The Research Of Defect Magnetic Flux Leakage Testing Quantized Technology Based On BP Neural Network

In port, oil depot and petrochemical industry, tank is the special equipment of storing raw oil, liquid chemical material, and chemical products, the safety issue of tank is a significant event. Corrosion is one of the most important inducement of tank incident, the bottom of tank is the most susceptible to corrode, so the periodic detection of tank bottom is very important. Magnetic flux leakage testing is a common NDT method of tank bottom, this paper is based on magnetic flux leakage signals of tank bottom corrosion defects, the quantized problem of it was deeply studied.Two theoretical calculational methods of leakage magnetic field was contrasted and analysised based on magnetic flux leakage testing principle, then 3-D finite element analysis of corrosion defects leakage magnetic field was done by ANSYS software, simulation magnetic flux leakage signals of corrosion edfect was got. Considering the feature that corrosion defects leakage magnetic field vertical component can resist the influence of magnetic pole, in this paper, the vertical component of defect leakage magnetic signal was studied, and domain analysis, frequency domain analysis and time-frequency analysis of it was done, some signal analysis methods such as entropy spectral analysis and wavelet analysis were used in the link of magnetic flux leakage signal processing, waveform characteristic quantity of signal was extracted. Some factors which affect detection result such as corrosion products, detection speed and the planeness of detected plate etc. was analysised, the basis of improving defect detection rate was provided. Leakage magnetic field of corrosion defects with different shapes and different sizes was simulated by finite element software, qualitative analysis and quantitative analysis of relationship between defect shape, size and its leakage magnetic field was done, a series of relation curves between defects shapes, defects geometric parameters and defects leakage magnetic field was got, the theoretical basis of defect identification and defect quantifying was provided. Leakage magnetic field of corrosion defects with different shapes and different sizes was obtained through experimental study and finite element simulation, characteristic quantity of signal with differnet sensors was extracted, defect samples was established. Neural networks which were used to identify defect types and quantify defect sizes were established through defect samples, the mapping from defect leakage magnetic signal to defects geometric size was effected.Finally, the feasibility of quantitative method was tested via experimental study, the experimental results indicate that the method which was raised in this paper can quantify defect size exactly.