Sincere gratitude is also expressed to the members of the advisory committee at the postgraduate school of the University of Ulsan, who are Prof. MLPNN locator.
Introduction
- Research background
- Review of leakage detection and localization
- Leak detection method using the cable
- Leak detection method using the fiber sensor
- Mass and volume balance method
- Acoustic wave method
- Negative pressure wave method
- Transient model method
- Metrics for leak detection methods
- Research status of AE technology
- AE technology features
- AE detection principle
- Main contents and organization of this dissertation
This method is based on the mass and volume balance of the liquid in the pipeline [30]. Finally, the characteristics of the AE source can be evaluated through the interpretation of the data.
Propagation characteristics of elastic waves on pipes
- Acoustic impedance
- The basic concept of guided waves
- The basic theory of guided wave in a pipe
- Dispersion characteristics of guided wave in a pipe
- Dispersion curves of guided wave in a pipe
- Influence of the wall thickness and diameter on the group speed
- Wave attenuation
- Diffusion attenuation
- Structural scattering
- Energy absorption
- Theoretical analysis of the attenuation of leak-induced signals
- Conclusions
The dispersion curves of the group velocity of a guided wave in a pipe can be determined by the inner diameter, thickness, modulus of elasticity, Poisson's ratio, and wall density of the pipe [58]. For the torsional modal group velocity in Figure 2.5b, the dispersion curves coincide, so changing a does not affect its dispersion characteristics. For the torsional modal group velocity in Figure 2.6b, the dispersion curves coincide, so changing h does not affect its dispersion characteristics.
When elastic waves propagate in the medium, the energy of the wave decreases as the propagation distance increases, which is called the attenuation of elastic waves. For example, the attenuation factor α can be calculated from the slope of the attenuation and frequency. The AE signal energy is proportional to the integral of the square of the signal amplitude [64] and can be written as.
The influence of wall thickness and diameter on the dispersion curve of the guided wave in a tube is given in this chapter.
Cross-correlation method for leakage location
- Introduction
- Cross-correlation method for leakage detection
- Cross-correlating continuous signals in the time domain
- Cross-correlating continuous signals in the frequency domain
- CCF for discrete leak-induced signals
- GCC method for leakage location
- PHAT estimator
- SCOT estimator
- ML estimator
- The parameters affecting the CCF of leak-induced signals
- Conclusions
The speed of propagation of signals caused by leaks in pipes is a critical factor in locating leaks. The following research work is divided into two parameters: estimation of time delay and propagation speed of signals caused by leakage in pipes. Incorrect selection of the frequency range will result in an error or even no peak. d) Phase: The phase of the signals caused by the leakage is important to determine the time lag between the two measured signals.
Another parameter that affects the location accuracy is the propagation speed of the signals caused by the leakage in the pipes, which can be calculated using a theoretical formula and the physical properties of the pipes or by on-site measurements. The cross-correlation method for leak detection is based on the characteristics of signals caused by leaks in pipes. Due to the limited knowledge of the characteristics of signals caused by leakage in pipes, this method is difficult to apply in pipes.
It has been shown that the PHAT, SCOT and ML estimators designed to pre-filter the leak-induced signals preceding the CCF can sharpen the peak in the CCF.
Wave velocity measurement theory
- Analytical method
- Time-of-flight method
- Cross-correlation method
- Three sensor method
- Phase-frequency method for leakage location
- Modal analysis of guided waves
- Conclusions
To determine the transfer function, the length of a pipe is assumed to be l, and the energy of the leakage-induced signals measured at the position of the leakage is assumed to be p0. It can be concluded that the phase of the product of the Fourier transform of the leakage-induced signals measured by two sensors mounted at two locations varies linearly with the frequency w. The dispersive behavior of guided waves in a gas tube can be analyzed by the theory of guided waves of hollow cylinders [88].
The velocity of the AE signal can be estimated online by the peak frequency of the AE signal in combination with the known group velocity dispersive curve of the fundamental bending mode [9]. G is the shear modulus of the pipe wall material, E is Young's modulus of the pipe wall material. These methods require prior knowledge of the pipe's physical properties and a known leak-induced signal/transient event source.
These methods do not consider the dispersion characteristics of the AE signal, thus a new method based on the dispersion.
Filter based on wavelet transform and empirical mode decomposition
Brief review of the wavelet theory
- The wavelet transform
- The wavelet packet transform
This concept is represented by means of a waltz tree with J levels, where J is the number of repetitions of the basic step. It can be seen that the wave tree (dashed line) is part of the wave packet tree. Each node of the wavelet tree is indexed by a pair of integers (j, k), where j is the corresponding level of decomposition and k is the order of the position of the node at a particular level.
The width of the frequency range, which corresponds to each reconstruction signal Rj,k is Fj≈Fs/2j+1. The frequency range corresponding to each node of the wavelet packet tree is shown in Figure 5.3. It is observed that the natural order of the reconstruction signals Rj,k in a level j is not the same as the increasing frequency order.
