Research On Defect Recognition Method Of Pipeline Leakage Detection In Pipeline Based On Artificial Intelligence

The main method of oil and gas energy transportation is pipeline transportation.The damage of pipelines is mainly caused by the destruction of pipeline ferromagnetic materials,including corrosion,perforation,leakage,and pipe burst.It will not only cause major direct or indirect economic losses,but also cause serious environmental pollution.The internal magnetic flux leakage detection method is aimed at the most effective detection method for oil and gas pipelines.At present,the assessment of pipeline defects mainly relies on the analysis and interpretation of internal magnetic flux leakage inspection data.Oil and gas pipeline pipelines are thousands of kilometers away,and the amount of data that is detected by internal magnetic flux leakage is too large,currently,based on the image data visualized by the magnetic flux leakage detection data,the pipe character is marked by the method of manual interpretation,the defect identification and judgment are completed,and the excavation report of the pipeline defect report is given.Aiming at the problem of identifying flaws in pipeline magnetic flux leakage detection,a method of artificial intelligence based on deep learning theory was proposed,and the artificial intelligence method identifies the detection defect in the pipeline magnetic flux leakage.This paper mainly studies the theoretical methods and applications of artificial intelligence;the basic concepts of deep learning and neural network training methods for different data;by constructing a work map model of TensorFlow,a relatively advanced deep learning platform,a platform for the implementation of a deep neural network is built;the basic formats and types of pipeline magnetic flux leakage inspection data are studied.The pipeline magnetic flux leakage is used to detect the image data of different pipeline characteristics and defects,and then the pipeline magnetic flux leakage internal inspection image data set is established.the basic realization principle of convolutional neural network and the optimization method of the model are studied.Firstly,the convolutional neural network is used to map the input image data through the design of the convolution kernel.This method breaks through the parameters of the matrix multiplication of the traditional neural network.The training method realizes the sparse connection of parameters in the neural network and adopts a parameter sharing training method.In this paper,the pipeline magnetic flux leakage detection data set is taken as the training object,and deep convolutional neural network is used as the algorithm tool.Based on the deep learning framework TensorFlow,a 7-layer deep convolutional neural network is built,and the pipeline magnetic flux leakage internal inspection image data set is completed.The training identification process uses the cross-entropy loss function as the goal optimization function of the network training,and adds a regularization parameter penalty term to the objective function to achieve optimization of the convolutional neural network model.The parameter updating method in the loss function term in the cross entropy loss function is to use the back propagation algorithm to continuously calculate the gradient of the weight connection matrix of the neural network to complete the updating of the network parameters.The neural network convergence is based on the expectation of the neural network through the gradient in the stochastic gradient descent algorithm.Then the small-scale sample approximation of the detection data set in the pipeline magnetic flux leakage is performed.Finally,the neural network achieves the purpose of convergence.Finally,experiments have shown that the trained input convolutional neural network model is used to predict the new input pipeline magnetic flux leakage image data,and the recognition rate is about 90%.