the leak location validated the use of the fiber optic detection array for leak detection and localization.
Discussion
Comparing the measurements from repeated tests (or averaging the measurements from several tests) can improve the signal-to-noise ratio. The time-shifting technique as presented helps to identify the directional information of main wave reflections. As the sensor location can be easily changed along the pipe by pulling or pushing the cable, it enables multiple measurements at different (almost arbitrary) locations, providing much more information than the conventional measurement approach that requires a limited number of sensors at specific locations has. .
Since leaks are at fixed locations, the arrival time of the reflection caused by the leak will change as the sensor location changes, but the reflected pattern is predictable (related to distance and wave speed) and can be used to distinguish the leak signal from the background noise. Increased complexity of the measured signals is expected in aging water networks where multiple leaks may occur and deteriorated pipe sections (eg clogged and corroded pipe sections) also cause wave reflection. These signals from fixed functions are repeatable (i.e. cannot be excluded from averaging), so identifying a leaky signal is difficult.
Understanding the characteristics of the different types of reflections is essential: for a positive-step wave, the reflection from a leak is a negative step [as shown in Fig. 9(a)], and multiple leaks will produce multiple negative-step reflections at different times. ; a rebound from a discrete block is a positive step [6]; rebound from an extended blockage (or section with a thicker wall) is a positive step followed by a negative step [5, 17]; and the rebound from a section with extensive corrosion (or a section with thinner walls) is a negative step followed by a positive step [10]. For example, a leak will introduce a sinusoidal pattern at the resonant or anti-resonant frequency responses [42], while a section of pipe with a thicker or thinner wall will slightly shift the resonant frequencies but not introduce any sinusoidal pattern [5]. . When the pipe condition is complex, however, the time-lapse technique as presented may not be effective because the pressure traces will be complex.
In this case, the use of a wavelet separation technique can be helpful, as this technique extracts directed waves traveling up and down the pipe using pressure measurements from two sensors in close proximity [43]. Wave separation can reduce signal complexity and allow reflection-based analysis to focus on one side of the pipe at a time. This technique remains to be tested on FBG measurements and will be the subject of future research.
Conclusions
Analysis of measurement data from four fiber optic sensors confirmed the existence of the leak and its location. The results have confirmed that the in-tube fiber optic pressure sensor array can significantly improve pipeline fault detection based on hydraulic transients by providing wave direction information.
Acknowledgements
Simpson, In situ non-invasive condition assessment for cement mortar-lined metallic pipelines by time-domain fluid transient analysis, Struct. Beck, Instantaneous phase and frequency for detecting leaks and characteristics in a pipeline system, Struct. Ramos, Case studies of leak detection and location in water piping systems by inverse transient analysis, J.
Kim, Field study on non-invasive and non-destructive condition assessment for asbestos cement pipelines by time domain fluid transition analysis, Struct. Dinning, A fiber optic catheter for simultaneous measurement of longitudinal and circumferential muscle activity in the gastrointestinal tract, J. Grattan, Fiber optic based sensor technology for humidity and moisture measurement: Review of recent progress, Metg.
Yao, A fiber-optic method for assessing the health of pipelines subjected to earthquake-induced ground motion, Struct. Chen, An embedded fiber optic distributed sensor for pipeline leak detection and localization, Sens. Krak, Utility of a new distributed fiber optic acoustic sensor for leak detection, Distributed and Multiplexed Fiber Optic Sensors II, Society of Photo Optical, Boston, MA, 1992, p.
Lemckert, Fiber optic pressure sensing arrays for monitoring horizontal and vertical pressures generated by water waves, IEEE Sensors J. Simpson, Experimental verification of pipeline frequency response extraction and leak detection using the inverse repetition signal, J. Lambert, Frequency response diagram for pipeline leak detection: comparison of odd and even harmonics, J.