Research On Blind Source Separation And Identification Method Of Corrosion Acoustic Signal Of In-use Pipeline

The pipeline is an important part of the oil and gas transportation system,and the scale is developing towards large-scale and continuous development.Once the pipeline leaks,it often causes huge economic losses and environmental hazards.Corrosion is still the main reason for pipeline leakage.If corrosion is found in time in the early stage of pipeline corrosion and the location of corrosion is determined,remedial measures can be taken in time.Acoustic emission technology is a non-destructive testing technology suitable for on-line monitoring.In the service process of pipelines,when using acoustic emission on-line monitoring,the collected acoustic signals are complicated,the corrosion signals are mixed in the flowing signals,and the corrosion signals cannot be effectively separated.Cannot serve the purpose of timely warning.Therefore,in order to effectively separate the corrosion sound signal of the pipeline in use and accurately identify the corrosion phenomenon,a blind source separation algorithm based on time-frequency decomposition and four different identification algorithms are proposed,which have important theories for the acoustic monitoring of actual pipeline conditions.The main research content of this article is summarized as follows:(1)Summarize the current research status of corrosion sound signal processing methods and blind source separation technology at home and abroad,discuss in detail the types and causes of corrosion of pipelines,and explain the generation mechanism of corrosion sound emission sources and corrosion sound sources of pipelines in use.Based on the predicament of the existing pipeline corrosion acoustic monitoring,the experimental conditions,the existing technology and the experimental plan determine the feasibility and practicability of the thesis content,determine the research route of the thesis,and formulate the research content of the thesis;(2)Through the characteristic research of corrosion signal and pipeline flow signal,it is clear that the corrosion signal is a sudden acoustic signal;the pipeline flow is a continuous acoustic signal.According to the experimental conditions,the pipeline simulation acoustic signal acoustic detection experiment was designed.The lead-broken signal was used to simulate the pipeline corrosion signal,the sandpaper signal was used to simulate the pipeline flow signal,and the lead-broken sandpaper mixed signal was used to simulate the mixed signal of the corroded pipeline in use.The time-frequency domain characteristics of the three types of signals are explored separately.From the time-frequency domain results,it is feasible to use the three types of signals to simulate pipeline signals;(3)This paper proposes a blind source separation algorithm based on time-freque ncy decomposition.Time-frequency decomposition specifically includes: EMD time-freq uency decomposition,CEEMDAN time-frequency decomposition,VMD time-frequency decomposition,and the comparison of frequency domain diagrams after decomposition shows that EMD and CEEMDAN's Time-frequency decomposition has different degree s of modal aliasing,and VMD has a better decomposition effect.Finally,the VMD-Fa st ICA algorithm is selected to separate the mixed signal based on the kurtosis index,and the kurtosis greater than 3 is attributed to one type of remaining The results are attributed to another category,and finally achieved the goal of separation of the two t ypes of mixed signals;(4)In order to verify the effectiveness of the separation algorithm and the selection of effective feature parameters,different types of feature parameters are extracted based on the angle of time domain and entropy.In the time domain,in addition to the extraction of typical acoustic emission feature parameters,there are additional The characteristic parameters of peak-to-peak value,variance,mean,skewness,and margin factor are extracted;in the direction of entropy,there are extraction of time-frequency entropy based on EMD decomposition,wavelet entropy based on'Db8' wavelet decomposition and LMD decomposition Permutation entropy;(5)Based on the different feature parameters extracted,the SVM algorithm is introduced.In view of the influence of the selection of the core parameters of the SVM algorithm on the accuracy of the algorithm,two types of group optimization algorithm ideas are proposed to optimize the core parameters of the SVM,in order to study which type The recognition algorithm has more advantages in the identification of pipeline corrosion characteristic parameters.Comparing the two types of SVM optimization algorithms,weighted KNN algorithm,and Bagged Tree algorithm,it is finally found that the GWO-SVM algorithm has more advantages in recognition accuracy and operation speed.In the effective selection of parameters,the recognition rate of time-frequency entropy is higher,while the recognition rate of mean value and permutation entropy is poor.