Conclusions

All the work reported here, uses transient analysis of water hammer waves for leak detection. Water hammer occurs when a valve is closed rapidly and pressure wave is generated through the system. When this wave encounters a leak or other features such as a junction, a valve or a hydrant, additional reflection points are produced and can be captured using a single pressure transducer. All features are located by measuring the period of time which the pressure wave takes to propagate from the measuring point to the features and back. The main conclusions drawn from the research can be summarised as follows:

For the simulation test using SUNAS software, the instantaneous frequency and phase of the signal could be found using the HT and the HHT analysis. Using both these methods to define these properties in the pressure wave history, the locations of the discontinuities of the simple simulated pipeline network were found. It was also discovered that the size of the resistances affects the values of the instantaneous properties. Change in the resistance location showed the pattern of change in the peaks locations. In the HHT analysis, the unwanted ripple also present at the low frequency part of the spectrum which is seen in the EMD method, but not in the Hilbert transform.

The leak detection research has been extended to transient test in the laboratory. The experimental setup in the department of civil engineering is used to capture the pressure transient signals in the pipeline with and without a leak. Both simulated and experimental pressure signals were used to evaluate the performance of these transforms. The proposed technique predicts satisfactorily the location in simple pipeline systems as it can detect and locate the leak up to error of 0.5%. Moreover, experimental application of the proposed technique has shown great potential to locate a leak. From the three different location of leak test conducted the error was in the range of 0.44% to 1.96%. Meanwhile for the outlet of pipes, the error is about 0.5% to 1.49%. Therefore, analysis of the experimental tests corroborates the simulated test results when the HT and the HHT analysis are used. The analysis using HT and HHT can be considered the first application to leak detection and location. In addition, this technique has many advantages over the presently available techniques, such as simplicity of application, ease to use and thus can be considered economical.

167 In the experimental work, a simple device was developed that attaches onto a UK standard hydrant in order to produce the pressure transient in the test rig. Fitted to the hydrant end cap, there was also a pressure transducer which was in contact with the water in the hydrant. This device provides both pressure creation and acquires system response data from a single, standard access point. Turning off the power to the solenoid valve causes it to shut. This suddenly stops the water in the copper pipe creating a water hammer pulse that travels down the pipe, into the hydrant and hence into the main pipe.

Furthermore, the research was extended to present a comparative study of instantaneous frequency analysis techniques based on pressure transients recorded within a live water distribution network. It has shown that EMD is an effective method to remove the noise that can mask the signal in many real cases. The innovation of this technique is the analysis of Instantaneous Frequency (IF). The instantaneous frequency of the filtered signals are analysed using the Hilbert Transform (HT), the Normalized Hilbert transform (NHT), Direct Quadrature (DQ), Teager Energy Operator (TEO) and Cepstrum.

In the first test at Esholt training ground, it can be seen that almost methods can identify both leaks with an acceptable of error (within -0.56% to 3.83%). The HT method can capture most of the features but at the same time produces unnecessarily spikes. The improvement of HT called NHT and DQ are able to determine most of the features within a reasonable accuracy and clearer compared to HT. Meanwhile, TEO method is also very sensitive to the presence of the noise which will affect the accuracy of the analysis. The combination of EMD and Cepstrum gives the worst. In the second and third test are presented from Yorkshire Water’s distribution network, Sheffield. In the second test, all the proposed method capable of detect and locate the leak within acceptable of error. Using the test from first hydrant, the error was 0.03% to 3.21%. Meanwhile, the test from second hydrant the error was in the range of -4.48% to 0.15%. Moreover, in the third test, it is found that all methods are able to determine the occurrence of leak with an acceptable error (within 0.55% to 1.08%). Additionally, the NHT and DQ are able to identify most of the features with the missing feature by HT has been recovered by both NHT and DQ. As a comparison, the results produced are also compared with the method proposed by Taghvaie. The result is shown that most of the features are detected but at the same time it

168 also produces irrelevant reflections, which may be caused by the noise present in the filtered data using orthogonal wavelet transform (OWT).

The proposed method has shown that features and leak points along a pipeline can be determined by filtered signal via EMD and analysis of instantaneous frequency, with only a small error in distance. The NHT and DQ methods perform the best while the rest of the instantaneous frequency methods also produce acceptable results.