PSI - Issue 64

S.W. Jacobsz et al. / Procedia Structural Integrity 64 (2024) 1657–1664 Author name / Structural Integrity Procedia 00 (2019) 000 – 000

1662

6

72F6CLC - 0.5m

72F6CLC - 1.0m

72F6CLC - 1.5m

-0.002 0 0.002 0.004 0.006 0.008 0.01 0.012

6FTB - 0.5m

6FTB - 1.0m

6FTB - 1.5m

Δ BFS (GHz)

0

60

120

180

240

300

360

420

480

Time (Minutes)

Fig. 4. change in Brillouin Frequency Shift at the three leak locations for the 72F-6C-LC and 6F-TB cables.

4.6. Spatial data processing and presentation Before the first leak test, BFS baseline readings were acquired at two-hourly intervals over the preceding 24-hours. The baseline was taken as the average of the 12 sets of readings thus recorded. After the leak tests, another 24-hour average was acquired. The change in BFS, when subtracting the baseline from the post-leak average, is shown in Figure 5(a) for the 72F-6C-LC and 6F-TB cables. The left-hand side of the figures represent the 72F-6C-LC cable and the right-hand side the 6F-TB cable. (The 6F-TB cable was spliced to the 72F-6C-LC cable as shown in Figure 3.) The BFS profile is noisy, especially in the case of the tight buffered 6F-TB cable, potentially obscuring the ability to detect leaks, as sharp peaks may be mistaken for a leak by a leak detection algorithm. A simple algorithm to clean the data and allow leaks to be identified was devised through trial and error. A 25-point centered moving average was fitted to the data and the baseline prior to baseline subtraction. The improvement thus attained is evident from the black curve in Figure 5(b) showing a reduction in random spikes in the raw relative profile. The three spikes in black on the right-hand part of the profile represent the leak tests at leak locations 1, 4 and 7 respectively. The  BFS profile was subsequently baselined to zero by subtracting a 601-point moving average. The improvement is shown in black in Figure 5(c). As a last step, the cleaned profile was squared, the result of which is presented in Figure 5(d). The peaks occurring at 200m, 250m and 300m are very distinct from the rest of the profile and illustrate the success of the 6F-TB cable at detecting leaks. On the other hand, the 72F-6C-LC cable was only able to detect one leak, in this instance at 60m. (Leaks occurred at 10m, 60m and 110m.) Furthermore, a false leak was detected by the 72F-6C LC cable at 50m. The results indicate how, despite providing an initially more noisy BFS profile, TB cables are more efficient at leak detection than LC cables as they are able to respond to both thermal and mechanical strains. The volumes of water discharged during the leak tests were very small compared to the volumes of water that may be lost from actual pipe leaks. Actual leaks are unlikely to stop and will continue to influence the ground and therefore the leak detection system and are therefore be expected to be more prominent than the small leaks studied here. The fiber optic cables most sensitive to leaks were found to be the more flexible TB cables. Sensitivity reduced with increasing cable stiffness. The LC cables were significantly less sensitive to leaks compared to TB cables. The cables studied, in sequence of most to least sensitive, were: 2F-TB, 6F-TB, BRUsens-TB, 72F-6C-LC, 4F-DC-LC. 4.7. Repeatability of detection The first leak tests on 24 March demonstrated the performance of fiber optic cables to detect leaks. During this first leak test, the initially unsaturated soil in the trench was wetted up for the first time, undergoing significant deformation due to processes explained in Section 2. To assess whether the effectiveness of the system may reduce during subsequent leaks, repeat leak tests were carried out at previously tested locations. The results of both leak tests relative to the 23 March baseline are presented in Figure 6. The second test resulted is dated 2 April. Figure 6 shows that the sensitivity of the system had in fact increased after the first leak test. It is believed that this was due to the first inundation episode causing improved bedding of the soil in the trench around the cables. As moisture would have spread in the soil after the first leak test, negative pore pressure would have reestablished. These negative pore pressures would have been dissipated during the second leak test, again resulting in significant soil deformation, clearly detectable from the BFS records, even using the 72F-6C-LC cable (chainages 10m and 60m).