This fact requires the reordering of wave packet vectors cj,p, p=0,1,.,2j-1 using a new index of frequency order p at level j.
Brief review of the empirical mode decomposition theory
- Empirical mode decomposition
- The algorithm for signal filtering based on the EMD
This is explained by the fact that low-pass filters can transmit information about the high-frequency content of the signal due to frequency warping caused by downsampling. In the time domain, the mean value of the envelopes is 0, which is determined by local maxima and local minima. It involves the following steps leading to the decomposition of the signal S(t) into its component IMFs:.
The proposed filtering procedure takes into account this property of the decomposition to filter signals. Since the average envelope obtained from the upper and lower envelopes of y1 is different from zero, y1 does not meet the conditions imposed by the definition of IMFs, so it is not an IMF. This procedure is practical, mainly due to the empirical nature of the EMD method, and it can be applied to any signal since the EMD makes no assumption about the input time series.
These IMFs are then soft-thresholded, yielding tIMF1,.., tIMFN, which are thresholded versions of the original components.
Conclusions
For each IMF from 1 to N, a threshold, tn, n=1,..,N, is chosen and soft threshold (defined in Equation 5.13) is applied to individual IMFs as shown in Equation 5.13. The threshold tn is estimated using the following strategy: a window of noise is selected from the original signal and then the boundaries of this window are used to extract a noise region from IMFs. On the other hand, EMD makes no assumptions a priori about the composition of the signal.
Each IMF will be a single periodic oscillator, but otherwise cannot be predicted before it is empirically observed from the signal.
A modified leakage localization method using MLPNN
- Modified GCC location method
- MLPNN classifier
- Signal processing for leakage location
- Conclusions
Since multiplication in the frequency domain corresponds to convolution in the time domain [99], the GCC function between two measured signals can be derived in terms of the time delay τ by substituting Equation 6.4 into Equation 6.3, and expressed as The time delay between two measured signals is equal to the quotient of the offset of the GCC maximum peak and the sampling frequency. Therefore, the modified frequency weighting function can increase the degree of correlation between two measured signals and improve the accuracy of the time delay estimation.
The modified frequency weighting function differs from the PHAT processor function [100] in that the frequency weighting function involves pre-whitening of the leakage-induced signal. In the learning mode, the training sets of the MLPNN contain the input vectors and the output vectors. In the application mode, the trained MLPNN locator can locate the leak by using the energy ratios of the leak-induced signals in the frequency bands with greater power spectral density.
Determine the peak frequency of the leakage-induced signals. Estimate the group velocity of the leakage.
Laboratory experiments
- Experimental setup and data collection
- Characteristics of the frequency domain
- Characteristics of the signal energy ratios
- Characteristics of the time domain
- Leakage location analysis
- Conclusions
The leak-induced signals were generated by the frictional force between the jet of the leaking gas and the simulated leak orifice. To verify the consistency of weakening the pipe effect on two kinds of signals, which are impulse response signals and leakage-induced signals measured by keeping a distance of 14 m between the two sensors, while the distance between sensor 1 and the simulated leakage opening. The difference between the two fitting curves is small, which means that the attenuation effect of the tube on the AE signals is proportional to the propagation distance, regardless of any leakage-induced signals or impulse response signals.
In addition, the group velocity of the leakage-induced signal was determined from the peak frequency of the leakage-induced signal and the dispersion curve of the group velocity of the fundamental bending mode. The relative location errors in Table 7.2 show that the relative location errors increase as the distance between the two sensors increases, which implies that some high-frequency components of the leakage-induced signal fade as the propagation distance increases. The average of the relative location errors are 18% and 4%, respectively, which indicates that the average of the relative location errors obtained by the MLPNN localizer is reduced by 14% compared to that using the CCF location method.
The average of the relative location errors obtained by the MLPNN locator was reduced by 14%.
Conclusions
In addition, these methods do not consider the distribution characteristics of the AE signal, a new method based on the distribution characteristics of the AE signal is proposed. The GCC location method can compensate for the weakening effect of the different propagation paths on the leakage-induced signals, it can also increase the degree of the correlation between two measured signals and improve the accuracy of the time delay estimation. The average of the relative location errors obtained by the MLPNN locator was reduced by 14% compared to that using the CCF location method.
Geiger, "PIPELINE LEAK DETECTION TECHNOLOGIES AND EMERGENCY SHUTDOWN PROTOCOLS", Proceedings of the 2014 10th International Pipeline Conference, 2014. Zhang, Advances in Acoustic Emission Technology: Proceedings of the World Conference on Acoustic Emission–2013, Springer2014. Shehadeh, “Acoustic Emission Source Location for Steel Pipe and Pipeline Applications: The Role of Estimation of Arrival Time,” Proceedings of the Institution of Mechanical Engineers, Part E: Journal of Process Mechanical Engineering.
Maninder Pal, "Detecting and Locating Leaks in Polyethylene Pipes for Water Distribution", Proceedings of the World Engineering Congress II, 2010.
