Research On Detection And Identification Technology Of Seawall Defects In Oil Field

Because the coastal area where the onshore and offshore oil fields are located belongs to the strongly eroded coastal section,the marine environment is very complex and harsh,and the seawall is not only constantly affected by wind,wave and current,but also threatened by natural disasters such as storm surge and sea ice.At present,engineering quality problems and structural hazards exist to varying degrees in the seawalls in Shengli offshore oilfield.It is essential to discover and eliminate seawall quality hazards in time and provide a scientific and reasonable basis for the maintenance and reinforcement,which makes seawall defects detection and identification technology especially important.Focusing on the defects and hidden dangers of the main structure of the oil filed seawall,this thesis studies the application of integrated geophysical exploration technology in the detection of the seawall,and realizes the effective identification of the hidden problems of the structure through time-frequency analysis,inverse analysis,and defect anomaly feature identification technology research.This thesis has main done the following parts.According to the field situation of the oil filed seawall,the suitable defect detection method is determined and the optimization of the detection device of the high-density electric method is completed.According to the principle and characteristics of seismic image method and high density electrical method,the feasibility of using the combined geophysical exploration technology of seismic image method and high density electrical method to detect the defect of oil field seawall is analyzed.According to the geological simulation of the seawall site,the most suitable high-density electrical method detection device is determined.Aimming at the engineering practical application,the time domain information of seismic imaging method is mainly used,but the frequency domain information is seldom used,the timefrequency analysis method suitable for the seismic mapping method of oil field seawall is determined.The advantages and disadvantages of short-time fourier transform,wavelet transform and S transform and their application range are analyzed and compared.Three methods are compared and analyzed in the case of different frequency terrain deformation data.According to the characteristics of oil field seawall seismic imaging data,the S-transform is selected as the time-frequency analysis method,and the time-frequency diagram of single channel and whole data is obtained to determine the existence of defects,the information in time domain and frequency domain is comprehensively utilized.In view of the interference anomalies in the high-density electrical method data caused by the high resistance of the electrode itself,the existence of underground inhomogeneous body and geological noise,the data are preprocessed and then analyzed by inversion.Data preprocessing of high-density electrical method includes removing bad points,data splicing and terrain correction.Inversion analysis makes the imaging more intuitive,and can directly observe the electrical distribution of the underground medium,making the detection results more accurate.Aiming at the shortage of defect image samples of oil field seawall,we adopt an interdomain heterogeneous transfer learning method based on convolutional neural network.By pretraining the source domain data,the model can obtain excellent image feature extraction ability and good network parameters,and then transfer them to the target domain dataset to retrain the model.The model is trained in different layers by freezing technique to learn the features of the image data in the target domain.The comparative experimental results show that the transfer learning model from source domain to target domain can improve the recognition rate of small sample images and can show good feature extraction ability.