Research On Pipeline Space Identification Method Based On Ultrasonic Wave-guiding Technology

As a special transportation method,pipeline transportation has its unique merits,such as the transportation volume,small land occupation,short construction cycle,and low cost.It plays a very important role in the national economy and social development.However,with the aging of the pipelines,pipeline accidents often occur,which may cause significant casualties and property loss.So,it is of great significance to quickly identify the spatial distribution and damage status of the pipeline networks,which can be beneficial for the routine maintenance and repairing for the damage parts before accidents occur.For the purpose of detecting the damage in the pipeline,it is necessary to know the geometric distribution of the pipe network.At present,various types of nondestructive testing methods are adopted in the pipeline systems for the safety monitoring,including the pipeline pig and optical fiber leakage detection methods.For those methods,none of them can determine the geometric characteristics of the pipe network and re-construct a spatial database of the pipe network.Compared with traditional ultrasonic technology,ultrasonic guided wave is more convenient,more sensitive and more economical,and has high applicability.Therefore,this paper proposed to re-construct the geometric spatial features based on exploring the propagation characteristics of ultrasonic guided waves in different pipeline paths and extracting the characteristics of guided waves in different complex pipeline networks.To obtain the sufficient database,extensive pipeline network spatial numerical model library was established numerically for determining the pipeline network spatial distribution.It is meaningful for the pipeline systems that do not have network map or been extended to the new pipelines.In this paper,the 3D pipeline model with different pipeline lengths and different node types is established.T(0,1)mode,torsional wave of order 1 in symmetric mode,guided waves with6 cycles and the frequency of 200 k Hz are excited for simulation,and the time-domain data is collected and analyzed.MATLAB software is used to analyze the data,extract the sensitive characteristics of time domain signal,and determine the change rule of pipeline ultrasonic echo signal under the complexity of path and joint form.In the numerical simulation,164 models were built.With the signals produced BP neural network algorithm,S＿Kohonen neural network algorithm,and PNN neural network algorithm were utilized to classify and recognize geometric characteristics of different types of pipelines,respectively.It was found that PNN neural network algorithm has higher recognition rate for geometric characteristics of pipelines than the other two algorithms.BP neural network algorithm and RBF neural network algorithm are used to predict the length of ultrasonic guided wave propagation path of different types of pipelines,and both of them have good prediction.Based on the propagation characteristics of ultrasonic guided waves in pipelines,a experimental strategy was proposed to verify the numerical model and to identify the geometric characteristics of ultrasonic guided waves in pipelines.Through the detection of different types of pipelines,the echo signals were processed by wavelet noise reduction,the relatively high sensitive features were selected by PCA method;and also,the geometric characteristics of pipelines were identified by PNN neural network algorithm,and the propagation path of ultrasonic guided waves is predicted by BP neural network algorithm.The feasibility of identifying geometrical characteristics of pipeline network by ultrasonic guided wave has been proven.The experimental results show that the recognition rate of PNN neural network algorithm for pipelines with different geometrical characteristics is more than 80%,and the average error of BP neural network algorithm for predicting the propagation path length of pipeline ultrasonic guided wave is 6.01%.