Research On Damage Identification Algorithm Based On Pipeline Detection Signal

With the development of society,pipeline transportation has become one of the indispensable means of transportation in production and life,with good security,high efficiency,economic benefits and other advantages.In order to ensure the safe transportation of pipeline,the outer coating of anticorrosion layer is often used to protect the pipeline.Due to the harsh working environment,long service time and soil corrosion and other refractory factors,the external anticorrosion coating often produces cracks,holes and stripping defects,resulting in pipeline damage and leakage or even explosion,resulting in serious economic losses and casualties.Therefore,it is of great engineering value and theoretical significance to identify the damage of pipeline anticorrosion coating structure.In this paper,a pipeline damage identification method based on the combination of Ensemble Empirical Mode Decomposition(EEMD),Fast Fourier Transform(FFT),Isolation Forest(IForest),Miss Forest(MF),Kernel Principal Component Analysis(KPCA)and e Xtreme Gradient Boosting(XGBoost)classification regression algorithm is proposed by comprehensively analyzing the pipeline damage data and the advantages of each algorithm.The main research contents of this paper are as follows:1.Build pipeline damage database.The pipeline damage database is composed of six-dimensional dynamic response frequency domain features.Ultrasonic wave is stimulated by the sending probe of ultrasonic guided wave and the pipeline damage echo signal is acquired by the receiving probe.The acquired pipeline damage signal is de-noised and decomposed by Ensemble Empirical Mode Decomposition.The frequency domain features of pipeline damage signals were extracted by Fast Fourier Transform and the pipeline damage database was constructed.2.Isolation Forest,Miss Forest,Kernel Principal Component Analysis and e Xtreme Gradient Boosting were applied to pipeline damage identification.For the constructed pipeline damage data,outlier detection of Isolation Forest,data completion of Miss Forest and feature dimension reduction optimization of Kernel Principal Component Analysis were conducted successively.EXtreme Gradient Boosting classification algorithm is used to identify the damage types of pipeline anticorrosion layer,and e Xtreme Gradient Boosting regression algorithm is used to identify the damage size of pipeline anticorrosion layer.3.The pipeline damage identification method based on the combination of EEMD,FFT,IForest,MF,KPCA and XGBoost classification and regression algorithm is used to detect the damage of the double-layer structure of the pipeline anticorrosion coating.The results show that the classification algorithm can effectively identify the three damage types of the pipeline anticorrosion layer structure,namely,the crack,the hole and the peel,and the recognition accuracy reaches 90.9%.The regression algorithm can recognize the crack length,the hole radius and the peel length of the pipeline anticorrosion layer structure.Therefore,the pipeline damage identification method based on EEMD,FFT,IForest,MF,KPCA and XGBoost established in this paper has a certain reference value in the field of pipeline health detection and damage assessment.