Research On Non-destructive Testing Of Ferromagnetic Components With Weak Magnetic Imaging

Ferromagnetic components are used in various industrial fields,such as wire,pipeline and so on.The components result in fracture,corrosion,deformation and other defets because of detrimental conditions.This research aimed at realzing NDT and quantifying in a case of wire rope.This work could provide useful guidance and information for evaluating wire rope residual lifetime.In order to achieve this,the following aspects of theoretical research and engineering works are carried out in this dissertation.To overcome the disadvantages of traditional detection method,such as strong magnetic detection,eddy current method,ray method and ultrasonic guided wave,we proposed two weak magnetic detection method: one is RMF testing method which is based-on the RMF coercivity of ferromagnetic materials that ferromagnetic components would produce a weak and limited-time magnetic field after outer magnetic excitation.Through detecting and analyzing the residual magnetic field,the magnetic distribution could result in the surface and subsurface defects of wire rope.Because of low reliability of RMF method,we developed a new testing technology based-on UME method.The principle of the method is that the ferromagnetic materials would produce weak magnetic filed under an outer magnetic excitation.Magnetic characteristics of RMF and UME are analyzed based on the magnetic dipole model.The influence of defects parameters of MFL is analyzed and compared the MFL of RMF with UME.The results indicated that the signal of UME was higher than RMF.The UME signal was more reliably,higher SNR and difference than RMF through experimental comparison,and it provide more information between different broken wires.To restrain the system noise,a compressed sensing wavelet filtering was proposed to decrease the RMF system noise further.Aiming at the disadvantages of wavelet regarding an appropriate choice for the wavelet basis and uncertain analysis result,HHT is proposed for separate non-essential components.The consistency of array signal was promoted by the algorithm,and this filtering method is more general than wavelet in non-stationary signal.After the defect images are processed by sharpening target region and normalizing the center of defects.A multi-frame image superreconstruction method basing on Tikhonv regular principle was proposed,a high resolution defect image was obtained by this algorithm.The achieved recognition method was basing on artificial feature that includes texture,region description and invariant moment features.A BP and RBF neural network were separately used to classify broken-wire,and the limited error is one wire with accuracy rate of 91.43% through a trained RBF.