Study On MFL Signal Processing For Oil And Gas Pipeline And Evaluation Technique Of Pipeline Defects

With the enhancement of oil and gas transport status in a country's economy,more attention has been paid on safe running of pipelines. Pipeline in-line inspectiondevice based on MFL (Magnet Flux Leakage) is one of the most important tools. Theexisting evaluation technique on the defects of pipelines relies mainly on theinspectors' analysis on the MFL signal according to their experience, which leads tolarge workload, low efficiency and subjective results. The nonlinear characteristicbetween MFL signal and the defect geometric parameters of the pipeline, materialpermeability difference of the pipeline, the change of pipeline magnet structurecaused by the fluid pressure and the environment variation may result in direct orindirect influence on the MFL signals, which makes it more difficult to analyze thedefects. In this paper, we apply the magnetic dipole theory and finite element theorycombined with artificial neural network arithmetic. In this way, a defect evaluationtechnique which can overcome the former environmental influence and release thedependence of results on subjective factors is achieved. In addition, the pipelinedefect inspection device is optimized further to improve the total performance of theMFL inspection. The main researches in this paper are as follows:1.The wavelet de-nosing on MFL signal is studied and the image reconstruction and process technique is proposed, with which more verges and level information is preserved.2.The defect image transformation technique of grayscale to pseudo-color based on MFL is studied and improved equi-density pseudo-color coding arithmetic is proposed. According to the practical need, the non-linear pseudo-color coding arithmetic based on RGB tricolor, image pseudo-color coding arithmetic based on perceptive color space and improved equi-density pseudo-color coding arithmetic are integrated and preferably effect is obtained.3.The wavelet image compression theory is studied and the wavelet base function using for MFL image compression is designed, the wavelet coefficient threshold is calculated, the Jpeg table is modified and arithmetic coding algorithm is applied to accomplish the image compression with the VI software according to MFL image spectrum.4.By finite element theory, the defect model of oil and gas pipeline is created and the relationship between the defect geometric parameters and MFL signal and the IIrelationship sample database is set up.5.The center selection optimization algorithm of radial basis function neural network is proposed and the network model (RBFNN) predicting defect geometric parameters is designed. After the model is trained by the sample set created by the finite element simulation, the relationship between training target precision and the center set of radial function is studied and validated , by which the defect geometric parameters corresponding to MFL signal is predicted6.The wavelet basis function neural network (WBFNN) model predicting defect geometric parameters is designed, the improved ISODATA algorithm is used to classify the training sample set , to calculate the wavelet function center and to obtain the initial width of basis function, so the training speed and function approximating performance are improved.