Study On The Quantification Of Tubing Defects Based On Magnetic Flux Leakage Testing

Petroleum is very important as of strategic resources, and devotes a lot in developing international economy. Mechanical oil production is the main means, while tubing is the main equipment. During the process of production, tubing is constantly damaged; as a result, fatigues, corrosion, wear and fracture gradually come up. The damage will inevitably threat the safe production, and consequently a great loss of money. Tubing detection is important to find and evaluate the defects before it is too late. Magnetic flux leakage (MFL) for tubing is a nondestructive detection technique, which is pervasively used and does a great to ensure the safe production. The thesis surveys domestic and abroad MFL techniques for tubing, and studies carefully for tubing the MFL distribution, the MFL signals denoising, the defects localization and the MFL profile inversion. The thesis presents a technique to level the nondestructive detection automatically. The research is listed below.(1) General review of domestic and abroad literature. Broad reviews of the domestic and abroad literature in MFL detection techniques for oil tubing reveal the achievements and the present problems. Furthermore, MFL detection techniques in practice are surveyed on the spot. Based on the background, the procedure and the concept of the thesis are made up.(2) The study on MFL distribution for tubing defects. The finite element model is generated to model the typical types of tubing defects based on the research and generalization of the typical defects. By the model, a lot of finite element analysis is carried out on the different types and profile sizes of the defects. Consequently, the MFL distribution of the typical types of defects is obtained, and also the relation of the profile tubing defects and the MFL strength.(3) The study on denoising the MFL signals. MFL signals are apt to introduce noise. A new self-adaptive wavelet method based on particle swarm intelligent optimization to denoise the signals is proposed for threshold selection. Wavelet method is applied to denoise the MFL signals. The wavelet basis function and decomposition order are done by signal-to-noise ratio (SNR) and RMS error. Blind signal separation is also introduced in the method. The method lays a foundation to increase the SNR of MFL signals.(4) The classification model of inner-and-outer surface defects of tubing. The model, which is based on particle swarm intelligent optimization, is built on the SVM. The model avoids the influence of defects location to the quantitative evaluation. By the characteristics analysis of MFL signals, it extracts the axial and longitudinal composites of the signals in time and frequency domain. SVM is introduced into the position classification of tubing defects.(5) The recognition method for inner-and-outer surface defects. The method is based on the decision data fusion from multi-sensors. The method avoids the limits of single sensor detection, and thus improves the classification accuracy.(6) The direct inversion model for the irregular profiles of tubing defects based on multi-output SVM regression model. The currently adopted direct inversion method for tubing defects profiles is deficient. So, the thesis built such a model that introduces multi-output SVM regression model into the inversion model of tubing defects profiles.(7) The new recursive inversion method for defects profiles of tubing based on particle filter method. The method improves the quantitative evaluation accuracy of the tubing defects. It introduces the Bayesian theories. It carefully studied the influences of such observation model as the multiple regression models, the neutral network model and the SVM regression model. PSO optimization method is introduced to get more accuracy of the particle filter inversion. Consequently, it effectively decreases the particle degradation and impoverishment.In conclusion, the thesis presents a method to improve the detection and the quantitative evaluation of the MFL for tubing. The main work is on the MFL distribution of defects, the denoising method of MFL signals, the localization method of MFL signals, and the defects profiles inversion method. The method improves the detection accuracy MFL detection for tubing. Experiments identify the feasibility and effects of the method. As a result, the method technically provides a better accuracy in quantitative evaluation for tubing and an automatically leveling process. Hence, the presented method is important in both theory and practice.