Metadata Downloads
The health monitoring of pipelines has become extremely essential since pipeline networks were widely constructed. The primary aim of this process is to pinpoint any abnormality early in a pipeline to require maintenance or replacement as soon as possible because a defective pipe that has not been repaired would cause serious consequences for nearby nature and human. Pipeline failures can mainly be classified into a range of types including leaks, cracks, and corrosions. Cracks and corrosions, alongside their other effects, the easiest possibility would be leakage if a crack/corrosion grew to some extent to rupture a pipe. Hence, all the faults should immediately be identified once they occur.
Acoustic emission had long ago been utilized as the effective means to inspect materials and structures. This phenomenon occurs within a material from localized sources once the material is put stress and strain on. Acoustic emission is extremely sensitive to structural irregularities; therefore, AE signal–based techniques have been employing in recent studies of pipeline fault diagnostics. However, because the content of AE signals is extraordinarily complicated, which is resulting from the effects of wave attenuation and dispersion as well as the interference of ambient noises, the algorithms that scientists and researchers developed may not be reliable enough to extract accurate information about a defect from the signals. In this dissertation, what need to be improved in the pipeline fault diagnosis using AE signals will be considered, in which the identification of leak and crack is concentrated.
First, the thesis introduces a reliable method to detect a leak in a water pipeline based on AE wave attenuation. This approach segmented AE signals into short frames, calculated intermediate quantities that contain the symptoms of a leak and keeps its characteristic adequately stable even when the environmental conditions change, and trained a k–NN classifier using features extracted from the transformed signals to identify a leak in the pipeline. Experiments were conducted under different conditions to confirm the effectiveness of that method. The experimental results indicate that it offers better quality and more reliability than using features extracted directly from the AE signals to train the k–NN classifier. Moreover, the proposed method required less training data than existing techniques. The transformation method was highly accurate and worked well even when only a small amount of data was used to train the classifier, whereas the direct AE–based method returned misclassifications in some cases. In addition, the robustness of the proposed method was also tested by adding Gaussian noise to the AE signals. The proposed method was more resistant to noise than the direct AE–based method.
Leak localization is as important as leak detection in the general pipeline fault diagnosis. As a result, this dissertation proposes a method of leak localization using AE bursts, which is aimed particularly at industrial–fluid pipelines made from steel. The proposed approach exploited a burst phenomenon in AE signals and combined signal processing with a physical wave–propagation model in order to improve the leak localization. The algorithm sought AE bursts based on a detection theorem and then associated neighboring bursts into unique burst groups. Filtered AE events which are pairs of grouped bursts from two signal channels in turn allowed the extraction of precise location where a leak came about. The resulting localization method yielded a mean error of approximately 2.5% of the distance between two sensors, while this parameter returned by conventional approaches was greater than 10%. The combination of grouping and filtering in the methodology elucidates event concentration and reduces error.
Another technique is also offered for detecting a leak in a gas pipeline using a k–NN classifier and hybrid AE features in the thesis. This whole algorithm was embedded in an MCU to achieve a complete real–time pipeline leak detection system. First, AE signals were first recorded from a gas pipeline testbed under various conditions and offline investigated to synthesize the leak detection algorithm. The approach explored different features of normal/leaking states from corresponding datasets and eliminated inferior features to enhance the performance of leak detection. In order to obtain the robustness, alongside features were normalized, the trained k–NN classifier was adapted to the specific environment where the system was installed. Furthermore, the system decided the state of the pipeline on ALEOR and a defined threshold to reduce false alarms. The entire proposed system was implemented on the 32F746G–DISCOVERY board, and to verify this system, numerous real AE signals stored in a hard drive were transferred to the board. In the experiments, the implemented system executed the leak detection algorithm in a period shorter than the total input data time, thus guaranteeing the real–time characteristic. Additionally, the system yielded high ACA despite adding a white noise to input signal, and false alarms did not occur with a reasonable ALEOR threshold.
Finally, a novel approach is presented in the dissertation to detect and localize a crack in a pipeline transporting fluid under high pressure using AE signals. From signals acquired by two R15i-AST sensors at two ends of a fluid pipeline, the proposed method scanned peaks in the individual signal channels in the time–frequency domain and filtered out noise to obtain AE events. Subsequently, adjacent events were combined into grouped events, and these were picked and paired together on two sensor channels to localize emission sources using the TDOA technique. To improve the location accuracy, the mechanism only determined TOA of Rayleigh waves with a similar frequency in event pairs. Furthermore, the Rayleigh wave velocity was found by a PLB procedure. Additionally, false emission sources were eliminated by considering the wave energy attenuation characteristics in their propagation path. After locating the emission sources, the approach observed their distribution according to the position and time of occurrence. The variation in AE activity against applied load, which was established by counting the returned sources, could indicate irregular structural changes in a material. The location of the structural change could be surmised by the emission source distribution and density according to the position along the pipeline. Experimental results showed that the proposed method correctly diagnosed faults in the considered pipeline from AE signals, whereas a conventional approach (performed by detecting hits with a threshold) inaccurately localized AE sources and imprecisely exposed signs of abnormal structural transformation.
부이 꾸이 탕
Issued Date
Awarded Date
Acoustic emissioncrack detectionleak detectionpipeline health monitoring
Alternative Author(s)
일반대학원 전기전자컴퓨터공학과
울산대학교 일반대학원 전기전자컴퓨터공학과
울산대학교 논문은 저작권에 의해 보호 받습니다.
Appears in Collections:
Computer Engineering & Information Technology > 2. Theses (Ph.D)
Authorize & License
  • Authorize공개
Files in This Item:

Items in Repository are protected by copyright, with all rights reserved, unless otherwise indicated.