Research Intelligent Identification Of Defect Profile In The System Of Magnetic Flux Leakage Detecting

The periodic inspection of pipeline is one of the most important means to assure the safe operation of the gas pipeline network. In the past, the pipeline defect inspection was done based on the peculiarity of MFL signal mostly by the technicians who was experience in it, so the result can be artificially impacted by the technician greatly, and it also will take a long time to do it. Furthermore, the technique in pipeline inspection only can estimate the degree of pipeline defect before, but can not describe the profile and the figure of defect at detail. For above, we need a new approach applied in processing pipeline MFL signal to solve the problerms.In the paper, the basic theory of MFL detecting together with the structure and the working procedure of pipeline MFL detecting instrument are introduced. The factors including defect geometry parameters, defect surface and shape influence the MFL signal characteristics. The character of the different surface's and shape's pipeline MFL signal is achieved by using the finite element analysis software (ANSYS), which can manage to gain the data of simulated pipeline MFL signal. At the same time, the testing data is acquired by testing the defect made out on iron plate with the pipeline MFL detecting instrument, which can gain the factual date.The technique of neural network has the strong point and character to done the problerms, so it was employed, neural networks are employed for defect accurate recognition and calculation. The traditional BP Neural Network and Wavelet Basis Function Neural Networks can successfully predict or estimate defect shape and geometry parameters. The experiment, which the BP Neural Network is trained to identify defect shape and geometry, is done mainly in tow aspects. On the one hand, the BP Neural Network is trained by the data of simulated pipeline MFL signal. On the other hand, the BP Neural Network is trained by the factual data. As result, the latter have steady output and litter error.The multi-sensor pipeline MFL signal is gained the boundary by Canny algorithm, which can achieve the outline of defect. The quantity of sensor can influence the boundary's impact. The more the quantity of sensor is much, the more result is good. Studying Data Fusion is applied to disposal MFL signal in the paper.