Pipeline Leak Detection And Localization Based On Feature Extraction And Information Fusion

Leak accidents in long transportation pipleine occur frequently,which are influenced by natural corrosion,geographical environment and third party damage.As the important part and one of the key technologies in safe operation of pipeline,pipeline leak detection and location has an important and relistic role in meeting these objectives.With the rapid development of sensor technology,signal processing and analysis technology and wide application of digital pipeline,a large number of pipeline process data can be well measured and stored.How to extract useful information from a large number of measured data for leakage detection,it has become an urgent problem to be solved in the pipeline leak detection and localization.On the basis of the existing research,this paper proposes several leak detection and location mehods,which are mainly focus on feature extraction and information fusion of measure data,the main work is as follow:1.In oil pipeline leak detection and location,noise in the pressure signal collected at the end of the pipeline affects the accuracy of leak detection and the error of leakage location.To reduce the noise interference,an improved local mean decomposition?LMD?signal analysis method is proposed.The production functions?PFs?that are related to the leak signal can be exacted,and it is unnecessary to know the characteristics of leak signals or noise in advance.According to the cross-correlation function,there is a significant peak between the measured signals which are decomposed into a number of PFs.These reconstructed principles PF components are obtained,and a wavelet analysis is used to remove the noise in the reconstructed signal.On this basis,the signal features are extracted according to the time-domain feature and waveform feature,which are input into least squares twin support vector machine?LSTSVM?,LSTSVM to recognize pipeline leaks.According to the reconstructed signal after wavelet de-noising,the time delay estimate of the negative pressure signal at the end of the pipeline is obtained by the cross-correlation function,and the leak location is ultimately calculated by combining the time delay with the leak signal propagation velocity.A flow model for pipeline leakage is proposed based on Flowmaster software,where the collected data of the different working conditions is processed.The experimental results show that the proposed method can effectively identify different working conditions and accurately locate the leakage point.2.For data imbalance of pipeline working conditions will cause the pipeline leakage diagnosis accuracy decreased,a method of pipeline leak detection and location based on imbalance data is proposed.First,the imbalance data of different working conditions are processed by K-means clustering algorithm and under-sampled to achieve the balance data.Then the Fischer-Burmeister function is introduced into the learning process of the twin support vector machine?TWSVM?,in order to avoid the matrix inversion calculation,and the balance data are input into the improved TWSVM to distinguish the pipeline leakage.Leak location is obtained by the cross-correlation function method.A flow model of pipeline is put forward based on Flowmaster software,the proposed method is used to identify pipeline leakage.The experimental results show that the proposed method is more effective than the classical TWSVM and the Lagrange TWSVM to identify the pipeline leakage aperture and location.3.The leakage aperture cannot be easily identified,when an oil pipeline has small leaks.To address this issue,a leak aperture recognition method based on wavelet packet analysis?WPA?and a deep belief network?DBN?with independent component regression?ICR?is proposed.WPA is used to remove the noise in the collected sound velocity of the ultrasonic signal.Next,the de-noised sound velocity of the ultrasonic signal is input into the deep belief network with independent component regression(DBN_ICR)to recognize different leak apertures.However,the optimization of the weights of the DBN with the gradient leads to a local optimum and a slow learning rate,ICR is used to replace the gradient fine-tuning method in conventional DBN for improving the classification accuracy,and a Lyapunov function is constructed to prove the convergence of the DBN_ICR learning process.By analyzing the acquired ultrasonic sound velocity of different leak apertures,the results show that the proposed method can quickly and effectively identify different leakage apertures.4.We propose a branch pipeline leakage localization based on cyber-physical system?CPS?architecture to solve the problem that the data between multiple pipeline leakage monitoring systems cannot be shared and be located at the same time.Firstly,the singular point of pressure signals at the ends of multibranch pipeline is analyzed by wavelet packet analysis,so that the time feature samples could be established.Then,the samples are input into the improved twin support vector machine?ITWSVM?to distinguish the pipeline leakage localization.The simulation results show that the proposed method is more effective than the twin support vector machine,the radial basis function?RBF?neural networks and the back propagation?BP?neural networks.5.Leakage detection system often occurs the problem of leakage alarm,when an oil pipeline has small leaks.Based on the neural network and the D-S?Dempster-Shafer?evidential theory,a method was proposed for identifying small leaks.Firstly,the ultrasonic wave speed,flow,temperature signals were decomposed into 3 levels by the wavelet packet transform,the effective value of signals in various frequency bands were extracted and eigenvectors were constructed.Furthermore,the eigenvectors were put into the neural network and trained to realize the small leaks identification.Taking the preliminary identification as the independent evidence,a final identification was obtained according to the D-S evidential fusion algorithm;The change of the sound velocity of the ultrasonic signal cannot be easily detected,when an oil pipeline has small leaks.Based on the method of characteristics?MOC?theory,the leakage occurrence mechanism of pipeline and the propagation mechanism of the sound velocity of the ultrasonic signal along the pipeline were studied.Moreover,the computational models of generating the sound velocity of the ultrasonic signal,which are induced by pipeline leakage and calculated the propagation attenuation of the ultrasonic wave along the pipeline,were established.A small leakage localization method of pipeline based on information fusion is proposed,which is a combination of the sound velocity of the ultrasonic and flow rate signals.First,the time delay is estimated by modified cross-correlation analysis between the fusion signals at the ends of the pipeline;then,the leak position is calculated according to the information of pipeline length and negative pressure wave propagation velocity.Because the change of the information fusion signal induced by a small leak is larger than the change of sound velocity of the ultrasonic signal and the pressure signal,the proposed method can be used to locate the small leak in a pipeline.By analyzing the acquired data of the small leak in experimental pipeline,the results show that the proposed method can quickly and effectively locate the small leaks.